Archive for the ‘Medical Imaging Technology, Image Processing/Computing, MRI, CT, Nuclear Medicine, Ultra Sound’ Category

Two brothers with MEPAN Syndrome: A Rare Genetic Disorder

Reporter: Amandeep Kaur

In the early 40s, a married couple named Danny and Nikki, had normal pregnancy and delivered their first child in October 2011.  The couple was elated after the birth of Carson because they were uncertain about even conceiving a baby. Soon after birth, the parents started facing difficulty in feeding the newborn and had some wakeful nights, which they used to called “witching hours”. For initial six months, they were clueless that something was not correct with their infant. Shortly, they found issues in moving ability, sitting, and crawling with Carson. Their next half year went in visiting several behavioral specialists and pediatricians with no conclusion other than a suggestion that there is nothing to panic as children grow at different rates.

Later in early 2013, Caron was detected with cerebral palsy in a local regional center. The diagnosis was based on his disability to talk and delay in motor development. At the same time, Carson had his first MRI which showed no negative results. The parents convinced themselves that their child condition would be solved by therapies and thus started physical and occupational therapies. After two years, the couple gave birth to another boy child named Chase in 2013. Initially, there was nothing wrong with Chase as well. But after nine months, Chase was found to possess the same symptoms of delaying in motor development as his elder brother. It was expected that Chase may also be suffering from cerebral palsy. For around one year both boys went through enormous diagnostic tests starting from karyotyping, metabolic screen tests to diagnostic tests for Fragile X syndrome, lysosomal storage disorders, Friedreich ataxia and spinocerebellar ataxia. Gene panel tests for mitochondrial DNA and Oxidative phosphorylation (OXPHOS) deficiencies were also performed. No conclusion was drawn because each diagnostic test showed the negative results.

Over the years, the condition of boys was deteriorating as their movements became stiffer and ataxic, they were not able to crawl anymore. By the end of 2015, the boys had an MRI which showed some symmetric anomalies in their basal ganglia indicating a metabolic condition. The symptoms of Carson and Chase was not even explained by whole exome sequencing due to the absence of any positive result. The grievous journey of visits to neurologist, diagnostic tests and inconclusive results led the parents to rethink about anything happened erroneous due to them such as due to their lifestyle, insufficient intake of vitamins during pregnancy or exposure to toxic agents which left their sons in that situation.

During the diagnostic odyssey, Danny spent many restless and sleepless nights in searching PubMed for any recent cases with symptoms similar to his sons and eventually came across the NIH’s Undiagnosed Diseases Network (UDN), which gave a light of hope to the demoralized family. As soon as Danny discovered about the NIH’s Diseases Network, he gathered all the medical documents of both his sons and submitted the application. The submitted application in late 2015 got accepted a year later in December 2016 and they got their first appointment in early 2017 at the UDN site at Stanford. At Stanford, the boys had gone through whole-genome sequencing and some series of examinations which came back with inconclusive results. Finally, in February 2018, the family received some conclusive results which explained that the two boys suffer from MEPAN syndrome with pathogenic mutations in MECR gene.

  • MEPAN means Mitochondrial Enoyl CoA reductase Protein-Associated Neurodegeneration
  • MEPAN syndrome is a rare genetic neurological disorder
  • MEPAN syndrome is associated with symptoms of ataxia, optic atrophy and dystonia
  • The wild-type MECR gene encodes a mitochondrial protein which is involved in metabolic processes
  • The prevalence rate of MEPAN syndrome is 1 in 1 million
  • Currently, there are 17 patients of MEPAN syndrome worldwide

The symptoms of Carson and Chase of an early onset of motor development with no appropriate biomarkers and T-2 hyperintensity in the basal ganglia were matching with the seven known MEPAN patient at that time. The agonizing journey of five years concluded with diagnosis of rare genetic disorder.

Despite the advances in genetic testing and their low-cost, there are many families which still suffer and left undiagnostic for long years. To shorten the diagnostic journey of undiagnosed patients, the whole-exome and whole-genome sequencing can be used as a primary tool. There is need of more research to find appropriate treatments of genetic disorders and therapies to reduce the suffering of the patients and families. It is necessary to fill the gap between the researchers and clinicians to stimulate the development in diagnosis, treatment and drug development for rare genetic disorders.

The family started a foundation named “MEPAN Foundation” (https://www.mepan. org) to reach out to the world to educate people about the mutation in MECR gene. By creating awareness among the communities, clinicians, and researchers worldwide, the patients having rare genetic disorder can come closer and share their information to improve their condition and quality of life.

Reference: Danny Miller, The diagnostic odyssey: our family’s story, The American Journal of Human Genetics, Volume 108, Issue 2, 2021, Pages 217-218, ISSN 0002-9297, https://doi.org/10.1016/j.ajhg.2021.01.003 (https://www.sciencedirect.com/science/article/pii/S0002929721000033)




https://www.mepan. org

Other related articles were published in this Open Access Online Scientific Journal, including the following:

Effect of mitochondrial stress on epigenetic modifiers

Larry H. Bernstein, MD, FCAP, Curator, LPBI


The Three Parent Technique to Avoid Mitochondrial Disease in Embryo

Reporter and Curator: Dr. Sudipta Saha, Ph.D.


New Insights into mtDNA, mitochondrial proteins, aging, and metabolic control

Larry H. Bernstein, MD, FCAP, Curator, LPBI


Mitochondrial Isocitrate Dehydrogenase and Variants

Writer and Curator: Larry H. Bernstein, MD, FCAP


Update on mitochondrial function, respiration, and associated disorders

Larry H. Benstein, MD, FCAP, Gurator and writer


Read Full Post »

Happy 80th Birthday: Radioiodine (RAI) Theranostics: Collaboration between Physics and Medicine, the Utilization of Radionuclides to Diagnose and Treat: Radiation Dosimetry by Discoverer Dr. Saul Hertz, the early history of RAI in diagnosing and treating Thyroid diseases and Theranostics


Guest Author: Barbara Hertz



Celebrating eighty years of radionuclide therapy and the work of Saul Hertz

First published: 03 February 2021

Both authors contributed to the development, drafting and final editing of this manuscript and are responsible for its content.


March 2021 will mark the eightieth anniversary of targeted radionuclide therapy, recognizing the first use of radioactive iodine to treat thyroid disease by Dr. Saul Hertz on March 31, 1941. The breakthrough of Dr. Hertz and collaborator physicist Arthur Roberts was made possible by rapid developments in the fields of physics and medicine in the early twentieth century. Although diseases of the thyroid gland had been described for centuries, the role of iodine in thyroid physiology had been elucidated only in the prior few decades. After the discovery of radioactivity by Henri Becquerel in 1897, rapid advancements in the field, including artificial production of radioactive isotopes, were made in the subsequent decades. Finally, the diagnostic and therapeutic use of radioactive iodine was based on the tracer principal that was developed by George de Hevesy. In the context of these advancements, Hertz was able to conceive the potential of using of radioactive iodine to treat thyroid diseases. Working with Dr. Roberts, he obtained the experimental data and implemented it in the clinical setting. Radioiodine therapy continues to be a mainstay of therapy for hyperthyroidism and thyroid cancer. However, Hertz struggled to gain recognition for his accomplishments and to continue his work and, with his early death in 1950, his contributions have often been overlooked until recently. The work of Hertz and others provided a foundation for the introduction of other radionuclide therapies and for the development of the concept of theranostics.











Dr. Saul Hertz was Director of The Massachusetts General Hospital’s Thyroid Unit, when he heard about the development of artificial radioactivity. He conceived and brought from bench to bedside the successful use of radioiodine (RAI) to diagnose and treat thyroid diseases. Thus was born the science of theragnostics used today for neuroendocrine tumors and prostate cancer. Dr. Hertz’s work set the foundation of targeted precision medicine.

Keywords: Dr. Saul Hertz, nuclear medicine, radioiodine


How to cite this article:
Hertz B. A tribute to Dr. Saul Hertz: The discovery of the medical uses of radioiodine. World J Nucl Med 2019;18:8-12


How to cite this URL:
Hertz B. A tribute to Dr. Saul Hertz: The discovery of the medical uses of radioiodine. World J Nucl Med [serial online] 2019 [cited 2021 Mar 2];18:8-12. Available from: http://www.wjnm.org/text.asp?2019/18/1/8/250309



  • Dr Saul Hertz (1905-1950) discovers the medical uses of radioiodine

Barbara Hertz, Pushan Bharadwaj, Bennett Greenspan»

Abstract » PDF» doi: 10.24911/PJNMed.175-1582813482





  • Happy 80th Birthday: Radioiodine (RAI) Theranostics

Thyroid practitioners and patients are acutely aware of the enormous benefit nuclear medicine has made to mankind. This month we celebrate the 80th anniversary of the early use of radioiodine(RAI).

Dr. Saul Hertz predicted that radionuclides “…would hold the key to the larger problem of cancer in general,” and may just be the best hope for diagnosing and treating cancer successfully.  Yes, RAI has been used for decades to diagnose and treat disease.  Today’s “theranostics,” a term that is a combination of “therapy” and “diagnosis” is utilized in the treatment of thyroid disease and cancer. 

            This short note is to celebrate Dr. Saul Hertz who conceived and brought from bench to bedside the medical uses of RAI; then in the form of 25 minute iodine-128.  

On March 31st 1941, Massachusetts General Hospital’s Dr. Saul Hertz (1905-1950) administered the first therapeutic use of Massachusetts Institute of Technology (MIT) cyclotron produced RAI.  This landmark case was the first in Hertz’s clinical studies conducted with MIT, physicist Arthur Roberts, Ph.D.

[Photo – Courtesy of Dr Saul Hertz Archives ]

Dr Saul Hertz demonstrating RAI Uptake Testing

            Dr. Hertz’s research and successful utilization of radionuclides to diagnose and treat diseases and conditions, established the use of radiation dosimetry and the collaboration between physics and medicine and other significant practices.   Sadly, Saul Hertz (a WWII veteran) died at a very young age.  


About Dr. Saul Hertz

Dr. Saul Hertz (1905 – 1950) discovered the medical uses of radionuclides.  His breakthrough work with radioactive iodine (RAI) created a dynamic paradigym change integrating the sciences.  Radioactive iodine (RAI) is the first and Gold Standard of targeted cancer therapies.  Saul Hertz’s research documents Hertz as the first and foremost person to conceive and develop the experimental data on RAI and apply it in the clinical setting.

Dr. Hertz was born to Orthodox Jewish immigrant parents in Cleveland, Ohio on April 20, 1905. He received his A.B. from the University of Michigan in 1925 with Phi Beta Kappa honors. He graduated from Harvard Medical School in 1929 at a time of quotas for outsiders. He fulfilled his internship and residency at Mt. Sinai Hospital in Cleveland. He came back to Boston in 1931 as a volunteer to join The Massachusetts General Hospital serving as the Chief of the Thyroid Unit from 1931 – 1943.

Two years after the discovery of artifically radioactivity, on November 12, 1936 Dr. Karl Compton, president of the Massachusetts Institute of Technology (MIT), spoke at Harvard Medical School.  President Compton’s topic was What Physics can do for Biology and Medicine. After the presentation Dr. Hertz spontaneously asked Dr. Compton this seminal question, “Could iodine be made radioactive artificially?” Dr. Compton responded in writing on December 15, 1936 that in fact “iodine can be made artificially radioactive.”

Shortly thereafter, a collaboration between Dr. Hertz (MGH) and Dr. Arthur Roberts, a physicist of MIT, was established. In late 1937, Hertz and Roberts created and produced animal studies  involving 48 rabbits that demonstrated that the normal thyroid gland concentrated Iodine 128 (non cyclotron produced), and the hyperplastic thyroid gland took up even more Iodine.  This was a GIANT step for Nuclear Medicine.

In early 1941, Dr. Hertz administer the first therapeutic treatment of MIT Markle Cyclotron produced radioactive iodine (RAI) at the Massachusetts General Hospital.  This  led to the first series of twenty-nine patients with hyperthyroidism being treated successfully with RAI. ( see “Research” RADIOACTIVE IODINE IN THE STUDY OF THYROID PHYSIOLOGY VII The use of Radioactive Iodine Therapy in Hyperthyroidism, Saul Hertz and Arthur Roberts, JAMA Vol. 31 Number 2).

In 1937, at the time of the rabbit studies Dr Hertz conceived of RAI in therapeutic treatment of thyroid carsonoma.  In 1942 Dr Hertz gave clinical trials of RAI to patients with thyroid carcinoma.

After serving in the Navy during World War II, Dr. Hertz wrote to the director of the Mass General Hospital in Boston, Dr. Paxon on March 12, 1946, “it is a coincidence that my new research project is in Cancer of the Thyroid, which I believe holds the key to the larger problem of cancer in general.”

Dr. Hertz established the Radioactive Isotope Research Institute, in September, 1946 with a major focus on the use of fission products for the treatment of thyroid cancer, goiter, and other malignant tumors. Dr Samuel Seidlin was the Associate Director and managed the New York City facilities. Hertz also researched the influence of hormones on cancer.

Dr. Hertz’s use of radioactive iodine as a tracer in the diagnostic process, as a treatment for Graves’ disease and in the treatment of cancer of the thyroid remain preferred practices. Saul Hertz is the Father of Theranostics.

Saul Hertz passed at 45 years old from a sudden death heart attack as documented by an autopsy. He leaves an enduring legacy impacting countless generations of patients, numerous institutions worldwide and setting the cornerstone for the field of Nuclear Medicine. A cancer survivor emailed, The cure delivered on the wings of prayer was Dr Saul Hertz’s discovery, the miracle of radioactive iodine. Few can equal such a powerful and precious gift. 

To read and hear more about Dr. Hertz and the early history of RAI in diagnosing and treating thyroid diseases and theranostics see –



   References in https://www.wjnm.org/article.asp?issn=1450-1147;year=2019;volume=18;issue=1;spage=8;epage=12;aulast=Hertz




Hertz S, Roberts A. Radioactive iodine in the study of thyroid physiology. VII The use of radioactive iodine therapy in hyperthyroidism. J Am Med Assoc 1946;131:81-6.  Back to cited text no. 1
Hertz S. A plan for analysis of the biologic factors involved in experimental carcinogenesis of the thyroid by means of radioactive isotopes. Bull New Engl Med Cent 1946;8:220-4.  Back to cited text no. 2
Thrall J. The Story of Saul Hertz, Radioiodine and the Origins of Nuclear Medicine. Available from: http://www.youtube.com/watch?v=34Qhm8CeMuc. [Last accessed on 2018 Dec 01].  Back to cited text no. 3
Braverman L. 131 Iodine Therapy: A Brief History. Available from: http://www.am2016.aace.com/presentations/friday/F12/hertz_braverman.pdf. [Last accessed on 2018 Dec 01].  Back to cited text no. 4
Hofman MS, Violet J, Hicks RJ, Ferdinandus J, Thang SP, Akhurst T, et al. [177Lu]-PSMA-617 radionuclide treatment in patients with metastatic castration-resistant prostate cancer (LuPSMA trial): A single-centre, single-arm, phase 2 study. Lancet Oncol 2018;19:825-33.  Back to cited text no. 5
Krolicki L, Morgenstern A, Kunikowska J, Koiziar H, Krolicki B, Jackaniski M, et al. Glioma Tumors Grade II/III-Local Alpha Emitters Targeted Therapy with 213 Bi-DOTA-Substance P, Endocrine Abstracts. Vol. 57. Society of Nuclear Medicine and Molecular Imaging; 2016. p. 632.  Back to cited text no. 6
Baum RP, Kulkarni HP. Duo PRRT of neuroendocrine tumours using concurrent and sequential administration of Y-90- and Lu-177-labeled somatostatin analogues. In: Hubalewska-Dydejczyk A, Signore A, de Jong M, Dierckx RA, Buscombe J, Van de Wiel CJ, editors. Somatostatin Analogues from Research to Clinical Practice. New York: Wiley; 2015.  Back to cited text no. 7



From: htziev@aol.com” <htziev@aol.com>

Reply-To: htziev@aol.com” <htziev@aol.com>

Date: Tuesday, March 2, 2021 at 11:04 AM

To: “Aviva Lev-Ari, PhD, RN” <AvivaLev-Ari@alum.berkeley.edu>

Subject: Dr Saul Hertz : Discovery for the Medical Uses of RADIOIODINE (RAI) MARCH 31ST: 80 Years


Other related articles published in this Open Access Online Scientific Journal include the following:


Experience with Thyroid Cancer

Author: Larry H. Bernstein, MD, FCAP



New Guidelines and Meeting Information on Advanced Thyroid Cancer as Reported by Cancer Network (Meeting Highlights)

Reporter: Stephen J. Williams, Ph.D.


The Experience of a Patient with Thyroid Cancer

Interviewer and Curator: Larry H Bernstein, MD, FCAP



Parathyroids and Bone Metabolism

Writer and Curator: Larry H. Bernstein, MD, FCAP



Thyroid Function and Disorders

Writer and Curator: Larry H. Bernstein, MD, FCAP


Summary and Perspectives: Impairments in Pathological States: Endocrine Disorders, Stress Hypermetabolism and Cancer

Author and Curator: Larry H. Bernstein, MD, FCAP


Introduction to Impairments in Pathological States: Endocrine Disorders, Stress Hypermetabolism and Cancer

Author and Curator: Larry H. Bernstein, MD, FCAP


Metformin, Thyroid-Pituitary Axis, Diabetes Mellitus, and Metabolism

Larry H, Bernstein, MD, FCAP, Author and Curator
and Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/9/27/2014/Metformin,_thyroid-pituitary_ axis,_diabetes_mellitus,_and_metabolism

Autophagy-Modulating Proteins and Small Molecules Candidate Targets for Cancer Therapy: Commentary of Bioinformatics Approaches

Author and Curator: Larry H Bernstein, MD, FCAP and Article Architect: Aviva Lev-Ari, PhD, RN



Neural Activity Regulating Endocrine Response

Writer and Curator: Larry H. Bernstein, MD, FCAP



Pituitary Neuroendocrine Axis

Writer and Curator: Larry H. Bernstein, MD, FCA


On the Influence of Hormones on Cancer

VOLUME 4: Human Reproductive System, Genomic Endocrinology and Cancer Types

(Series D: BioMedicine & Immunology) Kindle Edition. On Amazon.com  since February 2, 2021


Read Full Post »

Early Details of Brain Damage in COVID-19 Patients

Reporter: Irina Robu, PhD


COVID-19 has currently claimed more American lives than World War I, Vietnam War and the Korean war combined. And while it is mainly a respiratory disease, COVID-19 infection affects other organs, including the brain. Researchers at Harvard-affiliated Massachusetts General Hospital found that COVID patients with neurological symptoms show more than some metabolic disturbances in the brain as patients who have suffered oxygen deprivation.

During the course of the pandemic, thousand patients with COVID-19 have been seen at MGH and the severity of the neurological symptoms varies from temporary loss of smell to more severe symptoms such as dizziness, confusion, seizures, and stroke. According to the principal investigator of the study, Eva Maria Ratai, Department of Radiology used 3 Tesla Magnetic Resonance Spectroscopy (MRS) to identify neurochemical abnormalities even the structural imagining findings are normal. COVID-19 patients’ brains showed N-acetyl-aspartate (NAA) reduction, choline elevation, and myo-inositol elevation, comparable to what is seen with these metabolites in other patients with leukoencephalopathy after hypoxia without COVID.

Their research indicated that one of patients with COVID-19 indicate the most severe white matter damage, whereas another had COVID-19 associated necrotizing leukoencephalopathy at the time of imaging. And the patient that experience cardiac arrest showed subtle white matter changes on structural MR. The control cases included one patient with damage due to hypoxia from other causes: one with sepsis-related white matter damage, and a normal, age-matched, healthy volunteer.

The main question still remains whether the decrease in the oxygen of the brain is causing the white matter to change or whether the virus itself is attacking white matter. The conclusion is that MRS can be used as a disease and therapy monitoring tool.


Small study reveals details of brain damage in COVID-19 patients

Read Full Post »

COVID concern in Cardiology: Asymptomatic patients who have been previously infected demonstrating evidence on MRI of scarring or myocarditis

Reporters: Justin D. Pearlman, MD, PhD, FACC and Aviva Lev-Ari, PhD, RN


The Voice of Dr. Justin D. Pearlman, MD, PhD, FACC

Indeed, many viruses can cause inflammation and weakening of the heart.

So far there is no established action to take for prevention, and management is based on clinical manifestations of heart failure: shortness of breath, particularly if worse laying flat or worse with exertion, leg swelling (edema), blood tests showing elevated brain natriuretic peptide (BNP or proBNP, a marker of heart muscle strain), and a basic metabolic panel that may show “pre-renal azotemia” (elevation of BUN and Creatinine, typically in a ratio >20:1) and/or hyponatremia (sodium concentration below 135 mEq/dL). If any of the above are suspected, it is reasonable to get transthoracic echocardiography for systolic and diastolic function. If either systolic or diastolic function by ultrasound show significant impairment not improved by usual therapy (diuretic, ACEI/ARB/ARNI, blocker, aldosterone inhibitor e.g. spironolactone) then an MRI scar map may be considered (MRI scar maps show retention of gadolinium contrast agent by injured heart muscle, first demonstrated by Dr. Justin Pearlman during angiogenesis research MRI studies).

There is no controversy in the above, the controversy is a rush to expanded referral for cardiac MRI without clear clinical evidence of heart impairment, at a stage when there is no established therapy for possible detection of myocarditis (cardiac inflammation). General unproven measures for inflammation may include taking ginger and tumeric supplements if well tolerated by the stomach, drinking 2 cups/day of Rooibos Tea if well tolerated by the liver.

Canakinumab was recommended by one research group to treat inflammation and risk to the heart if the blood test hsCRP is elevated (in addition to potential weakening of muscle, inflammation activates complement, makes atherosclerosis lesions unstable, and thus may elevate risk of heart attack, stroke, renal failure or limb loss from blocked blood delivery). The canakinumab studies were published in NEJM and LANCET with claims of significant improvement in outcomes, but that was not approved by FDA or confirmed by other groups, even though it has biologic plausibility. https://www.thelancet.com/journals/lancet/article/PIIS0140-67361732247-X/fulltext


Some Heart Societies Agree on Cautions for COVID-Myocarditis Screening

— Official response has been modest, though

Such evidence of myocardial injury and inflammation on CMR turned up in a German study among people who recovered from largely mild or moderate cases of COVID-19 compared with healthy controls and risk factor-matched controls.

Then an Ohio State University study showed CMR findings suggestive of myocarditis in 15% of collegiate athletes after asymptomatic or mild SARS-CoV-2 infection.

But an open letter from some 50 medical professionals across disciplines emphasized that “prevalence, clinical significance and long-term implications” of such findings aren’t known. The letter called on the 18 professional societies to which it was sent on Tuesday to release clear guidance against CMR screening in the general population to look for post-COVID heart damage in the absence of symptoms.

The Society for Cardiac Magnetic Resonance quickly responded with a brief statement from its chief executive officer, Chiara Bucciarelli-Ducci, MD, PhD, agreeing that routine CMR in asymptomatic patients after COVID-19 “is currently not justified… and it should not be encouraged.”

She referred clinicians to the multisociety guidelines on clinical indications of CMR when deciding whether to scan COVID-19 patients. “While CMR is an excellent imaging tool for diagnosing myocarditis in patients with suspected disease, we do not recommend its use in patients without symptoms,” she added.

The American Heart Association didn’t put out any written statement but offered spokesperson Manesh Patel, MD, chair of its Diagnostic and Interventional Cath Committee.

“The American Heart Association’s position on this is that in general we agree that routine cardiac MRI should not be conducted unless in the course of a study” for COVID-19 patients, he said. “There’s a lot of evolving information around people with COVID, and certainly asymptomatic status, whether it’s recent or prior, it’s not clearly known what the MRI findings will mean or what the long-term implications are without both a control group and an understanding around population.”

The ACC opted against taking a stand. It provided MedPage Today with the following statement from ACC President Athena Poppas, MD:

“We appreciate the authors’ concerns about the potential mischaracterization of the long-term impact of myocarditis after a COVID-19 diagnosis and the need for well-designed clinical trials and careful, long term follow-up. The pandemic is requiring everyone make real-time decisions on how to best care for heart disease patients who may be impacted by COVID-19. The ACC is committed to helping synthesize and provide the most up-to-date, high quality information possible to the cardiovascular care team. We will continue to review and assess the scientific data surrounding cardiac health and COVID-19 and issue guidance to help our care team.”

While the open letter noted that some post-COVID patients have been asking for CMR, Walsh noted that primary care would likely see the brunt of any such influx. She personally has not had any patients ask to be screened.



Effect of interleukin-1β inhibition with canakinumab on incident lung cancer in patients with atherosclerosis: exploratory results from a randomised, double-blind, placebo-controlled trial



Inflammation in the tumour microenvironment mediated by interleukin 1β is hypothesised to have a major role in cancer invasiveness, progression, and metastases. We did an additional analysis in the Canakinumab Anti-inflammatory Thrombosis Outcomes Study (CANTOS), a randomised trial of the role of interleukin-1β inhibition in atherosclerosis, with the aim of establishing whether inhibition of a major product of the Nod-like receptor protein 3 (NLRP3) inflammasome with canakinumab might alter cancer incidence.


We did a randomised, double-blind, placebo-controlled trial of canakinumab in 10 061 patients with atherosclerosis who had had a myocardial infarction, were free of previously diagnosed cancer, and had concentrations of high-sensitivity C-reactive protein (hsCRP) of 2 mg/L or greater. To assess dose–response effects, patients were randomly assigned by computer-generated codes to three canakinumab doses (50 mg, 150 mg, and 300 mg, subcutaneously every 3 months) or placebo. Participants were followed up for incident cancer diagnoses, which were adjudicated by an oncology endpoint committee masked to drug or dose allocation. Analysis was by intention to treat. The trial is registered with ClinicalTrials.govNCT01327846. The trial is closed (the last patient visit was in June, 2017).


Baseline concentrations of hsCRP (median 6·0 mg/L vs 4·2 mg/L; p<0·0001) and interleukin 6 (3·2 vs 2·6 ng/L; p<0·0001) were significantly higher among participants subsequently diagnosed with lung cancer than among those not diagnosed with cancer. During median follow-up of 3·7 years, compared with placebo, canakinumab was associated with dose-dependent reductions in concentrations of hsCRP of 26–41% and of interleukin 6 of 25–43% (p<0·0001 for all comparisons). Total cancer mortality (n=196) was significantly lower in the pooled canakinumab group than in the placebo group (p=0·0007 for trend across groups), but was significantly lower than placebo only in the 300 mg group individually (hazard ratio [HR] 0·49 [95% CI 0·31–0·75]; p=0·0009). Incident lung cancer (n=129) was significantly less frequent in the 150 mg (HR 0·61 [95% CI 0·39–0·97]; p=0·034) and 300 mg groups (HR 0·33 [95% CI 0·18–0·59]; p<0·0001; p<0·0001 for trend across groups). Lung cancer mortality was significantly less common in the canakinumab 300 mg group than in the placebo group (HR 0·23 [95% CI 0·10–0·54]; p=0·0002) and in the pooled canakinumab population than in the placebo group (p=0·0002 for trend across groups). Fatal infections or sepsis were significantly more common in the canakinumab groups than in the placebo group. All-cause mortality did not differ significantly between the canakinumab and placebo groups (HR 0·94 [95% CI 0·83–1·06]; p=0·31).


Our hypothesis-generating data suggest the possibility that anti-inflammatory therapy with canakinumab targeting the interleukin-1β innate immunity pathway could significantly reduce incident lung cancer and lung cancer mortality. Replication of these data in formal settings of cancer screening and treatment is required.


Novartis Pharmaceuticals.

Read Full Post »

Expanding 3D Printing in Cardiology

Reporter: Irina Robu, PhD

3D printing is a fabrication technique used to transform digital objects into physical models, which builds structures of arbitrary geometry by depositing material in successive layers on the basis of specific digital design. Even though, the use of 3D bioprinting in cardiovascular medicine is relatively new development, advancement within this discipline is occurring at such a rapid rate. Most cardiologists believed the costs would be too high for routine use such that the price tag was better for academic applications.

Now as the prices are starting to lower, the idea of using 3D printed models of organs vessels and tissue manufactured based on CT, MRI and echocardiography might be beneficial according to Dr. Fadi Matar, professor at University of South Florida. He and his cardiology colleagues use 3D printed models to allow them to view patient’s complex anatomies before deciding what treatments to pursue. The models allow them to calculate the size and exact placement of devices which has led to shorter procedure time and better outcome.

In a study published in Academic Radiology, David Ballard, professor at University School of Medicine appraised the costs of setting up a 3D printing lab including the commercial printer plus software, lab space, materials and staffing. According to Ballard’s team, the commercial printers start at $12,000 but can be as high as high as $500,000.

According to American Medical Association-approved Category III Current Procedural Terminology (CPT) codes allows cardiology relief from setting up a new 3D printing lab such as Codes 0559T and 0560T, for individually prepared 3D-printed anatomical models with one or more components (including arteries and veins) and Codes 0561T and 0562T, which are for the production of personalized 3D-printed cutting or drilling tools that use patient imaging data and often are used to guide or facilitate surgery.

These codes have been met with enthusiasm by teams eyeing 3D printing, but there are noteworthy limitations to Category III codes—which are temporary codes describing emerging technologies, services and procedures that are used for tracking effectiveness data. It is important to note that Category III codes are not reimbursed but often are a step toward reimbursement.

New and improved materials also might lead to a sharper focus on 3D printing in cardiology. Dr. Fadi Matar says companies are working on materials that better mimic elements of the heart. Such “mimicry” ought to enhance the value of 3D-printed models since they will give cardiologists more realistic insights into how specific devices will interact with an individual patient’s heart. Even with the complex modalities of using 3D bioprinting, in time there would be less obstacles to being able to set up a 3D bioprinter lab.



Read Full Post »

Artificial Intelligence Innovations in Cardiac Imaging

Reporter: Aviva Lev-Ari, PhD, RN

‘CTA-for-All’ fast-tracks intervention, improves LVO detection in stroke patients

A “CTA-for-All” stroke imaging policy improved large vessel occlusion (LVO) detection, fast-tracked intervention and improved outcomes in a recent study of patients with acute ischemic stroke (AIS), researchers reported in Stroke.

“Combined noncontrast computed tomography (NCCT) and CT angiography (CTA) have been championed as the new minimum standard for initial imaging of disabling stroke,” Mayer, a neurologist at Henry Ford Hospital in Detroit, and co-authors wrote in their paper. “Patient selection criteria that impose arbitrary limits on time from last known well (LKW) or baseline National Institutes of Health Stroke Scale (NIHSS) score may delay CTA and the diagnosis of LVO.”

“These findings suggest that a uniform CTA-for-All imaging policy for stroke patients presenting within 24 hours is feasible and safe, improves LVO detection, speeds intervention and can improve outcomes,” the authors wrote. “The benefit appears to primarily affect patients presenting within six hours of symptom onset.”




How to integrate AI into the cardiac imaging pipeline

Hsiao said physicians can expect “a little bit of generalization” from neural networks, meaning they’ll work okay on data that they’ve never seen, but they’re not going to produce perfect results the first time around. If a model was trained on 3T MRI data, for example, and someone inputs 1.5T MRI data, it might not be able to analyze that information comprehensively. If some 1.5T data were fed into the model’s training algorithm, though, that could change.

According to Hsiao, all of this knowledge means little without clinical validation. He said he and his colleagues are working to integrate algorithms into the clinical environment such that a radiologist could hit a button and AI could auto-prescribe a set of images. Even better, he said, would be the ability to open up a series and have it auto-prescribe itself.

“That’s where we’re moving next, so you don’t have to hit any buttons at all,” he said.




DiA Imaging, IBM pair to take the subjectivity out of cardiac image analysis




FDA clears Ultromics’ AI-based CV image analysis system

Smartphone app accurately finds, identifies CV implants—and fast

According to the study, the finalized model achieved 95% sensitivity and 98% specificity.

Ferrick et al. said that since their training sample size was somewhat small and limited to a single institution, it would be valuable to validate the model externally. Still, their neural network was able to accurately identify CIEDs on chest radiographs and translate that ability into a phone app.

“Rather than the conventional ‘bench-to-bedside’ approach of translational research, we demonstrated the feasibility of ‘big data-to-bedside’ endeavors,” the team said. “This research has the potential to facilitate device identification in urgent scenarios in medical settings with limited resources.”



Machine learning cuts cardiac MRI analysis from minutes to seconds

“Cardiovascular MRI offers unparalleled image quality for assessing heart structure and function; however, current manual analysis remains basic and outdated,” Manisty said in a statement. “Automated machine learning techniques offer the potential to change this and radically improve efficiency, and we look forward to further research that could validate its superiority to human analysis.”

It’s estimated that around 150,000 cardiac MRIs are performed in the U.K. each year, she said, and based on that number, her team thinks using AI to read scans could mean saving 54 clinician-days per year at every health center in the country.

“Our dataset of patients with a range of heart diseases who received scans enabled us to demonstrate that the greatest sources of measurement error arise from human factors,” Manisty said. “This indicates that automated techniques are at least as good as humans, with the potential soon to be ‘superhuman’—transforming clinical and research measurement precision.




General SOURCE

From: Cardiovascular Business <news@mail.cardiovascularbusiness.com>

Reply-To: Cardiovascular Business <news@mail.cardiovascularbusiness.com>

Date: Tuesday, December 17, 2019 at 9:31 AM

To: Aviva Lev-Ari <AvivaLev-Ari@alum.berkeley.edu>

Subject: Cardiovascular Imaging | December 2019

Read Full Post »

Cardiac MRI Imaging Breakthrough: The First AI-assisted Cardiac MRI Scan Solution, HeartVista Receives FDA 510(k) Clearance for One Click™ Cardiac MRI Package

Reporter: Aviva Lev-Ari, PhD, RN


HeartVista Receives FDA 510(k) Clearance for One Click™ Cardiac MRI Package, the First AI-assisted Cardiac MRI Scan Solution

The future of imaging is here—and FDA cleared.

LOS ALTOS, Calif.–(BUSINESS WIRE)–HeartVista, a pioneer in AI-assisted MRI solutions, today announced that it received 510(k) clearance from the U.S. Food and Drug Administration to deliver its AI-assisted One Click™ MRI acquisition software for cardiac exams. Despite the many advantages of cardiac MRI, or cardiac magnetic resonance (CMR), its use has been largely limited due to a lack of trained technologists, high costs, longer scan time, and complexity of use. With HeartVista’s solution, cardiac MRI is now simple, time-efficient, affordable, and highly consistent.

“HeartVista’s Cardiac Package is a vital tool to enhance the consistency and productivity of cardiac magnetic resonance studies, across all levels of CMR expertise,” said Dr. Raymond Kwong, MPH, Director of Cardiac Magnetic Resonance Imaging at Brigham and Women’s Hospital and Associate Professor of Medicine at Harvard Medical School.

A recent multi-center, outcome-based study (MR-INFORM), published in the New England Journal of Medicine, demonstrated that non-invasive myocardial perfusion cardiovascular MRI was as good as invasive FFR, the previous gold standard method, to guide treatment for patients with stable chest pain, while leading to 20% fewer catheterizations.

“This recent NEJM study further reinforces the clinical literature that cardiac MRI is the gold standard for cardiac diagnosis, even when compared against invasive alternatives,” said Itamar Kandel, CEO of HeartVista. “Our One Click™ solution makes these kinds of cardiac MRI exams practical for widespread adoption. Patients across the country now have access to the only AI-guided cardiac MRI exam, which will deliver continuous imaging via an automated process, minimize errors, and simplify scan operation. Our AI solution generates definitive, accurate and actionable real-time data for cardiologists. We believe it will elevate the standard of care for cardiac imaging, enhance patient experience and access, and improve patient outcomes.”

HeartVista’s FDA-cleared Cardiac Package uses AI-assisted software to prescribe the standard cardiac views with just one click, and in as few as 10 seconds, while the patient breathes freely. A unique artifact detection neural network is incorporated in HeartVista’s protocol to identify when the image quality is below the acceptable threshold, prompting the operator to reacquire the questioned images if desired. Inversion time is optimized with further AI assistance prior to the myocardial delayed-enhancement acquisition. A 4D flow measurement application uses a non-Cartesian, volumetric parallel imaging acquisition to generate high quality images in a fraction of the time. The Cardiac Package also provides preliminary measures of left ventricular function, including ejection fraction, left ventricular volumes, and mass.

HeartVista is presenting its new One Click™ Cardiac Package features at the Radiological Society of North America (RSNA) annual meeting in Chicago, on Dec. 4, 2019, at 2 p.m., in the AI Showcase Theater. HeartVista will also be at Booth #11137 for the duration of the conference, from Dec. 1 through Dec. 5.

About HeartVista

HeartVista believes in leveraging artificial intelligence with the goal of improving access to MRI and improved patient care. The company’s One Click™ software platform enables real-time MRI for a variety of clinical and research applications. Its AI-driven, one-click cardiac localization method received first place honors at the International Society for Magnetic Resonance in Medicine’s Machine Learning Workshop in 2018. The company’s innovative technology originated at the Stanford Magnetic Resonance Systems Research Laboratory. HeartVista is funded by Khosla Ventures, and the National Institute of Health’s Small Business Innovation Research program.

For more information, visit www.heartvista.ai


Reply-To: Kimberly Ha <kimberly.ha@kkhadvisors.com>

Date: Tuesday, October 29, 2019 at 11:01 AM

To: Aviva Lev-Ari <AvivaLev-Ari@alum.berkeley.edu>

Subject: HeartVista Receives FDA Clearance for First AI-assisted Cardiac MRI Solution

Read Full Post »

Disentangling molecular alterations from water-content changes in the aging human brain using quantitative MRI

Reporter: Dror Nir, PhD


Authors’ list: Shir Filo, Oshrat Shtangel, Noga Salamon, Adi Kol, Batsheva Weisinger, Sagiv Shifman & Aviv A. Mezer
Published in: Nature Communications volume 10, Article number: 3403 (2019)


It is an open question whether aging-related changes throughout the brain are driven by a common factor or result from several distinct molecular mechanisms. Quantitative magnetic resonance imaging (qMRI) provides biophysical parametric measurements allowing for non-invasive mapping of the aging human brain. However, qMRI measurements change in response to both molecular composition and water content. Here, we present a tissue relaxivity approach that disentangles these two tissue components and decodes molecular information from the MRI signal. Our approach enables us to reveal the molecular composition of lipid samples and predict lipidomics measurements of the brain. It produces unique molecular signatures across the brain, which are correlated with specific gene-expression profiles. We uncover region-specific molecular changes associated with brain aging. These changes are independent from other MRI aging markers. Our approach opens the door to a quantitative characterization of the biological sources for aging, that until now was possible only post-mortem.



The biology of the aging process is complex, and involves various physiological changes throughout cells and tissues1. One of the major changes is atrophy, which can be monitored by measuring macroscale brain volume reduction1,2. In some cases, atrophy can also be detected as localized microscale tissue loss reflected by increased water content3. This process is selective for specific brain regions and is thought to be correlated with cognitive decline in Alzheimer’s disease2,4,5. In addition to atrophy, there are molecular changes associated with the aging of both the normal and pathological brain5,6. Specifically, lipidome changes are observed with age, and are associated with several neurological diseases7,8,9,10,11.

It is an open question as to whether there are general principles that govern the aging process, or whether each system, tissue, or cell deteriorates with age for different reasons12,13. On one hand, the common-cause hypothesis proposes that different biological aging-related changes are the result of a single underlying factor14,15. This implies that various biomarkers of aging will be highly correlated16. On the other hand, the mosaic theory of aging suggests that there are several distinct aging mechanisms that have a heterogenous effect throughout the brain12,13. According to this latter view, combining different measurements of brain tissue is crucial in order to fully describe the state of the aging brain. To test these two competing hypotheses in the context of volumetric and molecular aging-related changes, it is essential to measure different biological aspects of brain tissue. Unfortunately, the molecular correlates of aging are not readily accessible by current in vivo imaging methods.

The main technique used for non-invasive mapping of the aging process in the human brain is magnetic resonance imaging (MRI)2,17,18,19. Advances in the field have led to the development of quantitative MRI (qMRI). This technique provides biophysical parametric measurements that are useful in the investigation and diagnosis of normal and abnormal aging20,21,22,23,24,25,26,27. qMRI parameters have been shown to be sensitive to the microenvironment of brain tissue and are therefore named in vivo histology28,29,30. Nevertheless, an important challenge in applying qMRI measurements is increasing their biological interpretability. It is common to assume that qMRI parameters are sensitive to the myelin fraction20,23,30,31,32,33, yet any brain tissue including myelin is a mixture of multiple lipids and proteins. Moreover, since water protons serve as the source of the MRI signal, the sensitivity of qMRI parameters to different molecular microenvironments may be confounded by their sensitivity to the water content of the tissue34,35. We hypothesized that the changes observed with aging in MRI measurements20,23,30,31,32,33,36 such as R1, R2, mean diffusivity (MD), and magnetization transfer saturation (MTsat)37, could be due to a combination of an increase in water content at the expense of tissue loss, and molecular alterations in the tissue.

Here, we present a qMRI analysis that separately addresses the contribution of changes in molecular composition and water content to brain aging. Disentangling these two factors goes beyond the widely accepted “myelin hypothesis” by increasing the biological specificity of qMRI measurements to the molecular composition of the brain. For this purpose, we generalize the concept of relaxivity, which is defined as the dependency of MR relaxation parameters on the concentration of a contrast agent38. Instead of a contrast agent, our approach exploits the qMRI measurement of the local non-water fraction39 to assess the relaxivity of the brain tissue itself. This approach allows us to decode the molecular composition from the MRI signal. In samples of known composition, our approach provides unique signatures for different brain lipids. In the live human brain, it produces unique molecular signatures for different brain regions. Moreover, these MRI signatures agree with post-mortem measurements of the brain lipid and macromolecular composition, as well as with specific gene-expression profiles. To further validate the sensitivity of the relaxivity signatures to molecular composition, we perform direct comparison of MRI and lipidomics on post-mortem brains. We exploit our approach for multidimensional characterization of aging-related changes that are associated with alterations in the molecular composition of the brain. Finally, we evaluate the spatial pattern of these changes throughout the brain, in order to compare the common-cause and the mosaic theories of aging in vivo.



Different brain lipids have unique relaxivity signatures
The aging process in the brain is accompanied by changes in the chemophysical composition, as well as by regional alterations in water content. In order to examine the separate pattern of these changes, we developed a model system. This system was based on lipid samples comprising common brain lipids (phosphatidylcholine, sphingomyelin, phosphatidylserine, phosphatidylcholine-cholesterol, and phosphatidylinositol-phosphatidylcholine)7. Using the model system, we tested whether accounting for the effect of the water content on qMRI parameters provides sensitivity to fine molecular details such as the head groups that distinguish different membrane phospholipids. The non-water fraction of the lipid samples can be estimated by the qMRI measurement of lipid and macromolecular tissue volume (MTV, for full glossary of terms see Supplementary Table 1)39. By varying the concentration of the lipid samples, we could alter their MTV and then examine the effect of this manipulation on qMRI parameters. The parameters we estimated for the lipid samples were R1, R2, and MTsat. The potential ambiguity in the biological interpretation of qMRI parameters is demonstrated in Fig. 1a. On one hand, samples with similar lipid composition can present different R1 measurements (Fig. 1a, points 1 & 2). On the other hand, scanning samples with different lipid compositions may result in similar R1 measurements (Fig. 1a, points 2 & 3). This ambiguity stems from the confounding effect of the water content on the MR relaxation properties.

Screenshot 2019-08-01 at 14.36.20

We evaluated the dependency of different qMRI parameters on the non-water fraction estimated by MTV. This analysis revealed strong linear dependencies (median R2 = 0.74, Fig. 1a, b and Supplementary Fig. 1a, b). These linear MTV dependencies change as a function of the lipid composition, reflecting the inherent relaxivity of the different lipids. We could therefore use the MTV derivatives of qMRI parameters (dqMRIdMTV, i.e., the slope of the linear relationship between each qMRI parameter and MTV) as a measure that is sensitive to molecular composition. By accounting for the Multidimensional Dependency on MTV (“MDM”) of several qMRI parameters, a unique MRI relaxivity signature was revealed for each lipid (Fig. 1c). This implies that the water-related ambiguity demonstrated in the inset of Fig. 1a can be removed by measuring the MTV dependencies (Fig. 1c). Creating mixtures of several lipids provided supportive evidence for the generality of our framework. Figure 1d and Supplementary Fig. 1c show that the qMRI measurements of a mixture can be predicted by summing the MTV dependencies of pure lipids (for further details see Supplementary Note 1 and Supplementary Fig. 2). Furthermore, we used this biophysical model to predict the lipid composition of a mixture from its MDM measurements (Fig. 1e). This model provided a good estimation of the sphingomyelin (Spg) and phosphatidylserine (PS) content (R2 > 0.64) but failed to predict phosphatidylcholine (PtdCho) content (for further details see Supplementary Note 2). While lipids are considered to be a major source of the MRI signal in the brain 40,41,42,43,44,45, our approach can be applied to other compounds to reveal differences in the MRI signal between different proteins, sugars, and ions (Supplementary Fig. 1d). Hence, the relationships between qMRI parameters and MTV account for the effect of water on MRI measurements and could be of use in quantifying the biological and molecular contributions to the MRI signal of water protons.

The tissue relaxivity of the human brain is region-specific.
In order to target age-related changes in molecular composition, we applied the same approach for the human brain (Fig. 2a).

Screenshot 2019-08-01 at 14.41.35

We found that the linear dependency of qMRI parameters on MTV is not limited to in vitro samples and a similar relationship was also evident in the human brain (Fig. 2b and Supplementary Figs. 3–5). Importantly, different brain regions displayed a distinct dependency on MTV. Therefore, the relaxivity of brain tissue is region-specific. Figure 2b provides an example for the regional linear trends of R1 and MTsat in a single subject. Remarkably, while the thalamus and the pallidum presented relatively similar R1 dependencies on MTV, their MTsat dependencies were different (p < 0.001, two-sample t-test). Compared to these two brain regions, frontal white-matter demonstrated different dependencies on MTV (p < 0.001, two-sample t-test). A better separation between brain regions can therefore be achieved by combining the MTV dependencies of several qMRI parameters (MTsat, MD, R1 and R2). The MTV derivatives of qMRI parameters are consistent across subjects (Fig. 2c and Supplementary Fig. 6), with good agreement between hemispheres (Supplementary Fig. 5). Moreover, they provide a novel pattern of differentiation between brain regions, which is not captured by conventional qMRI methods (Supplementary Fig. 7). In our lipid sample experiments, the MDM approach revealed unique relaxivity signatures of different lipids (Fig. 1c). Therefore, we attribute the observed diversity in the MTV derivatives of qMRI parameters across brain regions to the intrinsic heterogeneity in the chemophysical microenvironment of these regions. The multidimensional dependency of various qMRI parameters on MTV can be represented by the space of MTV derivatives to reveal a unique chemophysical MDM signature for different brain regions (Fig. 2d, see explanatory scheme of the MDM method in Supplementary Fig. 8). Fig. 2 figure2 The MDM method provides region-specific signatures in the in vivo human brain. a Representative MTV, MTsat, and R1 maps. b Calculating the MDM signatures. The dependency of R1 (left) and MTsat (right) on MTV in three brain regions of a single subject. For each region, MTV values were pooled into bins (dots are the median of each bin; shaded area is the median absolute deviation), and a linear fit was calculated (colored lines). The slopes of the linear fit represent the MTV derivatives of R1 and MTsat and vary across brain regions. c The reliability of the MDM method across subjects. Variation in the MTV derivatives of R1 (left) and MTsat (right) in young subjects (N = 23). Different colors represent 14 brain regions (see legend). Edges of each box represent the 25th, and 75th percentiles, median is in black, and whiskers extends to extreme data points. Different brain regions show distinct MTV derivatives. d Unique MDM signatures for different brain regions (in different colors). Each axis is the MTV derivative (“MDM measurements”) of a different qMRI parameter (R1, MTsat, R2, and MD). The range of each axis is in the legend. Colored traces extend between the MDM measurements, shaded areas represent the variation across subjects (N = 23). An overlay of all MDM signatures is marked with dashed lines Full size image The in vivo MDM approach captures ex vivo molecular profiles To validate that the MDM signatures relate to the chemophysical composition of brain tissue, we compared them to a previous study that reported the phospholipid composition of the human brain7. First, we established the comparability between the in vivo MRI measurements and the reported post-mortem data. MTV measures the non-water fraction of the tissue, a quantity that is directly related to the total phospholipid content. Indeed, we found good agreement between the in vivo measurement of MTV and the total phospholipid content across brain regions (R2 = 0.95, Fig. 3a). Söderberg et al.7 identified a unique phospholipid composition for different brain regions along with diverse ratios of phospholipids to proteins and cholesterol. We compared this regional molecular variability to the regional variability in the MDM signatures. To capture the main axes of variation, we performed principal component analysis (PCA) on both the molecular composition of the different brain regions and on their MDM signatures. For each of these two analyses, the first principal component (PC) explained >45% of the variance. The regional projection on the first PC of ex vivo molecular composition was highly correlated (R2 = 0.84, Fig. 3b) with the regional projection on the first PC of in vivo MDM signatures. This confirms that brain regions with a similar molecular composition have similar MDM. Supplementary Fig. 9a provides the correlations of individual lipids with MDM. Importantly, neither MTV nor the first PC of standard qMRI parameters was as strongly correlated with the ex vivo molecular composition as the MDM (Supplementary Fig. 9b, c). We next used the MDM measurements as predictors for molecular properties of different brain regions. Following our content predictions for lipids samples (Fig. 1e), we constructed a weighted linear model for human data (for further details see Supplementary Note 3). To avoid over fitting, we reduced the number of fitted parameters by including only the MDM and the molecular features that accounted for most of the regional variability. The MTV derivatives of R1 and MTsat accounted for most of the variance in MDM. Thus, we used these parameters as inputs to the linear model, while adjusting their weights through cross validation. We tested the performance of this model in predicting the three molecular features that account for most of the variance in the ex vivo molecular composition. Remarkably, MRI-driven MDM measurements provided good predictions for the regional sphingomyelin composition (R2 = 0.56, p < 0.05 for the F-test, Fig. 3c) and the regional ratio of phospholipids to proteins (R2 = 0.56, p < 0.05 for the F-test, Fig. 3c).

Screenshot 2019-08-01 at 14.44.06
Last, we compared the cortical MDM signatures to a gene co-expression network based on a widespread survey of gene expression in the human brain46. Nineteen modules were derived from the gene network, each comprised of a group of genes that co-varies in space. Six out of the nineteen gene modules were significantly correlated with the first PC of MDM. Interestingly, the first PC of MDM across the cortex was correlated most strongly with the two gene modules associated with membranes and synapses (Fig. 4, for further details see Supplementary Note 4 and Supplementary Figs. 10 and 11).

Screenshot 2019-08-01 at 14.47.04

Post-mortem validation for the lipidomic sensitivity of MDM.
The aforementioned analyses demonstrate strong agreement between in vivo MDM measurements and ex vivo molecular composition based on a group-level comparison of two different datasets. Strikingly, we were able to replicate this result at the level of the single brain. To achieve this we performed MRI scans (R1, MTsat, R2, MD, and MTV mapping) followed by histology of two fresh post-mortem porcine brains (Fig. 5a, b). First, we validated the qMRI estimation of MTV using dehydration techniques. MTV values estimated using MRI were in agreement with the non-water fraction found histologically (adjusted R2 = 0.64, p < 0.001 for the F-test, Fig. 5c).

Screenshot 2019-08-01 at 14.50.12
Next, we estimated the lipid composition of different brain regions. Thin-layer chromatography (TLC) was employed to quantify seven neutral and polar lipids (Supplementary Table 2 and Supplementary Fig. 12a). In accordance with the analysis in Fig. 3, we performed PCA to capture the main axes of variation in lipidomics, standard qMRI parameters, and MDM. Figure 5d shows that MTV did not correlate with the molecular variability across the brain, estimated by the 1st PC of lipidomics. Likewise, the molecular variability did not agree with the 1st PC of standard qMRI parameters (Fig. 5e).

Last, we applied the MDM approach to the post-mortem porcine brain. Similar to the human brain, different porcine brain regions have unique MDM signatures (Fig. 5f, g and Supplementary Fig. 12b). Remarkably, we found that agreement between lipid composition and MRI measurements emerges at the level of the MDM signatures. The molecular variability across brain regions significantly correlated with the regional variability in the MDM signatures (adjusted R2 = 0.3, p < 0.01 for the F-test, Fig. 5h). Excluding from the linear regression five outlier brain regions where the histological lipidomics results were 1.5 standard deviations away from the center yielded an even stronger correlation between MDM signatures and lipid composition (adjusted R2 = 0.55, p < 0.001 for the F-test, Supplementary Fig. 12c). This post-mortem analysis validates that the MDM approach allows us to capture molecular information using MRI at the level of the individual brain.

Disentangling water and molecular aging-related changes.
After establishing the sensitivity of the MDM signatures to the molecular composition of the brain, we used them to evaluate the chemophysical changes of the aging process. To assess aging-related changes across the brain, we scanned younger and older subjects (18 older adults aged 67 ± 6 years and 23 younger adults aged 27 ± 2 years). First, we identified significant molecular aging-related changes in the MDM signatures of different brain regions (Figs. 6 and 7, right column; Supplementary Fig. 13). Next, we tested whether the changes in MRI measurements, observed with aging, result from a combination of changes in the molecular composition of the tissue and its water content. We found that although it is common to attribute age-related changes in R1 and MTsat to myelin28,30,36, these qMRI parameters combine several physiological aging aspects. For example, using R1 and MTsat we identified significant aging-related changes in the parietal cortex, the thalamus, the parietal white-matter and the temporal white-matter (Figs. 6 and 7, left column). However, the MDM approach revealed that these changes have different biological sources (Figs. 6 and 7, middle columns; see Supplementary Figs. 14–17 for more brain regions).

Screenshot 2019-08-01 at 14.51.53

Screenshot 2019-08-01 at 14.54.44

Screenshot 2019-08-01 at 14.56.06

In agreement with the mosaic hypothesis, we identified distinct aging patterns for different brain regions. For example, in the hippocampus we found a change in R2* values related to a higher iron concentration with age, along with significant reduction in the total hippocampal volume (Fig. 8a). This age-related shrinkage was not accompanied by lower MTV values, indicating conserved tissue density (Fig. 7b). In addition, there was no significant difference in the hippocampal MDM signature with age (Fig. 7b). Cortical gray-matter areas also exhibited similar trends of volume reduction without major loss in tissue density (Fig. 8a). Unlike the gray matter, in the white matter we did not find volume reduction or large iron accumulation with age (Fig. 8a). However, we did find microscale changes with age in tissue composition, as captured by the MDM signature (Figs. 6a and 7c, and Supplementary Fig. 13), accompanied by a significant density-related decline in MTV (Fig. 8a). These findings are consistent with previous histological studies49,50,51 (see Discussion), and provide the ability to monitor in vivo the different components of the aging mosaic.

Last, to test whether the different biological aging trajectories presented in Fig. 8a share a common cause, we evaluated the correlations between them (Fig. 8b). Importantly, the chemophysical trajectory did not correlate significantly with the iron or volume aging patterns. The spatial distribution of water-related changes was found to correlate with iron content alterations (R2 = 0.27) and chemophysical alterations (R2 = 0.25). However, the strongest correlation between aging-related changes was found in volume and iron content (R2 = 0.77). As shown previously, this correlation may be explained to some extent by a systematic bias in automated tissue classification23. Additional analysis revealed that the different dimensions of the MDM signature capture distinct patterns of aging-related changes (Supplementary Fig. 30). Hence, complementary information regarding the various chemophysical mechanisms underlying brain aging could be gained by combining them.



Normal brain aging involves multiple changes, at both the microscale and macroscale level. MRI is the main tool for in vivo evaluation of such age-related changes in the human brain. Here, we propose to improve the interpretation of MRI findings by accounting for the fundamental effect of the water content on the imaging parameters. This approach allows for non-invasive mapping of the molecular composition in the aging human brain.

Our work is part of a major paradigm shift in the field of MRI toward in vivo histology30,36,52. The MDM approach contributes to this important change by providing a hypothesis-driven biophysical framework that was rigorously developed. We demonstrated the power of our framework, starting from simple pure lipid phantoms to more complicated lipid mixtures, and from there, to the full complexity of the brain. In the brain, we show both in vivo and post-mortem validations for the molecular sensitivity of the MDM signatures. Early observations relate different qMRI parameters to changes in the fraction of myelin20,23,30,31,32,33,36. The current approach enriches this view and provides better sensitivity to the molecular composition and fraction of myelin and other cellular tissues.

We developed a unique phantom system of lipid samples to validate our method. While the phantom system is clearly far from the complexity of brain tissue, its simplicity allowed us to verify the specificity of our method to the chemophysical environment. Remarkably, our approach revealed unique signatures for different lipids, and is therefore sensitive even to relatively subtle details that distinguish one lipid from another. We chose to validate our approach using membrane lipids based on previous experiments40,41,42,43,44,45. Nevertheless, we do acknowledge the fact that brain tissue comprises many other compounds beside lipids, such as proteins, sugars, and ions. As we have shown, these other compounds also exhibit unique dependency on MTV. The effect of such compounds, along with other factors such as microstructure, and multi-compartment organization28 is probably captured when we apply the MDM approach to the in vivo human brain. Therefore, the phantoms were made to examine the MRI sensitivity for the chemophysical environment, and the human brain data was used to measure the true biological effects in a complex in vivo environment.

Our relaxivity approach captures the molecular signatures of the tissue, but is limited in its abilities to describe the full complexity of the chemophysical environment of the human brain. For example, R1 and R2, which are used to generate the MDM signatures, are also sensitive to the iron content23,48,52. However, we found that most of our findings cannot be attributed to alterations in iron content as measured with R2* (for more details see Supplementary Note 5). While there is great importance in further isolating different molecular components, we argue that accounting for the major effect of water on qMRI parameters (for R2 distributions see Supplementary Fig. 5) is a crucial step towards more specific qMRI interpretation.

We provide evidence from lipids samples and post-mortem data for the sensitivity of the MDM signatures to the molecular environment (Figs. 1e, 3b, and 5h). The variability of MDM values between human brain regions also correlated with specific gene-expression profiles (Fig. 4). While the comparison of in vivo human brain measurements to previously published ex vivo findings is based on two different datasets, these measurements are highly stable across normal subjects and the intersubject variabilities are much smaller than the regional variability. The agreement between the modalities provides strong evidence for the ability of our method to capture molecular information.

Remarkably, we were able to demonstrate the sensitivity of MDM signatures to lipid composition using direct comparison on post-mortem porcine brains. Even though there are many challenges in scanning post-mortem tissue, segmenting it, and comparing it to anatomically relevant histological results, we were able to replicate our in vivo findings. We provide histological validation for the MRI estimation of MTV. Moreover, we find that while standard qMRI parameters and MTV do not explain the lipidomic variability across the brain, the MDM signatures are in agreement with histological results. Lipids constitute the majority of the brain’s dry weight and are known to be important for maintaining neural conduction and chemical balance53,54. The brain lipidome was shown to have a great deal of structural and functional diversity and was found to vary according to age, gender, brain region, and cell type55. Disruptions of the brain lipid metabolism have been linked to different disorders, including Alzheimer’s disease, Parkinson’s disease, depression, and anxiety7,8,11,54,55,56,57. Our results indicate that the MDM approach enhances the consistency between MRI-driven measurements and lipidomics, compared with standard qMRI parameters.

The simplicity of our model, which is based on a first-order approximation of qMRI dependencies, has great advantages in the modeling of complex environments. Importantly, we used lipids samples to show that the contributions of different mixture-components can be summed linearly (Fig. 1d). For contrast agents, the relaxivity is used to characterize the efficiency of different agents. Here, we treated the tissue itself, rather than a contrast material, as an agent to compute the relaxivity of the tissue. While relaxivity is usually calculated for R1 and R2, we extended this concept to other qMRI parameters. Our results showed that the tissue relaxivity changes as a function of the molecular composition. This suggests that the relaxivity of the tissue relates to the surface interaction between the water and the chemophysical environment. A theoretical formulation for the effect of the surface interaction on proton relaxation has been proposed before58,59. Specifically, a biophysical model for the linear relationship between R1 and R2 to the inverse of the water content (1/WC = 1/(1 – MTV)) was suggested by Fullerton et al.43. Interestingly, 1/WC varies almost linearly with MTV in the physiological range of MTV values. Applying our approach with 1/WC instead of MTV produces relatively similar results (Supplementary Fig. 28). However, using MTV as a measure of tissue relaxivity allowed us to generalize the linear model to multiple qMRI parameters, thus producing multidimensional MDM signatures.

We show that the MDM signatures allow for better understanding of the biological sources for the aging-related changes observe with MRI. Normal brain aging involves multiple changes, at both the microscale and macroscale levels. Measurements of macroscale brain volume have been widely used to characterize aging-associated atrophy. Our method of analysis can complement such findings and provide a deeper understanding of microscale processes co-occurring with atrophy. Moreover, it allows us to test whether these various microscale and macroscale processes are caused by a common factor or represent the aging mosaic. Notably, we discovered that different brain regions undergo different biological aging processes. Therefore, combining several measurements of brain tissue is crucial in order to fully describe the state of the aged brain. For example, the macroscale aging-related volume reduction in cortical gray areas was accompanied by conserved tissue density, as estimated by MTV, and region-specific chemophysical changes, as estimated by the MDM. In contrast, in white-matter areas both MDM and MTV changed with age. These microscale alterations were not accompanied by macroscale volume reduction. Our in vivo results were validated by previous histological studies, which reported that the cortex shrinks with age, while the neural density remains relatively constant49,50. In contrast, white matter was found to undergo significant loss of myelinated nerve fibers during aging51. In addition, we found that the shrinkage of the hippocampus with age is accompanied with conserved tissue density and chemophysical composition. This is in agreement with histological findings, which predict drastic changes in hippocampal tissue composition in neurological diseases such as Alzheimer, but not in normal aging49,50,60,61. In contrast, hippocampal macroscale volume reduction was observed in both normal and pathological aging2.

It should be noted that most of the human subjects recruited for this study were from the academic community. However, the different age groups were not matched for variables such as IQ and socioeconomic status. In addition, the sample size in our study was quite small. Therefore, the comparison we made between the two age groups may be affected by variables other than age. Our approach may benefit from validation based on larger quantitative MRI datasets27,62. Yet, we believe we have demonstrated the potential of our method to reveal molecular alterations in the brain. Moreover, the agreement of our findings with previous histological aging studies supports the association between the group differences we measured and brain aging. Our results suggest that the MDM approach may be very useful in differentiating the effects of normal aging from those of neurodegenerative diseases. There is also great potential for applications in other brain research fields besides aging. For example, our approach may be used to advance the study and diagnosis of brain cancer, in which the lipidomic environment undergoes considerable changes63,64,65.

To conclude, we have presented here a quantitative MRI approach that decodes the molecular composition of the aging brain. While common MRI measurements are primarily affected by the water content of the tissue, our method employed the tissue relaxivity to expose the sensitivity of MRI to the molecular microenvironment. We presented evidence from lipid samples, post-mortem porcine brains and in vivo human brains for the sensitivity of the tissue relaxivity to molecular composition. Results obtained by this method in vivo disentangled different biological processes occurring in the human brain during aging. We identified region-specific patterns of microscale aging-related changes that are associated with the molecular composition of the human brain. Moreover, we showed that, in agreement with the mosaic theory of aging, different biological age-related processes measured in vivo have unique spatial patterns throughout the brain. The ability to identify and localize different age-derived processes in vivo may further advance human brain research.


Phantom construction
The full protocol of lipids phantom preparation is described in Shtangel et al.66.

In short, we prepared liposomes from one of the following lipids: phosphatidylserine (PS), phosphatidylcholine (PtdCho), phosphatidylcholine-cholesterol (PtdCho-Chol), Phosphatidylinositol-phosphatidylcholine (PI-PtdCho), or sphingomyelin (Spg). These phantoms were designed to model biological membranes and were prepared from lipids by the hydration–dehydration dry film technique67. The lipids were dissolved over a hot plate and vortexed. Next, the solvent was removed to create a dry film by vacuum-rotational evaporation. The samples were then stirred on a hot plate at 65 °C for 2.5 h to allow the lipids to achieve their final conformation as liposomes. Liposomes were diluted with Dulbecco’s phosphate buffered saline (PBS), without calcium and magnesium (Biological Industries), to maintain physiological conditions in terms of osmolarity, ion concentrations and pH. To change the MTV of the liposome samples we varied the PBS to lipid volume ratios66. Samples were then transferred to the phantom box for scanning in a 4 mL squared polystyrene cuvettes glued to a polystyrene box, which was then filled with ~1% SeaKem Agarose (Ornat Biochemical) and ~0.0005 M Gd (Gadotetrate Melumine, (Dotarem, Guerbet)) dissolved in double distilled water (ddw). The purpose of the agar with Gd (Agar-Gd) was to stabilize the cuvettes, and to create a smooth area in the space surrounding the cuvettes that minimalized air–cuvette interfaces. In some of our experiments we used lipid mixtures composed of several lipids. We prepared nine mixtures containing different combinations of two out of three lipids (PtdChol, Spg and PS) in varying volume ratios (1:1,1:2,2:1). For each mixture, we prepared samples in which the ratio between the different lipid components remained constant while the water-to-lipid volume fraction varied.

For the bovine serum albumin (BSA) phantoms, samples were prepared by dissolving lyophilized BSA powder (Sigma Aldrich) in PBS. To change the MTV of these phantoms, we changed the BSA concentration. For the BSA + Iron phantoms, BSA was additionally mixed with a fixed concentration of 50 µg/mL ferrous sulfate heptahydrate (FeSO4*7H2O). Samples were prepared in their designated concentrations at room temperature. Prepared samples were allowed to sit overnight at 4 ℃ to ensure BSA had fully dissolved, without the need for significant agitation, which is known to cause protein cross-linking. Samples were then transferred to the phantom box for scanning.

For Glucose and Sucrose phantoms, different concentrations of D-( + )-Sucrose (Bio-Lab) and D-( + )-Glucose (Sigma) were dissolved in PBS at 40 ℃. Samples were allowed to reach room temperature before the scan.

MRI acquisition for phantoms

Data was collected on a 3 T Siemens MAGNETOM Skyra scanner equipped with a 32-channel head receive-only coil at the ELSC neuroimaging unit at the Hebrew University.

For quantitative R1 & MTV mapping, three-dimensional (3D) Spoiled gradient (SPGR) echo images were acquired with different flip angles (α = 4°, 8°, 16°, and 30°). The TE/TR was 3.91/18 ms. The scan resolution was 1.1 × 1.1 × 0.9 mm. The same sequence was repeated with a higher resolution of 0.6 × 0.6 × 0.5 mm. The TE/TR was 4.45/18 ms. For calibration, we acquired an additional spin-echo inversion recovery (SEIR) scan. This scan was done on a single slice, with adiabatic inversion pulse and inversion times of TI = 2000, 1200, 800, 400, and 50. The TE/TR was 73/2540 ms. The scan resolution was 1.2 mm isotropic.

For quantitative T2 mapping, images were acquired with a multi spin-echo sequence with 15 equally spaced spin echoes between 10.5 ms and 157.5 ms. The TR was 4.94 s. The scan resolution was 1.2 mm isotropic. For quantitative MTsat mapping, images were acquired with the FLASH Siemens WIP 805 sequence. The TR was 23 ms for all samples except PI:PtdCho for which the TR was 72 ms. Six echoes were equally spaced between 1.93 ms to 14.58 ms. The on-resonance flip angle was 6°, the MT flip angle was 220°, and the RF offset was 700. We used 1.1-mm in-plane resolution with a slice thickness of 0.9 mm. For samples of sucrose and glucose, MTsat mapping was done similar to the human subjects, based on 3D Spoiled gradient (SPGR) echo image with an additional MT pulse. The flip angle was 10°, the TE/TR was 3.91/28 ms. The scan resolution was 1 mm isotropic.

Estimation of qMRI parameters for phantoms

MTV and R1 estimations for the lipids samples were computed based on a the mrQ39 (https://github.com/mezera/mrQ) and Vista Lab (https://github.com/vistalab/vistasoft/wiki) software. The mrQ software was modified to suit the phantom system66. The modification utilizes the fact that the Agar-Gd filling the box around the samples is homogeneous and can, therefore, be assumed to have a constant T1 value. We used this gold standard T1 value generated from the SEIR scan to correct for the excite bias in the spoiled gradient echo scans. While the data was acquired in two different resolutions (see “MRI acquisition”), in our analysis we use the median R1 and MTV of each lipid sample and these are invariant to the resolution of acquisition (Supplementary Fig. 1e). Thus, we were able to use scans with different resolutions without damaging our results. T2 maps were computed by implementing the echo‐modulation curve (EMC) algorithm68.

For quantitative MTsat mapping see the “MTsat estimation” section for human subjects.

MDM computation for phantoms

We computed the dependency of each qMRI parameter (R1, MTsat, and R2) on MTV in different lipids samples. This process was implemented in MATLAB (MathWorks, Natwick, MI, USA). To manipulate the MTV values, we scanned samples of the same lipid in varying concentrations. We computed the median MTV of each sample, along with the median of qMRI parameters. We used these data points to fit a linear model across all samples of the same lipid. The slope of this linear model represents the MTV derivative of the linear equation. We used this derivative estimate of three qMRI parameters (R1, R2, and MTsat) to compute the MDM signatures. The same procedure was used for the MDM computation of lipid mixtures.


MDM modeling of lipid mixtures

We tested the ability of MDM to predict the composition of lipid mixtures. For this analysis we used nine mixture phantoms (see “Phantom construction”), along with the three phantoms of the pure lipid constituents of the mixtures (PS, Spg, and Ptd-Cho).

In order to predict the qMRI parameters of a lipid mixture (Fig. 1d) we used Supplementary Eq. 1 (Supplementary Note 1). To further predict the composition of the mixtures (Fig. 1e) we used Supplementary Eq. 5 (Supplementary Note 2). We solved this equation using the QR factorization algorithm.


Human experiments complied with all relevant ethical regations. The Helsinki Ethics Committee of Hadassah Hospital, Jerusalem, Israel approved the experimental procedure. Written informed consent was obtained from each participant prior to the procedure.

Human subjects

Human measurements were performed on 23 young adults (aged 27 ± 2 years, 11 females), and 18 older adults (aged 67 ± 6 years, five females). Healthy volunteers were recruited from the community surrounding the Hebrew University of Jerusalem.

MRI acquisition for human subjects

Data was collected on a 3 T Siemens MAGNETOM Skyra scanner equipped with a 32-channel head receive-only coil at the ELSC neuroimaging unit at the Hebrew University.

For quantitative R1, R2*, & MTV mapping, 3D Spoiled gradient (SPGR) echo images were acquired with different flip angles (α = 4°, 10°, 20°, and 30°). Each image included five equally spaced echoes (TE = 3.34–14.02 ms) and the TR was 19 ms (except for six young subjects for which the scan included only one TE = 3.34 ms). The scan resolution was 1 mm isotropic. For calibration, we acquired additional spin-echo inversion recovery scan with an echo-planar imaging (EPI) read-out (SEIR-epi). This scan was done with a slab-inversion pulse and spatial-spectral fat suppression. For SEIR-epi, the TE/TR was 49/2920 ms. TI were 200, 400, 1,200, and 2400 ms. We used 2-mm in-plane resolution with a slice thickness of 3 mm. The EPI read-out was performed using 2 × acceleration.

For quantitative T2 mapping, multi‐SE images were acquired with ten equally spaced spin echoes between 12 ms and 120 ms. The TR was 4.21 s. The scan resolution was 2 mm isotropic. T2 scans of four subjects (one young, three old) were excluded from the analysis due to motion.

For quantitative MTsat mapping, 3D Spoiled gradient (SPGR) echo image were acquired with an additional MT pulse. The flip angle was 10°, the TE/TR was 3.34/27 ms. The scan resolution was 1 mm isotropic.

Whole-brain DTI measurements were performed using a diffusion-weighted spin-echo EPI sequence with isotropic 1.5-mm resolution. Diffusion weighting gradients were applied at 64 directions and the strength of the diffusion weighting was set to b = 2000 s/mm2 (TE/TR = 95.80/6000 ms, G = 45mT/m, δ = 32.25 ms, Δ = 52.02 ms). The data includes eight non-diffusion-weighted images (b = 0). In addition, we collected non-diffusion-weighted images with reversed phase-encode blips. For five subjects (four young, one old) we failed to acquire this correction data and they were excluded from the diffusion analysis.

Anatomical images were acquired with 3D magnetization prepared rapid gradient echo (MP-RAGE) scans for 24 of the subjects (14 from the younger subjects, 10 from the older subjects). The scan resolution was 1 mm isotropic, the TE/TR was 2.98/2300 ms. Magnetization Prepared 2 Rapid Acquisition Gradient Echoes (MP2RAGE) scans were acquired for the rest of the subjects. The scan resolution was 1 mm isotropic, the TE/TR was 2.98/5000 ms.

Estimation of qMRI parameters for human subjects

Whole-brain MTV and R1 maps, together with bias correction maps of B1 + and B1-, were computed using the mrQ software39,69 (https://github.com/mezera/mrQ). Voxels in which the B1 + inhomogeneities were extrapolated and not interpolated were removed from the MTV and R1 maps. While we did not correct our MTV estimates for R2*, we showed that employing such a correction does not significantly change our results (see Supplementary Note 6, Supplementary Figs. 20–27). MTV maps of four subjects had bias in the lower part of the brain and they were therefore excluded from the analysis presented in Fig. 3, which includes ROIs in the brainstem.

Whole-brain T2 maps were computed by implementing the echo‐modulation curve (EMC) algorithm68. To combine the MTV and T2 we co-registered the quantitative MTV map to the T2 map. We used the ANTS software package70 to calculate the transformation and to warp the MTV map and the segmentation. The registration was computed to match the T1 map to the T2 map. Next, we applied the calculated transformation to MTV map (since MTV and T1 are in the same imaging space) and resampled the MTV map to match the resolution of the T2 map. The same transformation was also applied to the segmentation. R2 maps were calculated as 1/T2.

Whole-brain MTsat maps were computed as described in Helms et al.37. The MTsat measurement was extracted from Eq. (1):

Where SMT is the signal of the SPGR scan with additional MT pulse, α is the flip angle and TR is the repetition time. Mo (the equilibrium magnetization parameter), B1 (the transmit inhomogeneity), and R1 estimations were computed from the non-MT weighted SPGR scans, during the pipeline described under “MTV & R1 estimation”. Registration of the SMT image to the imaging space of the MTV map was done using a rigid-body alignment (R1, B1, and MO are all in the same space as MTV).

Diffusion analysis was done using the FDT toolbox in FSL71,72. Susceptibility and eddy current induced distortions were corrected using the reverse phase-encode data, with the eddy and topup commands73,74. MD maps were calculated using vistasoft (https://github.com/vistalab/vistasoft/wiki). We used a rigid-body alignment to register the corrected dMRI data to the imaging space of the MTV map (Flirt, FSL). In order to calculate the MD-MTV derivatives, we resampled the MTV map and the segmentation to match the dMRI resolution.

We used the SPGR scans with multiple echoes to estimate R2*. Fitting was done through the MPM toolbox75. As we had four SPGR scans with variable flip angles, we averaged the R2* maps acquired from each of these scans for increased SNR.

Human brain segmentation

Whole-brain segmentation was computed automatically using the FreeSurfer segmentation algorithm76. For subjects who had an MP-RAGE scan, we used it as a reference. For the other subjects the MP2RAGE scan was used as a reference. These anatomical images were registered to the MTV space prior to the segmentation process, using a rigid-body alignment. Sub-cortical gray-matter structures were segmented with FSL’s FIRST tool77. To avoid partial volume effects, we removed the outer shell of each ROI and left only the core.

MDM computation in the human brain

We computed the dependency of each qMRI parameter (R1, MTsat, MD, and R2) on MTV in different brain areas. This process was implemented in MATLAB (MathWorks, Natwick, MI, USA). For each ROI, we extracted the MTV values from all voxels and pooled them into 36 bins spaced equally between 0.05 and 0.40. This was done so that the linear fit would not be heavily affected by the density of the voxels in different MTV values. We removed any bins in which the number of voxels was smaller than 4% of the total voxel count in the ROI. The median MTV of each bin was computed, along with the median of the qMRI parameter. We used these data points to fit the linear model across bins using Eq. (2):

The slope of this linear model (“a”) represents the MTV derivative of the linear equation. We used this derivative estimate to compute the MDM signatures.

For each subject, ROIs in which the total voxel count was smaller than a set threshold of 500 voxels for the MTsat and R1 maps, 150 voxels for the MD map, and 50 voxels for the R2 map were excluded.

Principal component analysis (PCA) in the human brain

To estimate the variability in the MDM signatures across the brain, we computed the first principal component (PC) of MDM. For each MDM dimension (MTV derivatives of R1, MTsat, MD, and R2), we evaluated the median of the different brain areas across the young subjects. As each MDM dimension has different units, we then computed the z-score of each dimension across the different brain area. Finally, we performed PCA. The variables in this analysis were the different MDM dimensions, and the observations were the different brain areas. From this analysis, we derived the first PC that accounts for most of the variability in MDM signatures across the brain. To estimate the median absolute deviations (MAD) across subjects of each MDM measurement in the PC basis, we applied the z-score transformation to the original MAD and then projected them onto the PC basis.

To compute the first PC of standard qMRI parameters we followed the same procedure, but used R1, MTsat, MD, and R2 instead of their MTV derivatives.

For the first PC of molecular composition, we followed the same procedure, but used the phospholipid composition and the ratio between phospholipids to proteins and cholesterol as variables. The data was taken from eight post-mortem human brains7. Brains were obtained from individuals between 54 and 57 years of age, which were autopsied within 24 h after death.

Linear model for prediction of human molecular composition

We used MDM measurements in order to predict the molecular composition of different brain areas (Fig. 3c). For this analysis we used Supplementary Eq. 5 in the Supplementary Note 2. We solved this equation using QR factorization algorithm (for more details see Supplementary Note 3).

Gene-expression dataset

For the gene-expression analysis we followed the work of Ben-David and Shifman46. Microarray data was acquired from the Allen Brain Atlas (http://human.brain-map.org/well_data_files) and included a total of 1340 microarray profiles from donors H0351.2001 and H0351.2002, encompassing the different regions of the human brain. The donors were 24 and 39 years old, respectively, at the time of their death, with no known psychopathologies. We used the statistical analysis described by Ben-David and Shifman46. They constructed a gene network using a weighted gene co-expression network analysis. The gene network included 19 modules of varying sizes, from 38 to 7385 genes. The module eigengenes were derived by taking the first PC of the expression values in each module. In addition, we used the gene ontology enrichment analysis described by Ben-David and Shifman to define the name of each module. The colors of the different modules in the Fig. 4 and Supplementary Fig. 10 are the same as in the original paper.

Next, we matched between the gene-expression data and the MRI measurements. This analysis was done on 35 cortical regions extracted from FreeSurfer cortical parcellation. We downloaded the T1-weighted images of the two donors provided by the Allen Brain Atlas (http://human.brain-map.org/mri_viewers/data) and used them as a reference for FreeSurfer segmentation. We then found the FreeSurfer label of each gene-expression sample using the sample’s coordinates in brain space. We removed samples for which the FreeSurfer label and the label provided in the microarray dataset did not agree (there were 72 such samples out of 697 cortical samples). For each gene module, we averaged over the eigengenes of all samples from the same cortical area across the two donors.

Last, we compared the cortical eigengene of each module to the projection of cortical areas on the first PC of MDM. In addition, we compared the modules’ eigengenes to the MTV values of the cortical areas and to the projection of cortical areas on the first PC of standard qMRI parameters (Supplementary Fig. 10). These 57 correlations were corrected for multiple comparisons using the FDR method.

Brain region’s volume computation

To estimate the volume of different brain regions, we calculated the number of voxels in the FreeSurfer segmentation of each region (see “Brain segmentation”).

R2* correction for MTV
To correct the MTV estimates for R2* we used Eq. (3):

Where MTVC is the corrected MTV.

Statistical analysis

The statistical significance of the differences between the age groups was computed using an independent-sample t-test (alpha = 0.05, both right and left tail) and was corrected for multiple comparisons using the false-discovery rate (FDR) method. For this analysis, MRI measurements of both hemispheres of bilateral brain regions were joined together. R2 measurements were adjusted for the number of data points. All statistical tests were two-sided.

Post-mortem tissue acquisition

Two post-mortem porcine brains were purchased from BIOTECH FARM.

Post-mortem MRI acquisition

Brains were scanned fresh (without fixation) in water within 6 h after death. Data was collected on a 3 T Siemens MAGNETOM Skyra scanner equipped with a 32-channel head receive-only coil at the ELSC neuroimaging unit at the Hebrew University.

For quantitative R1, R2*, & MTV mapping, 3D Spoiled gradient (SPGR) echo images were acquired with different flip angles (α = 4°, 10°, 20°, and 30°). Each image included five equally spaced echoes (TE = 4.01 – 16.51 ms) and the TR was 22 ms. The scan resolution was 0.8 mm isotropic. For calibration, we acquired additional spin-echo inversion recovery scan with an echo-planar imaging (EPI) read-out (SEIR-epi). This scan was done with a slab-inversion pulse and spatial-spectral fat suppression. For SEIR-epi, the TE/TR was 49/2920 ms. TI were 50, 200, 400, 1200 ms. The scan resolution was 2 mm isotropic. The EPI read-out was performed using 2 × acceleration.

For quantitative T2 mapping, multi‐SE images were acquired with ten equally spaced spin echoes between 12 and 120 ms. The TR was 4.21 s. The scan resolution was 2 mm isotropic.

For quantitative MTsat mapping, 3D Spoiled gradient (SPGR) echo image were acquired with an additional MT pulse. The flip angle was 10°, the TE/TR was 4.01/40 ms. The scan resolution was 0.8 mm isotropic.

Whole-brain DTI measurements were performed using a diffusion-weighted spin-echo EPI sequence with isotropic 1.5-mm resolution. Diffusion weighting gradients were applied at 64 directions and the strength of the diffusion weighting was set to b = 2000 s/mm2 (TE/TR = 95.80/6000 ms, G = 45mT/m, δ = 32.25 ms, Δ = 52.02 ms). The data includes eight non-diffusion-weighted images (b = 0).

For anatomical images, 3D magnetization prepared rapid gradient echo (MP-RAGE) scans were acquired. The scan resolution was 1 mm isotropic, the TE/TR was 2.98/2300 ms.

Histological analysis

Following the MRI scans the brains were dissected. Total of 42 brain regions were identified. Four samples were excluded as we were not able to properly separate the WM from the GM. One sample was excluded as we could not properly identify its anatomical origin. Additional two samples were too small for TLC analysis.

The non-water fraction (MTV) was determined by desiccation, also known as the dry-wet method. A small fraction of each brain sample (~0.25 g) was weighed. In order to completely dehydrate the fresh tissues, they were left for several days in a vacuum dessicator over silica gel at 4 °C. The experiment ended when no further weight loss occurred. The MTV of each brain sample was calculated based on the difference between the wet (Wwet) and dry (Wdry) weights of the tissue (Eq. 4):

For lipid extraction and lipidomics analysis78, Brain samples were weighted and homogenized with saline in plastic tubes on ice at concentration of 1 mg/12.5 µL. Two-hundred fifty microliters from each homogenate were utilized for lipid extraction and analysis with thin-layer chromatography (TLC). The lipid species distribution was analyzed by TLC applying 150 µg aliquots. Samples were reconstituted in 10 µL of Folch mixture and spotted on Silica-G TLC plates. Standards for each fraction were purchased from Sigma Aldrich (Rehovot, Israel) and were spotted in separate TLC lanes, i.e., 50 µg of triacylglycerides (TG), cholesterol (Chol), cholesteryl esters (CE), free fatty acids (FFA), lysophospholipids (Lyso), sphingomyelin (Spg), phosphatidylcholine (PtdCho), phosphatidylinositol (PI), phosphatidylserine (PS), and phosphatidylethanolamine (PE). Plates were then placed in a 20 × 20 cm TLC chamber containing petroleum ether, ethyl ether, and acetic acid (80:20:1, v/v/v) for quantification of neutral lipids or chloroform, methanol, acetic acid, and water (65:25:4:2, v:v:v:v) for quantification of polar lipids and run for 45 min. TG, Chol, CE, FFA, phospholipids (PL), Lyso, Spg, PtdCho, PI, PS, and PE bands were visualized with Iodine, scanned and quantified by Optiquant after scanning (Epson V700). Lyso, CE, TG, and PI were excluded from further analysis as their quantification was noisy and demonstrated high variability across TLC plates. This analysis was conducted under the guidance of Prof. Alicia Leikin-Frenkel in the Bert Strassburger Lipid Center, Sheba, Tel Hashomer.

Estimation of qMRI parameters in the post-mortem brain

Similar to human subjects.

Brain segmentation of post-mortem brain

Brain segmentation was done manually. Five tissue samples were excluded as we could not identify their origin location in the MRI scans.

MDM computation in the post-mortem brain

We computed the dependency of each qMRI parameter (R1, MTsat, MD, and R2) on MTV in different brain areas similarly to the analysis of the human subjects.

Principal component analysis (PCA) in the post-mortem brain

To estimate the variability in the MDM signatures across the brain, we computed the first principal component (PC) of MDM. PCA analysis was performed with four variables corresponding to the MDM dimensions (MTV derivatives of R1, MTsat, MD, and R2), and 30 observations corresponding to the different brain regions. As each MDM dimension has different units, we first computed the z-score of each dimension across the different brain areas prior to the PCA. From this analysis we derived the first PC that accounts for most of the variability in MDM signatures across the brain.

To compute the first PC of standard qMRI parameters we followed the same procedure, but used R1, MTsat, MD, and R2 instead of their MTV derivatives.

To estimate the variability in the lipid composition across the brain, we computed the first principal component (PC) of lipidomics. PCA analysis was performed with seven variables corresponding to the different polar and neutral lipids (Chol, FFA, PL, Spg, PtdCho, PS, PE), and 30 observations corresponding to the different brain regions. From this analysis, we derived the first PC that accounts for most of the variability in lipid composition across the brain.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Code availability

A toolbox for computing MDM signatures is available at [https://github.com/MezerLab/MDM_toolbox].

The code generating the figures of in the paper is available at [https://github.com/MezerLab/MDM_Gen_Figs].

Peters, R. Ageing and the brain. Postgrad. Med. J. 82, 84–88 (2006).

Lockhart, S. N. & DeCarli, C. Structural imaging measures of brain aging. Neuropsychol. Rev. 24, 271–289 (2014).

Wozniak, J. R. & Lim, K. O. Advances in white matter imaging: a review of in vivo magnetic resonance methodologies and their applicability to the study of development and aging. Neurosci. Biobehav. Rev. 30, 762–774 (2006).

Frisoni, G. B., Fox, N. C., Jack, C. R., Scheltens, P. & Thompson, P. M. The clinical use of structural MRI in Alzheimer disease. Nat. Rev. Neurol. 6, 67–77 (2010).

Mrak, R. E., Griffin, S. T. & Graham, D. I. Aging-associated changes in human brain. J. Neuropathol. Exp. Neurol. 56, 1269–1275 (1997).

Yankner, B. A., Lu, T. & Loerch, P. The aging brain. Annu. Rev. Pathol. 3, 41–66 (2008).

Söderberg, M., Edlund, C., Kristensson, K. & Dallner, G. Lipid compositions of different regions of the human brain during aging. J. Neurochem. 54, 415–423 (1990).

Lauwers, E. et al. Membrane lipids in presynaptic function and disease. Neuron 90, 11–25 (2016).

Li, Q. et al. Changes in lipidome composition during brain development in humans, chimpanzees, and Macaque monkeys. Mol. Biol. Evol. 34, 1155–1166 (2017).

Müller, C. P. et al. Brain membrane lipids in major depression and anxiety disorders. Biochim. Biophys. Acta-Mol. Cell Biol. Lipids 1851, 1052–1065 (2015).

Naudí, A. et al. Lipidomics of human brain aging and Alzheimer’s disease pathology. Int. Rev. Neurobiol. 122, 133–189 (2015).

Walker, L. C. & Herndon, J. G. Mosaic aging. Med. Hypotheses 74, 1048–1051 (2010).

Cole, J. H., Marioni, R. E., Harris, S. E. & Deary, I. J. Brain age and other bodily ‘ages’: implications for neuropsychiatry. Mol. Psychiatry 1 (2018). https://doi.org/10.1038/s41380-018-0098-1.

Hayflick, L. Biological aging is no longer an unsolved problem. Ann. N. Y. Acad. Sci. 1100, 1–13 (2007).

Christensen, H., Mackinnon, A. J., Korten, A. & Jorm, A. F. The ‘common cause hypothesis’; of cognitive aging: evidence for not only a common factor but also specific associations of age with vision and grip strength in a cross-sectional analysis. Psychol. Aging 16, 588–599 (2001).

Cole, J. H. et al. Brain age predicts mortality. Mol. Psychiatry 23, 1385–1392 (2018).

Sowell, E. R., Thompson, P. M. & Toga, A. W. Mapping changes in the human cortex throughout the span of life. Neuroscience 10, 372–392 (2004).

Fjell, A. M. & Walhovd, K. B. Structural brain changes in aging: courses, causes and cognitive consequences. Rev. Neurosci. 21, 187–221 (2010).

Gunning-Dixon, F. M., Brickman, A. M., Cheng, J. C. & Alexopoulos, G. S. Aging of cerebral white matter: a review of MRI findings. Int. J. Geriatr. Psychiatry 24, 109–117 (2009).

Callaghan, M. F. et al. Widespread age-related differences in the human brain microstructure revealed by quantitative magnetic resonance imaging. Neurobiol. Aging 35, 1862–1872 (2014).

Yeatman, J. D., Wandell, B. A. & Mezer, A. A. Lifespan maturation and degeneration of human brain white matter. Nat. Commun. 5, 4932 (2014).

Cox, S. R. et al. Ageing and brain white matter structure in 3,513 UK Biobank participants. Nat. Commun. 7, 13629 (2016).

Lorio, S. et al. Disentangling in vivo the effects of iron content and atrophy on the ageing human brain. Neuroimage 103, 280–289 (2014).

Gracien, R.-M. et al. Evaluation of brain ageing: a quantitative longitudinal MRI study over 7 years. Eur. Radiol. 27, 1568–1576 (2017).

Draganski, B. et al. Regional specificity of MRI contrast parameter changes in normal ageing revealed by voxel-based quantification (VBQ). Neuroimage 55, 1423–1434 (2011).

Tardif, C. L. et al. Investigation of the confounding effects of vasculature and metabolism on computational anatomy studies. Neuroimage 149, 233–243 (2017).

Carey, D. et al. Quantitative MRI provides markers of intra-, inter-regional, and age-related differences in young adult cortical microstructure. Neuroimage 182, 429–440 (2017).

Cercignani, M., Dowell, N. G. & Tofts, P. S. Quantitative MRI of the Brain: Principles of Physical Measurement. (CRC Press, United States, 2018).

Basser, P. J. & Pierpaoli, C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J. Magn. Reson. Ser. B 111, 209–219 (1996).

Weiskopf, N., Mohammadi, S., Lutti, A. & Callaghan, M. F. Advances in MRI-based computational neuroanatomy. Curr. Opin. Neurol. 28, 313–322 (2015).

Winklewski, P. J. et al. Understanding the physiopathology behind axial and radial diffusivity changes—what do we know? Front. Neurol. 9, 92 (2018).

Heath, F., Hurley, S. A., Johansen-Berg, H. & Sampaio-Baptista, C. Advances in noninvasive myelin imaging. Dev. Neurobiol. 78, 136–151 (2018).

Lutti, A., Dick, F., Sereno, M. I. & Weiskopf, N. Using high-resolution quantitative mapping of R1 as an index of cortical myelination. Neuroimage 93, 176–188 (2014).

Filo, S. & Mezer, A. A. in Quantitative MRI of the Brain: Principles of Physical Measurement (eds Cercignani, M., Dowell, N. G. & Tofts, P. S.) 55–72 (CRC Press, United States, 2018).

Fullerton, G. D., Cameron, I. L. & Ord, V. A. Frequency dependence of magnetic resonance spin-lattice relaxation of protons in biological materials. Radiology 151, 135–138 (1984).

Does, M. D. Inferring brain tissue composition and microstructure via MR relaxometry. Neuroimage 182, 136–148 (2018).

Helms, G., Dathe, H., Kallenberg, K. & Dechent, P. High-resolution maps of magnetization transfer with inherent correction for RF inhomogeneity and T 1 relaxation obtained from 3D FLASH MRI. Magn. Reson. Med. 60, 1396–1407 (2008).

Rohrer, M., Bauer, H., Mintorovitch, J., Requardt, M. & Weinmann, H. -J. Comparison of magnetic properties of MRI contrast media solutions at different magnetic field strengths. Investig. Radiol. 40, 715–724 (2005).

Mezer, A. et al. Quantifying the local tissue volume and composition in individual brains with magnetic resonance imaging. Nat. Med. 19, 1667–1672 (2013).

Koenig, S. H. Cholesterol of myelin is the determinant of gray‐white contrast in MRI of brain. Magn. Reson. Med. 20, 285–291 (1991).

Koenig, S. H., Brown, R. D., Spiller, M. & Lundbom, N. Relaxometry of brain: why white matter appears bright in MRI. Magn. Reson. Med. 14, 482–495 (1990).

Kucharczyk, W., Macdonald, P. M., Stanisz, G. J. & Henkelman, R. M. Relaxivity and magnetization transfer of white matter lipids at MR imaging: importance of cerebrosides and pH. Radiology 192, 521–529 (1994).

Fullerton, G. D., Potter, J. L. & Dornbluth, N. C. NMR relaxation of protons in tissues and other macromolecular water solutions. Magn. Reson. Imaging 1, 209–226 (1982).

Morawski, M. et al. Developing 3D microscopy with CLARITY on human brain tissue: towards a tool for informing and validating MRI-based histology. Neuroimage 182, 417–428 (2018).

Leuze, C. et al. The separate effects of lipids and proteins on brain MRI contrast revealed through tissue clearing. Neuroimage 156, 412–422 (2017).

Ben-David, E. & Shifman, S. Networks of neuronal genes affected by common and rare variants in autism spectrum disorders. PLoS Genet. 8, e1002556 (2012).

Zecca, L., Youdim, M. B. H., Riederer, P., Connor, J. R. & Crichton, R. R. Iron, brain ageing and neurodegenerative disorders. Nat. Rev. Neurosci. 5, 863–873 (2004).

Langkammer, C. et al. Quantitative MR imaging of brain iron: a postmortem validation study. Radiology 257, 455–462 (2010).

Freeman, S. H. et al. Preservation of neuronal number despite age-related cortical brain atrophy in elderly subjects without Alzheimer disease. J. Neuropathol. Exp. Neurol. 67, 1205–1212 (2008).

Burke, S. N. & Barnes, C. A. Neural plasticity in the ageing brain. Nat. Rev. Neurosci. 7, 30–40 (2006).

Bowley, M. P., Cabral, H., Rosene, D. L. & Peters, A. Age changes in myelinated nerve fibers of the cingulate bundle and corpus callosum in the rhesus monkey. J. Comp. Neurol. 518, 3046–3064 (2010).

Callaghan, M. F., Helms, G., Lutti, A., Mohammadi, S. & Weiskopf, N. A general linear relaxometry model of R1 using imaging data. Magn. Reson. Med. 73, 1309–1314 (2015).

Piomelli, D., Astarita, G. & Rapaka, R. A neuroscientist’s guide to lipidomics. Nat. Rev. Neurosci. 8, 743–754 (2007).

Sethi, S., Hayashi, M. A., Sussulini, A., Tasic, L. & Brietzke, E. Analytical approaches for lipidomics and its potential applications in neuropsychiatric disorders. World J. Biol. Psychiatry 18, 506–520 (2017).

Fantini, J. & Yahi, N. Brain Lipids in Synaptic Function and Neurological Disease: Clues to Innovative Therapeutic Strategies for Brain Disorders. (Academic Press, United States, 2015).

Shinitzky, M. Patterns of lipid changes in membranes of the aged brain. Gerontology 33, 149–154 (1987).

Martin, M., Dotti, C. G. & Ledesma, M. D. Brain cholesterol in normal and pathological aging. Biochim. Biophys. Acta-Mol. Cell Biol. Lipids 1801, 934–944 (2010).

Calucci, L. & Forte, C. Proton longitudinal relaxation coupling in dynamically heterogeneous soft systems. Prog. Nucl. Magn. Reson. Spectrosc. 55, 296–323 (2009).

Halle, B. Molecular theory of field-dependent proton spin-lattice relaxation in tissue. Magn. Reson. Med. 56, 60–72 (2006).

West, M. J., Coleman, P. D., Flood, D. G. & Troncoso, J. C. Differences in the pattern of hippocampal neuronal loss in normal ageing and Alzheimer’s disease. Lancet (Lond., Engl.) 344, 769–772 (1994).

West, M. J., Kawas, C. H., Stewart, W. F., Rudow, G. L. & Troncoso, J. C. Hippocampal neurons in pre-clinical Alzheimer’s disease. Neurobiol. Aging 25, 1205–1212 (2004).

Slater, D. A. et al. Evolution of white matter tract microstructure across the life span. Hum. Brain Mapp. 40, 2252–2268 (2019).

Jarmusch, A. K. et al. Lipid and metabolite profiles of human brain tumors by desorption electrospray ionization-MS. Proc. Natl Acad. Sci. U.S.A. 113, 1486–1491 (2016).

Wenk, M. R. The emerging field of lipidomics. Nat. Rev. Drug Discov. 4, 594–610 (2005).

Eberlin, L. S. et al. Classifying human brain tumors by lipid imaging with mass spectrometry. Cancer Res. 72, 645–654 (2012).

Shtangel, O. & Mezer, A. A phantom system designed to assess the effects of membrane lipids on water proton relaxation. bioRxiv 387845 (2018). https://doi.org/10.1101/387845.

Akbarzadeh, A. et al. Liposome: methods of preparation and applications. Liposome Technol. 6, 102 (2013).

Ben-Eliezer, N., Sodickson, D. K. & Block, K. T. Rapid and accurate T 2 mapping from multi-spin-echo data using Bloch-simulation-based reconstruction. Magn. Reson. Med. 73, 809–817 (2015).

Mezer, A., Rokem, A., Berman, S., Hastie, T. & Wandell, B. A. Evaluating quantitative proton-density-mapping methods. Hum. Brain Mapp. 37, 3623–3635 (2016).

Avants, B. B., Tustison, N. & Song, G. Advanced normalization tools (ANTS). Insight J. (2009). http://hdl.handle.net/10380/3113

Smith, S. M. et al. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage (2004). https://doi.org/10.1016/j.neuroimage.2004.07.051.

Behrens, T. E. J. et al. Characterization and propagation of uncertainty in diffusion-weighted MR imaging. Magn. Reson. Med. (2003). https://doi.org/10.1002/mrm.10609.

Andersson, J. L. R., Skare, S. & Ashburner, J. How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. Neuroimage (2003). https://doi.org/10.1016/S1053-8119(03)00336-7.

Andersson, J. L. R. & Sotiropoulos, S. N. An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage (2016). https://doi.org/10.1016/j.neuroimage.2015.10.019.

Weiskopf, N. et al. Quantitative multi-parameter mapping of R1, PD*, MT, and R2* at 3T: a multi-center validation. Front. Neurosci. (2013). https://doi.org/10.3389/fnins.2013.00095.

Fischl, B. FreeSurfer. Neuroimage 62, 774–781 (2012).

Patenaude, B., Smith, S. M., Kennedy, D. N. & Jenkinson, M. A Bayesian model of shape and appearance for subcortical brain segmentation. Neuroimage (2011). https://doi.org/10.1016/j.neuroimage.2011.02.046.

Shomonov-Wagner, L., Raz, A. & Leikin-Frenkel, A. Alpha linolenic acid in maternal diet halts the lipid disarray due to saturated fatty acids in the liver of mice offspring at weaning. Lipids Health Dis. (2015). https://doi.org/10.1186/s12944-015-0012-7.

Download references


This work was supported by the ISF grant 0399306, awarded to A.A.M. We acknowledge Ady Zelman for the assistance in collecting the human MRI data. We thank Assaf Friedler for assigning research lab space and advising on the lipid sample experiments. We thank Inbal Goshen for assigning research lab space and advising on the protein and ion samples as well as the porcine brain experiments. We thank Magnus Soderberg for advising on histological data interpretation. We are grateful to Brian A. Wandell, Jason Yeatman, Hermona Soreq, Ami Citri, Mark Does, Yaniv Ziv, Ofer Yizhar, Shai Berman, Roey Schurr, Jonathan Bain, Asier Erramuzpe Aliaga, Menachem Gutman, and Esther Nachliel for their critical reading of the manuscript and very useful comments. We thank Prof. Alicia Leikin-Frenkel for her guidance with the TLC analysis. We thank Rona Shaharabani for guidance and support in the post-mortem experiments.


The Edmond and Lily Safra Center for Brain Sciences, The Hebrew University of Jerusalem, Jerusalem, 9190401, Israel
Shir Filo, Oshrat Shtangel, Noga Salamon, Adi Kol, Batsheva Weisinger & Aviv A. Mezer
Department of Genetics, The Institute of Life Sciences, The Hebrew University of Jerusalem, Jerusalem, 9190401, Israel
Sagiv Shifman
S.F., O.S., and A.A.M. conceived of the presented idea. S.F. and A.A.M. wrote the manuscript and designed the figures. S.F. collected the human and non-human brain datasets and analyzed them. O.S. performed the phantom experiments and analyzed them. B.W. performed the phantom experiments for non-lipid compounds. N.S. performed the gene-expression analysis. S.S. assisted and instructed with the gene-expression analysis. A.K. performed the porcine brain dissection.

Corresponding author

Correspondence to Aviv A. Mezer.

Ethics declarations & Competing interests

A.A.M, S.F., O.S. and the Hebrew University of Jerusalem have filed a patent application describing the technology used to measure MDM in this work. The other authors declare no competing interests.

Read Full Post »

Artificial Intelligence and Cardiovascular Disease

Reporter and Curator: Dr. Sudipta Saha, Ph.D.


Cardiology is a vast field that focuses on a large number of diseases specifically dealing with the heart, the circulatory system, and its functions. As such, similar symptomatologies and diagnostic features may be present in an individual, making it difficult for a doctor to easily isolate the actual heart-related problem. Consequently, the use of artificial intelligence aims to relieve doctors from this hurdle and extend better quality to patients. Results of screening tests such as echocardiograms, MRIs, or CT scans have long been proposed to be analyzed using more advanced techniques in the field of technology. As such, while artificial intelligence is not yet widely-used in clinical practice, it is seen as the future of healthcare.


The continuous development of the technological sector has enabled the industry to merge with medicine in order to create new integrated, reliable, and efficient methods of providing quality health care. One of the ongoing trends in cardiology at present is the proposed utilization of artificial intelligence (AI) in augmenting and extending the effectiveness of the cardiologist. This is because AI or machine-learning would allow for an accurate measure of patient functioning and diagnosis from the beginning up to the end of the therapeutic process. In particular, the use of artificial intelligence in cardiology aims to focus on research and development, clinical practice, and population health. Created to be an all-in-one mechanism in cardiac healthcare, AI technologies incorporate complex algorithms in determining relevant steps needed for a successful diagnosis and treatment. The role of artificial intelligence specifically extends to the identification of novel drug therapies, disease stratification or statistics, continuous remote monitoring and diagnostics, integration of multi-omic data, and extension of physician effectivity and efficiency.


Artificial intelligence – specifically a branch of it called machine learning – is being used in medicine to help with diagnosis. Computers might, for example, be better at interpreting heart scans. Computers can be ‘trained’ to make these predictions. This is done by feeding the computer information from hundreds or thousands of patients, plus instructions (an algorithm) on how to use that information. This information is heart scans, genetic and other test results, and how long each patient survived. These scans are in exquisite detail and the computer may be able to spot differences that are beyond human perception. It can also combine information from many different tests to give as accurate a picture as possible. The computer starts to work out which factors affected the patients’ outlook, so it can make predictions about other patients.


In current medical practice, doctors will use risk scores to make treatment decisions for their cardiac patients. These are based on a series of variables like weight, age and lifestyle. However, they do not always have the desired levels of accuracy. A particular example of the use of artificial examination in cardiology is the experimental study on heart disease patients, published in 2017. The researchers utilized cardiac MRI-based algorithms coupled with a 3D systolic cardiac motion pattern to accurately predict the health outcomes of patients with pulmonary hypertension. The experiment proved to be successful, with the technology being able to pick-up 30,000 points within the heart activity of 250 patients. With the success of the aforementioned study, as well as the promise of other researches on artificial intelligence, cardiology is seemingly moving towards a more technological practice.


One study was conducted in Finland where researchers enrolled 950 patients complaining of chest pain, who underwent the centre’s usual scanning protocol to check for coronary artery disease. Their outcomes were tracked for six years following their initial scans, over the course of which 24 of the patients had heart attacks and 49 died from all causes. The patients first underwent a coronary computed tomography angiography (CCTA) scan, which yielded 58 pieces of data on the presence of coronary plaque, vessel narrowing and calcification. Patients whose scans were suggestive of disease underwent a positron emission tomography (PET) scan which produced 17 variables on blood flow. Ten clinical variables were also obtained from medical records including sex, age, smoking status and diabetes. These 85 variables were then entered into an artificial intelligence (AI) programme called LogitBoost. The AI repeatedly analysed the imaging variables, and was able to learn how the imaging data interacted and identify the patterns which preceded death and heart attack with over 90% accuracy. The predictive performance using the ten clinical variables alone was modest, with an accuracy of 90%. When PET scan data was added, accuracy increased to 92.5%. The predictive performance increased significantly when CCTA scan data was added to clinical and PET data, with accuracy of 95.4%.


Another study findings showed that applying artificial intelligence (AI) to the electrocardiogram (ECG) enables early detection of left ventricular dysfunction and can identify individuals at increased risk for its development in the future. Asymptomatic left ventricular dysfunction (ALVD) is characterised by the presence of a weak heart pump with a risk of overt heart failure. It is present in three to six percent of the general population and is associated with reduced quality of life and longevity. However, it is treatable when found. Currently, there is no inexpensive, noninvasive, painless screening tool for ALVD available for diagnostic use. When tested on an independent set of 52,870 patients, the network model yielded values for the area under the curve, sensitivity, specificity, and accuracy of 0.93, 86.3 percent, 85.7 percent, and 85.7 percent, respectively. Furthermore, in patients without ventricular dysfunction, those with a positive AI screen were at four times the risk of developing future ventricular dysfunction compared with those with a negative screen.


In recent years, the analysis of big data database combined with computer deep learning has gradually played an important role in biomedical technology. For a large number of medical record data analysis, image analysis, single nucleotide polymorphism difference analysis, etc., all relevant research on the development and application of artificial intelligence can be observed extensively. For clinical indication, patients may receive a variety of cardiovascular routine examination and treatments, such as: cardiac ultrasound, multi-path ECG, cardiovascular and peripheral angiography, intravascular ultrasound and optical coherence tomography, electrical physiology, etc. By using artificial intelligence deep learning system, the investigators hope to not only improve the diagnostic rate and also gain more accurately predict the patient’s recovery, improve medical quality in the near future.


The primary issue about using artificial intelligence in cardiology, or in any field of medicine for that matter, is the ethical issues that it brings about. Physicians and healthcare professionals prior to their practice swear to the Hippocratic Oath—a promise to do their best for the welfare and betterment of their patients. Many physicians have argued that the use of artificial intelligence in medicine breaks the Hippocratic Oath since patients are technically left under the care of machines than of doctors. Furthermore, as machines may also malfunction, the safety of patients is also on the line at all times. As such, while medical practitioners see the promise of artificial technology, they are also heavily constricted about its use, safety, and appropriateness in medical practice.


Issues and challenges faced by technological innovations in cardiology are overpowered by current researches aiming to make artificial intelligence easily accessible and available for all. With that in mind, various projects are currently under study. For example, the use of wearable AI technology aims to develop a mechanism by which patients and doctors could easily access and monitor cardiac activity remotely. An ideal instrument for monitoring, wearable AI technology ensures real-time updates, monitoring, and evaluation. Another direction of cardiology in AI technology is the use of technology to record and validate empirical data to further analyze symptomatology, biomarkers, and treatment effectiveness. With AI technology, researchers in cardiology are aiming to simplify and expand the scope of knowledge on the field for better patient care and treatment outcomes.






























Read Full Post »

These twelve artificial intelligence innovations are expected to start impacting clinical care by the end of the decade.

Reporter: Gail S. Thornton, M.A.


This article is excerpted from Health IT Analytics, April 11, 2019.

 By Jennifer Bresnick

April 11, 2019 – There’s no question that artificial intelligence is moving quickly in the healthcare industry.  Even just a few months ago, AI was still a dream for the next generation: something that would start to enter regular care delivery in a couple of decades – maybe ten or fifteen years for the most advanced health systems.

Even Partners HealthCare, the Boston-based giant on the very cutting edge of research and reform, set a ten-year timeframe for artificial intelligence during its 2018 World Medical Innovation Forum, identifying a dozen AI technologies that had the potential to revolutionize patient care within the decade.

But over the past twelve months, research has progressed so rapidly that Partners has blown up that timeline. 

Instead of viewing AI as something still lingering on the distant horizon, this year’s Disruptive Dozen panel was tasked with assessing which AI innovations will be ready to fundamentally alter the delivery of care by 2020 – now less than a year away.

Sixty members of the Partners faculty participated in nominating and narrowing down the tools they think will have an almost immediate benefit for patients and providers, explained Erica Shenoy, MD, PhD, an infectious disease specialist at Massachusetts General Hospital (MGH).

“These are innovations that have a strong potential to make significant advancement in the field, and they are also technologies that are pretty close to making it to market,” she said.

The results include everything from mental healthcare and clinical decision support to coding and communication, offering patients and their providers a more efficient, effective, and cost-conscious ecosystem for improving long-term outcomes.

In order from least to greatest potential impact, here are the twelve artificial intelligence innovations poised to become integral components of the next decade’s data-driven care delivery system.


Nearly twenty percent of US patients struggle with a mental health disorder, yet treatment is often difficult to access and expensive to use regularly.  Reducing barriers to access for mental and behavioral healthcare, especially during the opioid abuse crisis, requires a new approach to connecting patients with services.

AI-driven applications and therapy programs will be a significant part of the answer.

“The promise and potential for digital behavioral solutions and apps is enormous to address the gaps in mental healthcare in the US and across the world,” said David Ahern, PhD, a clinical psychologist at Brigham & Women’s Hospital (BWH). 

Smartphone-based cognitive behavioral therapy and integrated group therapy are showing promise for treating conditions such as depression, eating disorders, and substance abuse.

While patients and providers need to be wary of commercially available applications that have not been rigorously validated and tested, more and more researchers are developing AI-based tools that have the backing of randomized clinical trials and are showing good results.

A panel of experts from Partners HealthCare presents the Disruptive Dozen at WMIF19.
A panel of experts from Partners HealthCare presents the Disruptive Dozen at WMIF19.

Source: Partners HealthCare


Natural language processing is already a routine part of many behind-the-scenes clinical workflows, but voice-first tools are expected to make their way into the patient-provider encounter in a new way. 

Smart speakers in the clinic are prepping to relieve clinicians of their EHR burdens, capturing free-form conversations and translating the content into structured documentation.  Physicians and nurses will be able to collect and retrieve information more quickly while spending more time looking patients in the eye.

Patients may benefit from similar technologies at home as the consumer market for virtual assistants continues to grow.  With companies like Amazon achieving HIPAA compliance for their consumer-facing products, individuals may soon have more robust options for voice-first chronic disease management and patient engagement.


Underreporting makes it difficult to know just how many people suffer from intimate partner violence (IPV), says Bharti Khurana, MD, an emergency radiologist at BWH.  But the symptoms are often hiding in plain sight for radiologists.

Using artificial intelligence to flag worrisome injury patterns or mismatches between patient-reported histories and the types of fractures present on x-rays can alert providers to when an exploratory conversation is called for.

“As a radiologist, I’m very excited because this will enable me to provide even more value to the patient instead of simply evaluating their injuries.  It’s a powerful tool for clinicians and social workers that will allow them to approach patients with confidence and with less worry about offending the patient or the spouse,” said Khurana.


Every second counts when a patient experiences a stroke.  In far-flung regions of the United States and in the developing world, access to skilled stroke care can take hours, drastically increasing the likelihood of significant long-term disability or death.

Artificial intelligence has the potential to close the gaps in access to high-quality imaging studies that can identify the type of stroke and the location of the clot or bleed.  Research teams are currently working on AI-driven tools that can automate the detection of stroke and support decision-making around the appropriate treatment for the individual’s needs.  

In rural or low-resource care settings, these algorithms can compensate for the lack of a specialist on-site and ensure that every stroke patient has the best possible chance of treatment and recovery.

AI revolutionizing stroke care

Source: Getty Images


The costs of healthcare administration are off the charts.  Recent data from the Center for American progress states that providers spend about $282 billion per year on insurance and medical billing, and the burdens are only going to keep getting bigger.

Medical coding and billing is a perfect use case for natural language processing and machine learning.  NLP is well-suited to translating free-text notes into standardized codes, which can move the task off the plates of physicians and reduce the time and effort spent on complying with convoluted regulations.

“The ultimate goal is to help reduce the complexity of the coding and billing process through automation, thereby reducing the number of mistakes – and, in turn, minimizing the need for such intense regulatory oversight,” Partners says.

NLP is already in relatively wide use for this task, and healthcare organizations are expected to continue adopting this strategy as a way to control costs and speed up their billing cycles.


AI will combine with another game-changing technology, known as FHIR, to unlock siloes of health data and support broader access to health information.

Patients, providers, and researchers will all benefit from a more fluid health information exchange environment, especially since artificial intelligence models are extremely data-hungry.

Stakeholders will need to pay close attention to maintaining the privacy and security of data as it moves across disparate systems, but the benefits have the potential to outweigh the risks.

“It completely depends on how everyone in the medical community advocates for, builds, and demands open interfaces and open business models,” said Samuel Aronson, Executive Director of IT at Partners Personalized Medicine.

“If we all row in the same direction, there’s a real possibility that we will see fundamental improvements to the healthcare system in 3 to 5 years.”


Image-heavy disciplines have started to see early benefits from artificial intelligence since computers are particularly adept at analyzing patterns in pixels.  Ophthalmology is one area that could see major changes as AI algorithms become more accurate and more robust.

From glaucoma to diabetic retinopathy, millions of patients experience diseases that can lead to irreversible vision loss every year.  Employing AI for clinical decision support can extend access to eye health services in low-resource areas while giving human providers more accurate tools for catching diseases sooner.


The brain is still the body’s most mysterious organ, but scientists and clinicians are making swift progress unlocking the secrets of cognitive function and neurological disease.  Artificial intelligence is accelerating discovery by helping providers interpret the incredibly complex data that the brain produces.

From predicting seizures by reading EEG tests to identifying the beginnings of dementia earlier than any human, artificial intelligence is allowing providers to access more detailed, continuous measurements – and helping patients improve their quality of life.

Seizures can happen in patients with other serious illnesses, such as kidney or liver failure, explained, Bandon Westover, MD, PhD, executive director of the Clinical Data Animation Center at MGH, but many providers simply don’t know about it.

“Right now, we mostly ignore the brain unless there’s a special need for suspicion,” he said.  “In a year’s time, we’ll be catching a lot more seizures and we’ll be doing it with algorithms that can monitor patients continuously and identify more ambiguous patterns of dysfunction that can damage the brain in a similar manner to seizures.”


Malaria is a daily threat for approximately half the world’s population.  Nearly half a million people died from the mosquito-borne disease in 2017, according to the World Health Organization, and the majority of the victims are children under the age of five.

Deep learning tools can automate the process of quantifying malaria parasites in blood samples, a challenging task for providers working without pathologist partners.  One such tool achieved 90 percent accuracy and specificity, putting it on par with pathology experts.

This type of software can be run on a smartphone hooked up to a camera on a microscope, dramatically expanding access to expert-level diagnosis and monitoring.

AI for diagnosing and detecting malaria

Source: Getty Images


Artificial intelligence has made especially swift progress in diagnostic specialties, including pathology. AI will continue to speed down the road to maturity in this area, predicts Annette Kim, MD, PhD, associate professor of pathology at BWH and Harvard Medical School.

“Pathology is at the center of diagnosis, and diagnosis underpins a huge percentage of all patient care.  We’re integrating a huge amount of data that funnels through us to come to a diagnosis.  As the number of data points increases, it negatively impacts the time we have to synthesize the information,” she said.

AI can help automate routine, high-volume tasks, prioritize and triage cases to ensure patients are getting speedy access to the right care, and make sure that pathologists don’t miss key information hidden in the enormous volumes of clinical and test data they must comb through every day.

“This is where AI can have a huge impact on practice by allowing us to use our limited time in the most meaningful manner,” Kim stressed.


Suicide is the tenth leading cause of death in the United States, claiming 45,000 lives in 2016.  Suicide rates are on the rise due to a number of complex socioeconomic and mental health factors, and identifying patients at the highest risk of self-harm is a difficult and imprecise science.

Natural language processing and other AI methodologies may help providers identify high-risk patients earlier and more reliably.  AI can comb through social media posts, electronic health record notes, and other free-text documents to flag words or concepts associated with the risk of harm.

Researchers also hope to develop AI-driven apps to provide support and therapy to individuals likely to harm themselves, especially teenagers who commit suicide at higher rates than other age groups.

Connecting patients with mental health resources before they reach a time of crisis could save thousands of lives every year.


Radiology is already one of AI’s early beneficiaries, but providers are just at the beginning of what they will be able to accomplish in the next few years as machine learning explodes into the imaging realm.

AI is predicted to bring earlier detection, more accurate assessment of complex images, and less expensive testing for patients across a huge number of clinical areas.

But as leaders in the AI revolution, radiologists also have a significant responsibility to develop and deploy best practices in terms of trustworthiness, workflow, and data protection.

“We certainly feel the onus on the radiology community to make sure we do deliver and translate this into improved care,” said Alexandra Golby, MD, a neurosurgeon and radiologist at BWH and Harvard Medical School.

“Can radiology live up to the expectations?  There are certainly some challenges, including trust and understanding of what the algorithms are delivering.  But we desperately need it, and we want to equalize care across the world.”

Radiologists have been among the first to overcome their trepidation about the role of AI in a changing clinical world, and are eagerly embracing the possibilities of this transformative approach to augmenting human skills.”

“All of the imaging societies have opened their doors to the AI adventure,” Golby said.  “The community very anxious to learn, codevelop, and work with all of the industry partners to turn this technology into truly valuable tools. We’re very optimistic and very excited, and we look forward to learning more about how AI can improve care.”




Read Full Post »

Older Posts »