Posts Tagged ‘epilepsy’

Developing Deep Learning Models (DL) for the Instant Prediction of Patients with Epilepsy

Reporter: Srinivas Sriram, Research Assistant I
Research Team: Srinivas Sriram, Abhisar Anand

2021 LPBI Summer Intern in Data Science and Website Construction
This article reports on a research study conducted from January 2021 to May 2021.
This Research was completed before the 2021 LPBI Summer Internship that began on 6/15/2021.

The main criterion of this study was to utilize the dataset (shown above) to develop a DL network that could accurately predict new seizures based on incoming data. To begin the study, our research group did some exploratory data analysis on the dataset and we recognized the key defining pattern of the data that allowed for the development of the DL model. This pattern of the data can be represented in the graph above, where the lines representing seizure data had major spikes in extreme hertz values, while the lines representing normal patient data remained stable without any spikes. We utilized this pattern as a baseline for our model. 

Conclusions and Future Improvements:

Through our system, we were able to create a prototype solution that would predict when seizures happened in a potential patient using an accurate LSTM network and a reliable hardware system. This research can be implemented in hospitals with patients suffering from epilepsy in order to help them as soon as they experience a seizure to prevent damage. However, future improvements need to be made to this solution to allow it to be even more viable in the Healthcare Industry, which is listed below.

  • Needs to be implemented on a more reliable EEG headset (covers all neurons of the brain, less prone to electric disruptions shown in the prototype). 
  • Needs to be tested on live patients to deem whether the solution is viable and provides a potential solution to the problem. 
  • The network can always be fine-tuned to maximize performance. 
  • A better alert system can be implemented to provide as much help as possible. 

These improvements, when implemented, can help provide a real solution to one of the most common diseases faced in the world. 

Background Information:

Epilepsy is described as a brain disorder diagnostic category for multiple occurrences of seizures that happen within recurrent and/or a brief timespan. According to the World Health Organization, seizure disorders, including epilepsy, are among the most common neurological diseases. Those who suffer seizures have a 3 times higher risk of premature death. Epilepsy is often treatable, especially when physicians can provide necessary treatment quickly. When untreated, however, seizures can cause physical, psychological, and emotional, including isolation from others. Quick diagnosis and treatment prevent suffering and save lives. The importance of a quick diagnosis of epilepsy has led to our research team developing Deep Learning (DL) algorithms for the sole purpose of detecting epileptic seizures as soon as they occur. 

Throughout the years, one common means of detecting Epilepsy has emerged in the form of an electroencephalogram (EEG). EEGs can detect and compile “normal” and “abnormal “brain wave activity” and “indicate brain activity or inactivity that correlates with physical, emotional, and intellectual activities”. EEG waves are classified mainly by brain wave frequencies (EEG, 2020). The most commonly studied are delta, theta, alpha, sigma, and beta waves. Alpha waves, 8 to 12 hertz, are the key wave that occurs in normal awake people. They are the defining factor for the everyday function of the adult brain. Beta waves, 13 to 30 hertz, are the most common type of wave in both children and adults. They are found in the frontal and central areas of the brain and occur at a certain frequency which, if slow, is likely to cause dysfunction. Theta waves, 4 to 7 hertz, are also found in the front of the brain, but they slowly move backward as drowsiness increases and the brain enters the early stages of sleep. Theta waves are known as active during focal seizures. Delta waves, 0.5 to 4 hertz, are found in the frontal areas of the brain during deep sleep. Sigma waves, 12-16 hertz, are very slow frequency waves that occur during sleep. EEG detection of electrical brain wave frequencies can be used to detect and diagnose seizures based on their deviation from usual brain wave patterns.

In this particular research project, our research group hoped to develop a DL algorithm that when implemented on a live, portable EEG brain wave capturing device, could accurately predict when a particular patient was suffering from Epilepsy as soon as it occurred. This would be accomplished by creating a network that could detect when the brain frequencies deviated from the normal frequency ranges. 

The Study:

Line Graph representing EEG Brain Waves from a Seizure versus EEG Brain Waves from a normal individual. 

Source Dataset: https://archive.ics.uci.edu/ml/datasets/Epileptic+Seizure+Recognition

To expand more on the dataset, it is an EEG data set compiled by Qiuyi Wu and Ernest Fokoue (2021) from the work of medical researchers R.Andrzejak, M.D. et al. (2001) which had been made public domain through the UCI Machine Learning Repository We also confirmed fair use permission with UCI. The dataset had been gathered by Andrzejak during examinations of 500 patients with a chronic seizure disorder. R.G.Andrzejak, et al. (2001) recorded each entry in the EEG dataset used for this project within 23.6 seconds in a time-series data structure. Each row in the dataset represented a patient recorded. The continuous variables in the dataset were single EEG data points at that specific point in time during the measuring period. At the end of the dataset, was a y-variable that indicated whether or not the patient had a seizure during the period the data was recorded. The continuous variables, or the EEG data, for each patient, varied widely based on whether the patient was experiencing a seizure at that time. The Wu & Fokoue Dataset (2021) consists of one file of 11,500 rows, each with 178 sequential data points concatenated from the original dataset of 5 data folders, each including 100 files of EEG recordings of 23.6 seconds and containing 4097 data points. Each folder contained a single, original subset. Subset A contained EEG data gathering during epileptic seizure…. Subset B contained EEG data from brain tumor sites. Subset 3, from a healthy site where tumors had been located. Subsets 4 and 5 from non-seizure patients at rest with eyes open and closed, respectively. 

Based on the described data, our team recognized that a Recurrent Neural Network (RNN) was needed to input the sequential data and return an output of whether the sequential data was a seizure or not. However, we realized that RNN models are known to get substantially large over time, reducing computation speeds. To help provide a solution to this issue, our group decided to implement a long-short-term memory (LSTM) model. After deciding our model’s architecture, we proceeded to train our model in two different DL frameworks inside Python, TensorFlow, and PyTorch. Through various rounds of retesting and redesigning, we were able to train and develop two accurate models in each of the models that not only performed well while learning the data while training, but also could accurately predict new data in the testing set (98 percent accuracy on the unseen data). These LSTM networks could classify normal EEG data when the brain waves are normal, and then immediately predict the seizure data based on if a dramatic spike occurred in the data. 

After training our model, we had to implement our model in a real-life prototype scenario in which we utilized a Single Board Computer (SBC) in the Raspberry Pi 4 and a live capturing EEG headset in the Muse 2 Headband. The two hardware components would sync up through Bluetooth and the headband would return EEG data to the Raspberry Pi, which would process the data. Through the Muselsl API in Python, we were able to retrieve this EEG data in a format similar to the manner implemented during training. This new input data would be fed into our LSTM network (TensorFlow was chosen for the prototype due to its better performance than the PyTorch network), which would then output the result of the live captured EEG data in small intervals. This constant cycle would be able to accurately predict a seizure as soon as it occurs through batches of EEG data being fed into the LSTM network. Part of the reason why our research group chose the Muse Headband, in particular, was not only due to its compatibility with Python but also due to the fact that it was able to represent seizure data. Because none of our members had epilepsy, we had to find a reliable way of testing our model to make sure it worked on the new data. Through electrical disruptions in the wearable Muse Headband, we were able to simulate these seizures that worked with our network’s predictions. In our program, we implemented an alert system that would email the patient’s doctor as soon as a seizure was detected.

Individual wearing the Muse 2 Headband

Image Source: https://www.techguide.com.au/reviews/gadgets-reviews/muse-2-review-device-help-achieve-calm-meditation/

Sources Cited:

Wu, Q. & Fokoue, E. (2021).  Epileptic seizure recognition data set: Data folder & Data set description. UCI Machine Learning Repository: Epileptic Seizure Recognition. Jan. 30. Center for Machine Learning and Intelligent Systems, University of California Irvine.

Nayak, C. S. (2020). EEG normal waveforms.” StatPearls [Internet]. U.S. National Library of Medicine, 31 Jul. 2020, www.ncbi.nlm.nih.gov/books/NBK539805/#.

Epilepsy. (2019). World Health Organization Fact Sheet. Jun. https://www.who.int/ news-room/fact-sheet s/detail/epilepsy

Other Related Articles published in this Open Access Online Scientific Journal include the following:

Developing Deep Learning Models (DL) for Classifying Emotions through Brainwaves

Reporter: Abhisar Anand, Research Assistant I


Machine Learning (ML) in cancer prognosis prediction helps the researcher to identify multiple known as well as candidate cancer diver genes

Curator and Reporter: Dr. Premalata Pati, Ph.D., Postdoc


Deep Learning-Assisted Diagnosis of Cerebral Aneurysms

Reporter: Dror Nir, PhD


Developing Machine Learning Models for Prediction of Onset of Type-2 Diabetes

Reporter: Amandeep Kaur, B.Sc., M.Sc.


Deep Learning extracts Histopathological Patterns and accurately discriminates 28 Cancer and 14 Normal Tissue Types: Pan-cancer Computational Histopathology Analysis

Reporter: Aviva Lev-Ari, PhD, RN


A new treatment for depression and epilepsy – Approval of external Trigeminal Nerve Stimulation (eTNS) in Europe

Reporter: Howard Donohue, PhD (EAW)


Mutations in a Sodium-gated Potassium Channel Subunit Gene related to a subset of severe Nocturnal Frontal Lobe Epilepsy

Reporter: Aviva Lev-Ari, PhD, RN


Read Full Post »

Reporter: Aviva Lev-Ari, PhD, RN

Australian-led Team Reports on New Nocturnal Epilepsy Gene

October 22, 2012

NEW YORK (GenomeWeb News) – An international team led by investigators in Australia has linked mutations in a sodium-gated potassium channel subunit gene to a subset of severe nocturnal frontal lobe epilepsy cases.

As they reported online yesterday in Nature Genetics, the researchers began by testing a family with autosomal dominant nocturnal frontal lobe epilepsy, or ADNFLE. Affected members of the family often had not only typical ADNFLE symptoms, but also intellectual and/or psychiatric features that don’t usually characterize the disorder.

After narrowing in on a chromosome 9 region via linkage analyses in the family, the team identified ADNFLE-associated missense mutations in the sodium-gated potassium channel subunit gene KCNT1 by whole-exome sequencing in two affected family members. Follow-up testing on more than 100 other unrelated individuals with nocturnal frontal lobe epilepsy indicated that both inherited and de novo mutations in the gene can cause severe forms of the conditions that tend to include other co-morbidities.

“KCNT1 mutations were identified in two additional families and a sporadic case with severe ADNFLE and psychiatric features,” University of South Australia researcher Leanne Dibbens and the University of Melbourne’s Ingrid Scheffer, the study’s co-corresponding authors, and their colleagues wrote.

“These findings implicate the sodium-gated potassium channel complex in ADNFLE, and, more broadly, in the pathogenesis of focal epilepsies,” they added.

As the name suggests, ADNFLE is inherited in an autosomal dominant manner in affected families. Symptoms of the condition — including seizures that occur while individuals are asleep — generally appear in childhood, the researchers explained. And previous studies have implicated mutations to nicotinic acetylcholine receptor subunit genes in a subset of ADNFLE cases.

For the current study, the team focused on a multi-generational family with an especially severe form of ADNFLE that was accompanied by other symptoms such as intellectual disability and psychiatric disorders.

Genome-wide linkage analyses within the family led to a suspicious 2.36 million base stretch of sequence on chromosome 9, which housed almost 100 genes. Among them: two ion channel-coding genes, KCNT1 and GRIN1.

For two of the affected family members, the team turned to whole-exome sequencing to try to track down the most likely cause of ADNFLE. Indeed, missense mutations in KCNT1 that were predicted to be pathogenic turned up in one of the two exome sequences.

The mutation was not initially identified in the other family member’s exome sequence data, owing to low coverage, researchers explained. But it was subsequently shown to be present in both individuals by Sanger sequencing.

Consistent with the notion that this KCNT1 mutation could be related to ADNFLE pathogenesis, the investigators did not find it when they tested 111 unaffected, ancestry-matched individuals. Nor did it turn up in the dbSNP database, they reported, or in data generated for the 1000 Genomes Project or through the National Heart, Lung, and Blood Institute’s Exome Sequencing Project.

On the other hand, the team did find mutations in KCNT1 when it assessed another 108 unrelated individuals who either had ADNFLE or sporadically occurring nocturnal frontal lobe epilepsy.

That analysis helped the investigators track down two more ADNFLE-affected families with KCNT1 mutations that co-segregated with the disease, along with one case of sporadic nocturnal frontal lobe epilepsy including psychiatric features that seemed to stem from de novo mutations to KCNT1.

“[T]he phenotype associated with KNCT1 mutations is both more severe and more penetrant than that typically found with mutations affecting [nicotinic acetylcholine receptors],” the study’s authors noted.

In addition to showing more pronounced ADNFLE symptoms, they explained, the disease appears to manifest itself at a younger age in the cases linked to KCNT1 mutations.

Moreover, several cases that appear to be caused by alterations to KCNT1 also included intellectual disability, psychiatric, and/or behavioral features. The severity of such symptoms varied from one individual to the next — a pattern that the researchers speculated might be due to differences in the nature and extent of the KCNT1 mutation involved.

In addition to providing clues to help classify ADNFLE cases and offer genetic counseling for families affected by it, those involved in the study say the results should also prove useful for understanding — and potentially targeting — the processes that underlie this type of epilepsy.

“[T]his finding should provide new insights into the biological mechanisms underlying the pathogenesis of ADNFLE,” they concluded, “which may lead to targeted therapies addressing the serious co-morbidities as well as the debilitating seizure disorder.”

Read Full Post »

The eTNS System. (PRNewsFoto/NeuroSigma)

Reporter: Howard Donohue, PhD (EAW)

Following the arrival in the 1990s of a drug for treating depression called fluoxetine (better known by its brand name, Prozac) – a “selective serotonin reuptake inhibitor” (SSRI) – it’s probably fair to say that not many drugs have become as deeply engrained in the public’s general awareness as those of this type. Perhaps one reason for this could be the sheer number of people affected by depression and to whom SSRIs are relevant as a possible treatment (one study has estimated that depression affected upwards of 30 million Europeans in the year 2010 [1]). Perhaps another reason could be the various controversies that have surrounded SSRIs over the years, from stories of increased suicide risk in children [2] to evidence of biases and the “selective” publishing of clinical data favoring the effectiveness of these drugs [3]. Of course, despite the controversies, SSRIs (along with other classes of antidepressant drug) continue to be a mainstay, but let’s not forget, amid their popularity, that there are other ways to treat depressive illnesses. And in maximizing the benefits of treatment for the individual, it’s important to realize that any one of these approaches might work well for one person, but not for another. Among the non-pharmacologic ways to treat depression are psychological approaches, for example cognitive behavioral therapy, or alternatively, “brain stimulation” approaches such as electroconvulsive therapy (ECT). ECT is a method to induce a mild seizure in the patient by means of electrical activity applied to the brain via electrodes connected to the temples.

On the subject of ECT; you could be forgiven for thinking that it’s not very nice, especially if you’ve seen the plights of characters like Randle Patrick “Mac” McMurphy, portrayed by Jack Nicholson in One Flew Over the Cuckoo’s Nest or Russell Crowe’s portrayal of Dr. John Nash (based on the real-life Nobel Laureate in Economics by the same name) in A Beautiful Mind. Nonetheless, despite the treatment in Hollywood of ECT as a sinister, repressive, and even brutal procedure, the reality is obviously different and it continues to have a place in medical practice for the treatment of severely depressed patients to this day. This isn’t to say that controversies don’t exist within the medical community concerning certain side effects (such as memory loss), but in balancing this, we should remember that many – if not most – medical procedures have their drawbacks (hopefully, the benefits will far outweigh the drawbacks). Putting aside any thoughts on whether ECT is good or bad, it is recognition and consideration of the drawbacks that helps drive the evolution of medical technologies.

So, in illustrating the evolution that is happening in the field of brain stimulation for treating neurological disorders (in this case, depression and also epilepsy), the recent approval in Europe of an “external Trigeminal Nerve Stimulation” (eTNS) technique provides an excellent example. The technique, called the MonarchTM and exclusively licensed to Neurosigma Inc. (a Los Angeles-based medical device company) “for the adjunctive treatment of epilepsy and major depressive disorder, for adults and children 9 years and older”, is a non-invasive form of neuromodulation therapy [4]. It was invented at the University of California, Los Angleles (UCLA) and has been in development for over 10 years [4]. It works by using a low-energy stimulus to stimulate branches of the trigeminal nerve, a nerve that can affect the activity of several key brain regions believed to be involved in depression and epilepsy. In contrast to ECT, the stimulus is restricted to the soft tissues of the forehead without direct penetration to the brain, which thereby facilitates a non-invasive form of neuromodulation [4]. Following European approval, Neurosigma affirmed in a press release that eTNS is “supported by years of safety and compelling efficacy data generated in clinical trials conducted at UCLA and the University of Southern California (USC)” [4]. In realizing the future potential of eTNS, Neurosigma’s business strategy is now geared toward steps for its adoption at major epilepsy and depression centers in the EU, as well as endeavors to make it available to patients in the US and other countries [4].

To answer the question of whether eTNS will rise to prominence as an effective treatment in the fight against depression and epilepsy, only time will tell. But if it does, as well as being a valuable addition to the armamentarium against these debilitating diseases, maybe its non-invasive nature will mean that the film directors have a harder time in “demonizing” it for dramatic effect. Well anyway, let’s hope so.


  1. Wittchen et al. Eur Neuropsychopharmacol 2011: 21:655-79.
  2. http://news.bbc.co.uk/2/hi/health/3656110.stm
  3. Turner et al. N Engl J Med 2008; 358:252-60.
  4. http://www.prnewswire.com/news-releases/neurosigma-receives-ce-certification-168578146.html

Read Full Post »


Reporter: Howard Donohue, PhD (EAW)

The hypothalamic-pituitary-adrenal (HPA) axis – which can be thought of as a series of closely linked endocrine structures in the brain – has a key role in triggering the body’s stress response through the secretion of cortisol. In explaining how the HPA axis is itself regulated, for example how its activity is increased in response to a perceived environmental threat, we can infer that the diverse brain areas with which it shares neural interconnections have a crucial role (for a review, see [1]). An equally important question relates to how the activity of the HPA axis is returned to normal when the stress response is no longer needed. To answer this, it is well known that the same “neurosteroid” hormones released by the HPA axis that trigger stress-related biological adaptations also serve to dampen its activity through a “negative feedback” mechanism. In re-defining the biological model of how neurosteroids control the HPA axis, a study led by Jamie Maguire, PhD at Tufts University (Boston, MA) provides some fascinating insights [2]. Moreover, this study has some extremely interesting and counter-intuitive implications for understanding the functions of the “inhibitory” brain chemical gamma-aminobutyric acid (GABA), which is best known for opposing the effects of “excitatory” brain chemicals in order to balance the flow of electrical activity in the brain.

To study how the HPA axis is regulated by neurosteroids, Maguire’s team performed investigations in mice using the neurosteroid tetrahydrodeoxycorticosterone (THDOC). The investigators found that THDOC, when applied to a discrete population of cells in the thalamus called the paraventricular nucleus (PVN), resulted in a decrease in blood levels of corticosterone (the mouse equivalent of the human stress hormone, cortisol). This finding highlights the importance of the PVN as a key anatomical locus in the brain where neurosteroids act, and is consistent with the traditional view of neurosteroids as “negative regulators” of the HPA axis. However, in mice that underwent a stressful “restraint” procedure, it was found that a prior treatment with THDOC (thirty minutes before the stressful experience) resulted in augmentation of corticosterone levels (i.e. relative to mice that underwent the stressful experience but did not receive prior THDOC treatment). In parallel, it was shown that while application of THDOC normally decreased the electrical activity of PVN cells, it actually led to increases in mice that had undergone restraint. Taken together, these findings provide evidence that neurosteroids can have opposite effects on the HPA axis depending on the “stressed” state of the organism.

Thinking about how a neurosteroid hormone can exert opposite effects on PVN cells in the thalamus may be confusing, but what may be more confusing is that these different actions depend on the same “inhibitory” brain chemical, GABA (a neurotransmitter), as well as the same molecular “machinery” (or receptors) with which GABA interacts. This was demonstrated by using mice in which a particular sub-component (or subunit) of the GABA receptor, the gamma subunit, had been genetically deleted; neurosteroids had absolutely no effect on the activity of the HPA axis (neither positive nor negative) in these gamma subunit-deficient mice.

How is it possible to explain the seemingly paradoxical finding that neurosteroids can exert opposite effects on the HPA axis through the same neurotransmitter system? In addressing this question, it is important to remember that although neurotransmitters may be thought of as excitatory or inhibitory, their ability to trigger these effects depends solely on the molecular and cellular apparatus with which they interact. Normally, the inhibitory actions of GABA upon the electrical activity of nerve cells depend on the maintenance of an “electrochemical” gradient by a “transporter” molecule called KCC2 (which transports chloride ions out of cells). Maguire’s team showed that “dephosphorylation” (i.e. the removal of a small chemical moiety – the phosphate group – which is covalently bound at a specific site on the molecule) of KCC2 resulted in lower detectable levels of this transporter in the PVN. Similarly to innumerable other examples in biology where dephosphorylation (or the reverse, phosphorylation) serves as an exquisite regulatory mechanism for controlling the activity of molecular networks, removal of the phosphate group from KCC2 acts as a molecular “switch” that causes the breakdown of the electrochemical gradient. The outcome is that GABA has an excitatory influence on neural activity instead of the inhibitory influence with which it is usually associated.

In common with many important contributions to scientific understanding, these findings should serve as a reminder that it is often necessary to challenge and question what is already “accepted” in our theoretical models, in the light of unexpected and sometimes counter-intuitive experimental results. Whatever the line of scientific inquiry may be, the reward for doing so will be a deeper and more comprehensive understanding of the natural phenomena being studied. The findings of Maguire and colleagues, published in the Journal of Neuroscience, have possible therapeutic implications for disorders associated with disrupted function of the HPA axis, including epilepsy and depression.


1. http://www.nature.com/nrn/journal/v10/n6/full/nrn2647.html

2. http://www.jneurosci.org/content/31/50/18198.long


Read Full Post »

%d bloggers like this: