Advertisements
Feeds:
Posts
Comments

Posts Tagged ‘mammography breast cancer screening’


The debate on the function AI could or should realize in modern radiology is buoyant presenting wide spectrum of positive expectations and also fears.

The article: A Deep Learning Model to Triage Screening Mammograms: A Simulation Study that was published this month shows the best, and very much feasible, utility for AI in radiology at the present time. It would be of great benefit for radiologists and patients if such applications will be incorporated (with all safety precautions taken) into routine practice as soon as possible.

In a simulation study, a deep learning model to triage mammograms as cancer free improves workflow efficiency and significantly improves specificity while maintaining a noninferior sensitivity.

Background

Recent deep learning (DL) approaches have shown promise in improving sensitivity but have not addressed limitations in radiologist specificity or efficiency.

Purpose

To develop a DL model to triage a portion of mammograms as cancer free, improving performance and workflow efficiency.

Materials and Methods

In this retrospective study, 223 109 consecutive screening mammograms performed in 66 661 women from January 2009 to December 2016 were collected with cancer outcomes obtained through linkage to a regional tumor registry. This cohort was split by patient into 212 272, 25 999, and 26 540 mammograms from 56 831, 7021, and 7176 patients for training, validation, and testing, respectively. A DL model was developed to triage mammograms as cancer free and evaluated on the test set. A DL-triage workflow was simulated in which radiologists skipped mammograms triaged as cancer free (interpreting them as negative for cancer) and read mammograms not triaged as cancer free by using the original interpreting radiologists’ assessments. Sensitivities, specificities, and percentage of mammograms read were calculated, with and without the DL-triage–simulated workflow. Statistics were computed across 5000 bootstrap samples to assess confidence intervals (CIs). Specificities were compared by using a two-tailed t test (P < .05) and sensitivities were compared by using a one-sided t test with a noninferiority margin of 5% (P < .05).

Results

The test set included 7176 women (mean age, 57.8 years ± 10.9 [standard deviation]). When reading all mammograms, radiologists obtained a sensitivity and specificity of 90.6% (173 of 191; 95% CI: 86.6%, 94.7%) and 93.5% (24 625 of 26 349; 95% CI: 93.3%, 93.9%). In the DL-simulated workflow, the radiologists obtained a sensitivity and specificity of 90.1% (172 of 191; 95% CI: 86.0%, 94.3%) and 94.2% (24 814 of 26 349; 95% CI: 94.0%, 94.6%) while reading 80.7% (21 420 of 26 540) of the mammograms. The simulated workflow improved specificity (P = .002) and obtained a noninferior sensitivity with a margin of 5% (P < .001).

Conclusion

This deep learning model has the potential to reduce radiologist workload and significantly improve specificity without harming sensitivity.

Advertisements

Read Full Post »


Costs for breast screening are being driven higher by increased use of new imaging technologies such as digital mammography and MRI, workflows incorporating 2nd and 3rd remote-readings as quality control measure, use of computer-aided detection (CAD) applications and growth in aged population.

According to recent publication in JAMA, 40% of the annual spending is for screening women ages 75 and older. Under existing guidelines routine screening is not recommended for this age group since “There is insufficient evidence to assess the benefits and harms of screening mammography”

The study population comprised women of 66 to 100 years old. “Forty-two percent of the women in the study were younger than age 75, and almost 60% of this group had one or more screening mammograms. Women ages 75 to 85 represented 40% of the total; 42% of this group had mammograms. Women 85 years and older represented 18% of the total; only 15% of this group had mammograms. Women with multiple chronic health conditions were much less likely to have a mammogram (26.6%) than healthy women (47.4%).”

“Abstract

The Cost of Breast Cancer Screening in the Medicare Population.

Cary P. Gross, MD; Jessica B. Long, MPH; Joseph S. Ross, MD, MHS; Maysa M. Abu-Khalaf, MD; Rong Wang, PhD; Brigid K. Killelea, MD, MPH; Heather T. Gold, PhD; Anees B. Chagpar, MD, MA, MPH, MSc; Xiaomei Ma, PhD

JAMA Intern Med. 2013;():1-7. doi:10.1001/jamainternmed.2013.1397. Published online January 7, 2013

Background  Little is known about the cost to Medicare of breast cancer screening or whether regional-level screening expenditures are associated with cancer stage at diagnosis or treatment costs, particularly because newer breast cancer screening technologies, like digital mammography and computer-aided detection (CAD), have diffused into the care of older women.

Methods Using the linked Surveillance, Epidemiology, and End Results–Medicare database, we identified 137 274 women ages 66 to 100 years who had not had breast cancer and assessed the cost to fee-for-service Medicare of breast cancer screening and workup during 2006 to 2007. For women who developed cancer, we calculated initial treatment cost. We then assessed screening-related cost at the Hospital Referral Region (HRR) level and evaluated the association between regional expenditures and workup test utilization, cancer incidence, and treatment costs.

Results In the United States, the annual costs to fee-for-service Medicare for breast cancer screening-related procedures (comprising screening plus workup) and treatment expenditures were $1.08 billion and $1.36 billion, respectively. For women 75 years or older, annual screening-related expenditures exceeded $410 million. Age-standardized screening-related cost per beneficiary varied more than 2-fold across regions (from $42 to $107 per beneficiary); digital screening mammography and CAD accounted for 65% of the difference in screening-related cost between HRRs in the highest and lowest quartiles of cost. Women residing in HRRs with high screening costs were more likely to be diagnosed as having early-stage cancer (incidence rate ratio, 1.78 [95% CI, 1.40-2.26]). There was no significant difference in the cost of initial cancer treatment per beneficiary between the highest and lowest screening cost HRRs ($151 vs $115; P = .20).

Conclusions The cost to Medicare of breast cancer screening exceeds $1 billion annually in the fee-for-service program. Regional variation is substantial and driven by the use of newer and more expensive technologies; it is unclear whether higher screening expenditures are achieving better breast cancer outcomes.”

The study is mainly addressing the difference in costs between different regions of referrals. It would be interesting to explore the situation in the age group of 40 to 66 years old.

Written by:  Dr. Dror Nir, PhD.

Read Full Post »


Many feedbacks to my last post reflected radiologists’ perception of ultrasound as a low-tech, unreliable imaging device.

Ultrasounds most manifested limitation by radiologists is that its performance is too-much user-dependent. This opinion finds support in numerous clinical studies concluding that ultrasound-based assessment of a cancer patient varies with the operator.

How come that an imaging technology that is not only  low-cost, simple to operate and risk-free to the patient, but has also gained a leading position in certain domain, like obstetrics,  is perceived as the underdog when it comes  to cancer assessment? Could it be because of its positioning as a “multi-purpose” system, which requires only very basic training?

If indeed this is the case, it doesn’t require “rocket-science” to turn it around. It only needs designing dedicated ultrasound machines who offer a comprehensive solution to one specific clinical need. Using such machines will require highly skilled operators who will enjoy a superior workflow, reporting tools and proven clinical guidelines.

The unsatisfactory reality of mammography-based breast cancer screening, as evident by epidemiology data and expert-panels’ reports, opens the opportunity to transform ultrasound into a winner in the niche-market of breast cancer screening and diagnosis. It’s a significant market that justifies the investment in ultrasound systems dedicated to detection and characterisation of breast cancer lesions.

No doubt, that the ability to provide accurate and standardized interpretation of such ultrasound systems’ scans is a pre-requisite. Ultrasound-based tissue characterisation is a must for any application aiming at standardized image interpretation. A sample out-of present ultrasound-based technologies aiming at providing some level of tissue-characterisation are listed below. Recent clinical studies show promising results using these technologies. It is worth watching carefully to see if any of those could be part of a future ultrasound-based solution to breast cancer screening.

Solid Breast Lesions: Clinical Experience with US-guided Diffuse Optical Tomography Combined with Conventional US

Results: Of the 136 biopsied lesions, 54 were carcinomas and 82 were benign. The average total hemoglobin concentration in the malignant group was 223.3 μmol/L ± 55.8 (standard deviation), and the average hemoglobin concentration in the benign group was 122.5 μmol/L ± 80.6 (P = .005). When the maximum hemoglobin concentration of 137.8 μmol/L was used as the threshold value, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of DOT with US localization were 96.3%, 65.9%, 65.0%, 96.4%, and 76.5%, respectively. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of conventional US were 96.3%, 92.6%, 89.7%, 97.4%, and 93.4%, respectively. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of conventional US combined with DOT were 100%, 93.9%, 91.5%, 100%, and 96.3%, respectively.

Conclusion: US-guided DOT combined with conventional US improves accuracy compared with DOT alone.

Breast Lesions: Quantitative Elastography with Supersonic Shear Imaging—Preliminary Results

 

 

Results: All breast lesions were detected at Supersonic Shear Imaging. Malignant lesions exhibited a mean elasticity value of 146.6 kPa ± 40.05 (standard deviation), whereas benign ones had an elasticity value of 45.3 kPa ± 41.1 (P < .001). Complicated cysts were differentiated from solid lesions because they had elasticity values of 0 kPa (no signal was retrieved from liquid areas).

Conclusion: Supersonic Shear Imaging provides quantitative elasticity measurements, thus adding complementary information that potentially could help in breast lesion characterization with B-mode US.

 Distinguishing Benign from Malignant Masses at Breast US: Combined US Elastography and Color Doppler US—Influence on Radiologist Accuracy

Results: The Az of B-mode US, US elastography, and Doppler US (average, 0.844; range, 0.797–0.876) was greater than that of B-mode US alone (average, 0.771; range, 0.738–0.798) for all readers (P = .001 for readers 1, 2, and 3; P < .001 for reader 4; P = .002 for reader 5). When both elastography and Doppler scores were negative, leading to strict downgrading, the specificity increased for all readers from an average of 25.3% (75.4 of 298; range, 6.4%–40.9%) to 34.0% (101.2 of 298; range, 26.5%–48.7%) (P < .001 for readers 1, 2, 4, and 5; P = .016 for reader 3) without a significant change in sensitivity.

Conclusion: Combined use of US elastography and color Doppler US increases both the accuracy in distinguishing benign from malignant masses and the specificity in decision-making for biopsy recommendation at B-mode US.

Evaluation of breast lesions by contrast enhanced ultrasound: Qualitative and quantitative analysis

A 57-year-old woman with a no-palpable lesion in the outer upper quadrant of left breast. (a) Gray scale image show an indistinct, hypo-echoic lesion. (b) Contrast enhanced image obtained 35 s after contrast agent injection showing a homogeneously and hyper-enhanced lesion. (c) Micro flow image obtained 38 s after contrast agent injection showing the enhanced mass with several radial vessels (arrow). (d) The time-intensity curve analysis show the peak intensity is 145.69 (intensity/1000), time to peak is 15.08 s, ascending slope is 8.98, descending slope is 1.03, the area under the curve is 7783.34. Pathologic analyses show this is an invasive ductal carcinoma.

Results: Histopathologic analysis of the 91 lesions revealed 44 benign and 47 malignant. For qualitative analysis, benign and malignant lesions differ significantly in enhancement patterns (p < 0.05). Malignant lesions more often showed heterogeneous and centripetal enhancement, whereas benign lesions mainly showed homogeneous and centrifugal enhancement. The detectable rate of peripheral radial or penetrating vessels was significantly higher in malignant lesions than in benign ones (p < 0.001). For quantitative analysis, malignant lesions showed significantly higher (p = 0.031) and faster enhancement (p = 0.025) than benign ones, and its time to peak was significantly shorter (p = 0.002). The areas under the ROC curve for qualitative, quantitative and combined analysis were 0.910 (Az1), 0.768 (Az2) and 0.926(Az3) respectively. The values of Az1 and Az3 were significantly higher than that for Az2 (p = 0.024 and p = 0.008, respectively). But there was no significant difference between the values of Az1 and Az3 (p = 0.625).

Conclusions: The diagnostic performance of qualitative and combined analysis was significantly higher than that for quantitative analysis. Although quantitative analysis has the potential to differentiate benign from malignant lesions, it has not yet improved the final diagnostic accuracy.

 Breast HistoScanning: the development of a novel technique to improve tissue characterization during breast ultrasound

Results: In 17 normal testing volumes, 3% of isolated voxels were classified as abnormal. In 15 abnormal testing volumes, the subclassifiers differentiated between malignant and benign tissue. BHS in benign tissue showed <1% abnormal voxels in cyst, hamartoma, papilloma and benign fibrosis. The fibroadenomas differed showing <5% and <24% abnormal voxels. Abnormal voxels in cancers increased with the volume of cancer at pathology.

Conclusions: HistoScanning reliably discriminated normal from abnormal tissue and could distinguish between benign and malignant lesions.

Written by: Dror Nir, PhD

Read Full Post »


Automated Breast Ultrasound System (‘ABUS’) for full breast scanning: The beginning of structuring a solution for an acute need!

Writer: Dror Nir, PhD

GE Healthcare announced this week the acquisition of U-Systems, Inc. U-systems has developed the first and only Automated Breast Ultrasound System (ABUS) on the market – somo•v®, to receive FDA approval as an adjunct to mammography screening for breast cancer of; “asymptomatic women, with greater than 50 percent dense breast tissue and no prior breast interventions.”

somo•v® screen shot, showing mass in upper-outer quadrant of the left breast. Image courtesy of U-Systems.

I became aware of somo•v® already in 2004, when Prof. André Grivegnée, head of the breast screening unit at Jules Bordet – European oncology center in Brussels, Belgium, invited me to participate in a technology assessment of U-Systems’ somo•v® product. On that occasion, I also shared with U-System’s developers the idea of incorporating tissue characterisation into their product, an idea which they did not take on board. There is nothing more vivid to fully understand the meaning of this acquisition for breast cancer screening then the following quote from AuntMinnie’s report “GE taps interest in ABUS with U-Systems acquisition”:  “You know you’re onto something when the big boys come calling. GE Healthcare today announced its acquisition of automated breast ultrasound (ABUS) developer U-Systems, a move that highlights the rapid evolution of ABUS from a niche technology into a promising adjunct to screening mammography. “ First savvy: The reality of medical device startups is that it doesn’t matter how real and large is the need for your technology. Until one of the big boys will adopt it, it is prone to be considered as niche technology. I discussed the potential role of ABUS in future breast screening in my recent posts: Closing the Mammography gap; Introducing smart-imaging into radiologists’ daily practice.  As noted, in recent years, several ABUS systems were developed. An intriguing question is; why did GE choose to buy the somo•v® and not one of the other systems? Why now and not 2 or 3 years ago? The answer must have to do with the fact that in September 2012, somo•v® became the first ABUS system to receive premarket approval (PMA) for its application to use the system in a breast cancer screening environment. Until then, somo·v was indicated for use as an adjunct to mammography for B-mode ultrasonic imaging of a patient’s breast when used with an automatic scanning linear array transducer or a handheld transducer. The PMA has extended somo·v’s Indication For Use (IFU) allowing a claim that it increases breast cancer detection in a certain patients population. Second savvy: Having a PMA approval for a compelling indication for use, in a significant enough patient group, will dramatically increase “big boys” interest in your product. From the information available on the FDA site, one can get an insight into U-System’s regulatory strategy. They were smart enough to be satisfied with achieving a small step; increasing the detection rate of mammography-based screening. Therefore, the same radiologist who read the mammograms also read the ultrasound image. This increases the probability that your device’s sensitivity will not be worse than that of mammography. U-Systems did not try to go all the way to become an alternative to mammography. A claim that would significantly increase the complexity of the required clinical study; e.g. will require comparison of cancer detection-rates between modalities by independent, blinded-readers. Therefore, “the device is not intended to be used as a replacement for screening mammography”.   Third savvy: The most expensive component, in time and money, in a regulatory pathway are the clinical studies. A cost-effective regulatory strategy is linked to good understanding of the market segmentation. Identifying what kind of IFU differentiates your products from its competition in a large enough niche-market is key. It will also lead to the simplest clinical-study design possible. As an entrepreneur, I cannot help congratulating U-Systems’ team for pulling through continuous hurdles to reach the point all medical device startups are hoping for. They certainly picked up the right item to focus their efforts on: i.e. PMA approval for breast cancer screening. Finally, I will reiterate my vision that embedding real-time tissue characterization in an ultrasound system, capable of performing fast and standardized full breast scanning is: a. Technologically achievable; and b. in the long-term, will be an excellent alternative to mammography for breast cancer screening. Additional readings: Two studies related to  somo•v® will be discussed at the 2012 RSNA meeting: “ A study led by Dr. Rachel Brem of George Washington University Medical Center: ABUS plus mammography finds cancer early in women with dense tissue  Brem’s study found that ABUS enabled detection of early-stage cancers in women with dense breasts, giving healthcare providers time to start early treatment. In all, 88% of cancers found by ABUS alone in a group of 15,000 women were grade 1 or 2.” “A study presented by Maryellen Giger, PhD, of the University of Chicago: ABUS boosts mammography’s performance  this study results showthat adding ABUS to mammography for women with dense breast tissue improved sensitivity by 23.3 percentage points, from 38.8% for mammography alone to 63.1% for mammography plus ABUS.” As I mentioned already, there are other ultrasound modalities out there, some are ABUS and some are not. All are adjunct to mammography screening. Related studies will also be presented during that same meeting.

UPDATE (04-Aug-2013)

Here below is a recent publication on  the use of ABUS for better detection of breast cancer in patients presented with mammographically dense breast.

Improved breast cancer detection in asymptomatic women using 3D-automated breast ultrasound in mammographically dense breasts

  • Breast Cancer Research Institute, Nova Southeastern University College of Medicine, 5732 Canton Cove, Winter Springs, FL 32708, USA

Abstract

Automated breast ultrasound (ABUS)was performed in 3418 asymptomatic women with mammographically dense breasts. The addition of ABUS to mammography in women with greater than 50% breast density resulted in the detection of 12.3 per 1,000 breast cancers, compared to 4.6 per 1,000 by mammography alone. The mean tumor size was 14.3 mm and overall attributable risk of breast cancer was 19.92 (95% confidence level, 16.75 – 23.61) in our screened population. These preliminary results may justify the cost-benefit of implementing the judicious us of ABUS in conjunction with mammography in the dense breast screening population.

Keywords

  • Breast ultrasound;
  • 3-dimensional sonography;
  • Breast screening;
  • Dense breast;
  • Breast cancer;
  • Cancer detection

1. Introduction

Mammographic density as an independent risk factor for developing breast cancer has been documented since the 1970’s [1]. The appearance of breast tissue is variable among women. The appearance of density on mammography is the result of the relative proportion of breast stroma, which is less radiolucent compared to fat, accounting for increased breast density. Wolfe classified breast density as an independent risk factor for breast cancer in women [2] and [3]. Approximately 70 to 80% of breast cancers occur in women with no major predictors [4][5] and [6]. Population-based screening for early detection of breast cancer is therefore the primary strategy for reducing breast cancer mortality. Mammography has been used as the standard imaging method for breast cancer screening, with reduction in breast cancer mortality [7]. Breast density significantly reduces the ability to visualize cancers on mammography. The number of missed cancers is substantially increased in mammographically dense breasts, where the sensitivity is reported as low as 30 to 48%. [8]; and the odds of developing breast cancer 17.8 times higher [9]. Hand held ultrasound (HHUS) has been used to optimize the detection of cancers in mammographically dense breasts, but is limited due to technical factors, such as breast size, considerable user variability and reproducibility, technical skill, and time constraints, precluding HHUS as an effective screening modality for breast cancer [10][11] and [12]. Kelly described the use of 3D-automated breast ultrasound (ABUS) as an adjunct to mammography in the evaluation of non-palpable breast cancers in asymptomatic women. ABUS with mammography resulted in an increase in diagnostic yield from 3.6 per 1,000 with mammography alone, to 7.2 per 1,000 by adding ABUS, resulting in a mammography miss rate of 3.6 per 1,000 [13]. However, one of the limitations of the study was that it did not isolate dense breasts as an independent risk factor for developing breast cancer, where the detection rate should be expected to be higher. ABUS is FDA-approved in the United States for screening of women with dense breast parenchyma [14]. The purpose of this study was to demonstrate that ABUS increases the detection of non-palpable breast cancers in mammographically dense breasts when used as an adjunct diagnostic modality in asymptomatic women. This resulted in the subsequent detection of cancers missed by mammography of smaller size and stage, justifying the basis for the judicious use of implementing ABUS in conjunction with mammography in the dense breast screening population. The tabulated data was extrapolated based on known mammography screening utilization to show a cost-benefit of additional ABUS as a population based screening method.

2. Methods

2.1. Selection of participants

This study and the use of patient electronic health records were approved by an ethics committee appointed by the institute Board of Directors. The study design included two study groups, the control and test groups, in successive years. Each group was followed prospectively for 1 year. The control group consisted of women screened by digital mammography alone and stratified for breast density based on a Wolf classification of 50% or greater breast density (defined as the ‘mammographically dense breast’ for the purpose of this study). The second group consisted of women initially screening by digital mammography as having mammographically dense breasts, followed by automated breast ultrasound (ABUS). Each group was carefully selected on the basis of breast density and having no major pre-existing predictors of breast cancer, such personal or family history of breast cancer, or BRCA gene positive. In addition, the test group patients were not included in the screening group so as to eliminate impact on the results of the test group patients. The control group consisting of 4076 asymptomatic women designated as Wolf classification of 50% or greater breast density underwent stand-alone screening digital mammography between January 2009 and December 2009 using digital mammography (Selenia, Hologic Inc., Bedford, MA USA). The sensitivity, specificity, positive predictive value, and negative predictive value for biopsy recommendation were determined, in addition to data collection regarding the size and stage of cancers missed by mammography. The test group, consisting of 3418 asymptomatic women designated as Wolf classification of 50% or greater breast density, underwent stand-alone screening digital mammography between January 2010 and May 2011 using digital mammography (Selenia, Hologic Inc., Bedford, MA USA). This was followed by automated whole breast ultrasound (Somo-V. U-Systems, Sunnyvale, CA USA). The mammography-alone results were not used as control results in order to eliminate potential bias introduced by ABUS results on the mammography interpretations. In addition, mammography results were interpreted independently from ABUS results so as not to introduce bias. The sensitivity, specificity, positive predictive value, and negative predictive value for biopsy recommendation were determined, in addition to derived statistical data regarding the relative risk, and odds ratio for developing breast cancer.

2.2. Assessment of mammographic density

Mammographic density was assessed independently by radiologists on a dedicated mammography viewing workstation equipped with 5-Megapixel resolution. The radiologists were FDA-qualified in mammography, with at least 10 years experience in breast ultrasound, 24 months of which included ABUS. Two radiologists interpreted both the mammography and ABUS examinations under identical viewing conditions of 5-Megapixel resolution. The mammograms and ABUS studies were double read by two radiologists, with final consensus determination for each case. Mammograms were evaluated according to one of five categories of density (0%, 1 to 24%, 25 to 49%, 50 to 74%, and 75 to 100%) and only mammograms with breast density of 50% or greater were included in the control and test study groups.

2.3. 3D-Automated breast ultrasound evaluation

3D-Automated Breast Ultrasound (ABUS) is a computer-based system for evaluating the whole breast. The whole breast ultrasound system (Somo-V, U-Systems, Sunnyvale, CA USA) was used in combination with a 6 to 14 MHz broadband mechanical transducer attached to a rigid compression plate and arm, producing over 300 images per image acquisition obtained as coronal sweeps from the skin to the chest wall. The mechanical arm controls transducer speed and position, while a trained ultrasound technologist maintains appropriate contact pressure and vertical orientation to the skin. Interpretation and reporting time for an experienced radiologist is approximately 10 minutes per examination. The radiologist has cine functionality to simultaneously view breast images in the coronal, sagittal, and axial imaging planes.

2.4. Data collection

ABUS scan data was collected for location and size of breast masses and recorded in a radial or clock orientation consistent with American College of Radiology reporting lexicon. Studies were reported according to the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) six-point scale (0=incomplete, needs additional assessment; 1=normal; 2=benign; 3=probably benign; 4=suspicious; 5=highly suggestive of malignancy) [15] and [16]. For BI-RADS scores of 1, 2, and 3 on ABUS, patients were followed prospectively for 1 year to exclude cancers missed on both mammography and ABUS. For BI-RADS scores of 4 and 5, stereotactic hand held ultrasound (HHUS) biopsy was performed using 14 gauge or larger percutaneous biopsy. HHUS was employed because ABUS is presently not equipped with biopsy capability. If a benign non-high risk lesion was diagnosed, such as simple breast cysts, no further tissue sampling was performed. All non-cystic lesions were biopsied. Cystic lesions were identified as anechoic, thin walled lesions with posterior acoustic enhancement. All pathology proven breast malignancies were further staged using contrast volumetric/whole breast MR imaging (1.5T HDe Version 15.0/M4 with VIBRANT software, GE Medical Systems, Waukesha, WI USA.) with computer assisted detection (CADStream software, Merge Healthcare, Belleview WA USA). A final pathological stage was assigned by the pathologists in the usual manner in accordance with the American Joint Committee on Cancer (AJCC) TNM system guidelines. The pathologists were blinded to patient participation in the study and the method of cancer detection.

2.5. Statistical analysis

Calculations were made of the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), relative risk, odds risk, and attributable risk of breast cancer using MedCal version 12.2.1 software. Exact 95% confidence intervals (CI) were calculated for diagnostic yield. Statistical methods involved the Chi-square test statistic, which was used to compare the number of cancers detected by ABUS, based on the size of cancer. P-values of less than .05 were considered to indicate statistical significance. Attributable risk (AR) was calculated according to the following formula: AR=(RR − 1)Pc ÷ RR, where RR denotes relative risk of greater than 50%, and Pc prevalence of density of greater than 50% in case patients[17][18] and [19].

3. Results

Comparable interobserver diagnostic reliability (Kappa value of 0.98) was observed with mammography and ABUS examinations. In the control group (N=4076), the median age of participants with breast cancer (N=19) at the time of biopsy was 54 years, distributed as follows: 26% (5 out of 19) cancers occurred in women younger than age 50; 63% (12 out of 19) in women 50 to 69 years; and 11% (2 out of 19) over the age of 70 years. All cancers (N=19) were biopsy proven invasive ductal carcinoma. The sensitivity and specificity of stand-alone digital mammography were 76.00% (95% CI: 54.87% – 90.58%) and 98.2% (95% CI: 97.76% – 98.59%). The positive predictive value was 20.43% (95% CI: 12.78% – 30.05%) with a breast cancer prevalence rate of 0.60% (95% CI: 12.78% – 30.05%). The cancer detection rate was 4.6 per 1,000, with mean tumor size detected by mammography (N=19) of 21.3 mm. The average size of missed breast cancer (N=6) was 22.3 mm. The node positivity rate was 5% (1 of 19 cases). In the ABUS study group (N=3418), the median age of participants with breast cancer (N=42) at the time of biopsy was 57 years, distributed as follows: 17% (7 out of 42) cancers occurred in women younger than age 50; 64% (27 out of 42) in women 50 to 69 years; and 19% (8 out of 42) over the age of 70 years. The sensitivity and specificity of ABUS were 97.67% (95% CI: 87.67% – 99.61%) and 99.70%, (95% CI=99.46% – 99.86%), respectively, in mammographically dense breasts. The positive predictive value of ABUS was 80.77% (95% CI=67.46% – 90.36%), with a breast cancer prevalence rate of 1.25% (95% CI: 0.91% – 1.69%). The odds ratio of breast cancer in mammographically dense breasts determined by ABUS was 2.65 (95% CI: 1.54 – 4.57; P=0.0004). The cancer detection rate was 12.3 per 1,000. A 2.6-fold increase in cancer detection rate was observed between ABUS added to digital screening mammography compared to stand-alone digital screening mammography. Invasive breast cancer accounted for 81% (42 out of 52) solid breast masses detected by ABUS, of which 93% (39 out of 42) were invasive ductal carcinomas, and 7% (3 out of 42) were invasive lobular carcinomas. The mean tumor size detected by ABUS in patients with breast cancer (N=42) was 14.3 mm, distributed as follows: Stage 1A disease accounted for 83% (35 out of 42) of cases; 12% were Stage 2A (5 out of 42), and 5% were Stage 3A (2 out of 42). Stage 3A disease was associated with multifocal disease in both cases, one of which also was Level 1 axillary lymph node positive. The node positivity rate was 2% (1 in 42) of cases. The false positive rate of ABUS was 19.3%, with a negative predictive value of 99.97% (95% CI 99.83% – 100.00%). The pathologies associated with false positive results (N=10) were fibroadenomas and atypical epithelial neoplasms. We also used our data to extrapolate the theoretical cost-benefit of ABUS screening applied to a large screening population in the United States. Our analysis relied on the following assumptions: (1) Global Centers for Medicare and Medicaid reimbursement rate of breast ultrasound of $71 [20]; and (2) Estimated mean doubling time of a missed cancer of 250 days at the 95th percentile [21] and [22]. According to previously cited cancer kinetics models, a missed breast cancer should be clinically evident within 9 months[23]. When we considered the mean breast cancer size in our positive test subject group, 14.3 mm (N=42), we extrapolated a theoretical missed cancer size of 29.2 mm at 9 months in mammographically dense breasts, representative of Stage 2 or greater disease. In control subjects, a mean breast cancer size of 22.3 mm was consistent with stage 2 breast cancer. Incremental treatment cost assumptions, based on the global Centers for Medicare and Medicaid reimbursement rate between Stage 1 and Stage 2 breast cancer, were $24,002 and $34,469, respectively, for a cost differential of $10,467 [24]. Accordingly, the aggregate costs of screening 3418 ABUS patients in this study were $239,260, compared to the estimated aggregate costs of additional treatment in 26 potentially missed cancers (based on previously noted theoretical assumptions) of $275,557 based on a cancer miss rate of 0.77% (or 7.7 per 1,000).

4. Discussion

Table 1 shows the clinical indications for ordering an ABUS examination. Table 2 shows the distribution of breast cancer size according to age in the control and test study groups. The test group showed no statistical difference between size of the cancer and patient age at presentation. A significant increase in tumor size in the over 70 patients in control subjects was attributed to the more advanced tumor stage at presentation.Table 3 shows that stand-alone digital mammography was less sensitive than ABUS in breast cancer detection, with a 4-fold increase in positive predictive value of ABUS compared to stand-alone mammography in dense breasts. Our results showed that mammographic density of 50% or more was associated with an increased risk of breast cancer and resulted in a significant miss rate in asymptomatic women. Table 4 shows a statistically significant age-related attributable risk of developing breast cancer for mammographic density of 50% or greater. These observations are consistent with other studies which have shown an increased risk of breast cancer in dense breasts following negative mammography screening [2],[3][8] and [9]. We observed that breast cancer risk was highest in patients over age 70, where increased breast density was associated with an attributable risk of 29.6 (95% CI, 21.5 – 40.8). Fig. 1 shows box plots comparing case patients and control subjects according to age, with tumor sizes shown as a function of the odds ratio, relative risk, and attributable risk for each age category.

Table 1. Clinical criteria for ABUS screening
• As a supplement to mammography, screening for occult cancers in certain populations of women (such as those with dense fibroglandular breasts and/or with elevated risk of breast cancer);
• Imaging evaluation of non-palpable masses in women under 30 years of age who are not at high risk for development of breast cancer, and in lactating and pregnant women; and
• BI-RADS (American College of Radiology Breast Imaging Reporting and Data System) scoring classification class III, heterogeneously dense, with 50% to 74% or 75% to 100% breast density on mammography, without palpable mass.
Table 2. Breast cancer size according to method detection

T2

Table 3. Detection of breast cancer according to method
t3 

Table 4. Risk of breast cancer according to method detection

t4
  1-s2.0-S0899707112002872-gr1

Fig. 1. Breast Cancer Staging and Risk Assessment by Screening Method Detection. Box plots comparing case patients and control subjects according to age (boxes A through D). Tumor sizes are shown as a function of the odds ratio, relative risk, and attributable risk for each age category. Bars represent the highest and lowest observed values with respect to individual variables (individually labeled with arrows).

 

Our study also showed that 3D-Automated Breast Ultrasound (ABUS) was an effective screening modality in mammographically dense breasts. Our extrapolated data suggest a breast cancer miss rate of 7.7 per 1,000 in mammographically dense breasts in asymptomatic women, which is higher compared to the cancer miss rate of 3.6 per 1,000 reported by Kelly using ABUS [13]. We attribute the increased breast cancer miss rate due to breast density, which was isolated as the principal risk factor in our study. Other studies have shown that the attributable risk of breast cancer for a mammographic density of 50% or greater was 40% for all cancers detected less than 12 months after a negative screening mammogram, and as high as 50% in women less than the age of 50. This marked increase in the risk of breast cancer associated with mammographic density of 50% or greater up to 12 months following screening directly reflects cancers that were present at the time of screening but went undetected due to masking by dense breast parenchyma [25],[26][27][28] and [29]. In the final analysis, there is the issue of the theoretical cost-benefit of adding ABUS screening to mammography in an otherwise healthy population. The importance of screening mammographically dense breasts with ABUS has particular relevance based on the small size and early stage of breast cancers. Our study showed a mean tumor size of 14.3 mm, representing stage 1 disease, which was present in 81% of patients. From our data, we derived theoretical population-based costs as a basis for the cost-benefit of ABUS in the United States population. Our study compared the incremental costs of screening versus the costs of added treatment related to a change in the staging of missed cancers from Stage 1 to Stage 2. The costs of additional treatment outweighed the costs of screening by $32,808, which calculated to $9.60 added healthcare cost per patient in the 3418 participants in the study. In the United States, 48 million mammograms were performed annually, with a reported estimated miss rate of 10% [30]. When comparing control versus test patients, our study suggests a theoretical miss rate of 7.7 cancers per 1,000 mammograms, or 0.77%, which is considerably lower than the reported missed rate of 10%. Based on these theoretical assumptions, annual added ABUS screening of the entire U.S. population would cost $3.40-billion. However, in actual practice, ABUS would be used only in the mammographically dense breast, which would potentially reduce the screening costs by at least a factor of 0.8, bringing the cost closer to $2.72-billion. By contrast, the incremental costs of added treatment associated with stage 2 compared to stage 1 breast cancer in the U.S. population would be $3.82-billion, assuming a conservative cost basis of $10,467 per patient.. The cost-benefit of early detection of stage 1 disease results in a theoretical per capital annual cost savings of $22.75 per screened patient in the U.S. population, according to our model. However, we have no actual or derived data to support improved breast cancer mortality with the addition of ABUS as a universal screening modality. This is one of the major limitations of our study because actuarial analyses used to justify screening modalities are typically based on mortality statistics. With respect to five year survival statistics between stage 1 and stage 2 breast cancers, of 98% and 80%, respectively, one could construe the potential for a theoretical quality-of-life benefit based on judicious ABUS screening. Another limitation of our study is the relatively small screening population used in our study, emphasizing the need for continued research in order to validate ABUS as a viable and cost-effective population-based screening modality, which should be stratified for other risk factors for breast cancer, such as: personal or family history of breast cancer, BRCA genetic results, environmental factors (late parity, previous exposure to ionizing radiation, exogenous estrogen, smoking, and alcohol use), early menarche/late menopause, and ethnic/racial differences. At most imaging centers, mammography is the only screening method for breast cancer detection. Our study corroborates with the data derived from other studies that the principal mechanism for breast cancer in dense breast parenchyma is not rapid growth, but rather, the masking of coincident cancers that are missed on screening mammograms [9]. These findings further suggest that the addition of mammographic screening in patients with dense breast parenchyma is likely not to increase diagnostic yield in the detection of breast cancers. Therefore, emphasis should be placed on alternative imaging techniques for such women. To conclude, our study of a small representative dense breast screening population showed that the addition of ABUS was more effective than digital mammography alone. This study provides a platform for using ABUS as cost-effective approach to breast cancer detection in the judicious screening of asymptomatic women with excessive mammographic density, in whom the greatest risk is between screening mammography examinations.

References

    • [2]
    • JN Wolfe
    • Breast patterns as an index of risk for developing breast cancer
    • AJR Am J Roentgenol, 126 (1976), pp. 1130–1137
    • [8]
    • MT Mandelson, N Oestreicher, PL Porter et al.
    • Breast density as a predictor of mammographicdetection: comparison of interval and screen detected cancers
    • J Natl Cancer Inst, 92 (2000), pp. 1081–1087
    • [9]
    • NF Boyd, H Guo, LJ Martin et al.
    • Mammographic density and the risk and detection of breast cancer
    • N Engl J Med, 356 (2007), pp. 227–236
    • [10]
    • W Buchberger, P DeKoekkoik-Doll, P Springer, P Obrist, M Dunser
    • Incidental findings on sonography of the breast: clinical significance and diagnostic workup
    • AJR Am J Roentgenol, 173 (1999), pp. 921–927
    • [11]
    • DB Kopans
    • Breast cancer screening with ultrasonography
    • Lancet, 354 (1999), pp. 2096–2097
    • [12]
    • WA Berg, JD Blume, JB Cormack et al.
    • Combined screening with ultrasound and mammography versus mammography alone in women at elevated risk of breast cancer
    • JAMA, 299 (2008), pp. 2151–2163
    • [13]
    • KM Kelly, J Dean, WS Comulada, SJ Lee
    • Breast cancer detection using automated whole breast ultrasound and mammography in radiographically dense breasts
    • Eur Radiol, 20 (2010), pp. 734–742
    • [14]
    • United States Food and Drug Organization. Breast transilluminators. 74 FR 16214, April 9, 2009; Docket No. FDA-2012-N-0001, April 12, 2012.
    • [15]
    • C D’Orsi, L Bassett, W Berg et al.
    • ACR Breast Imaging Reporting and Data System (BIRADS)
    • (4th ed.)American College of Radiology, Reston, VA (2003)
    • [16]
    • BM Geller, WE Barlow, R Ballard-Barbash et al.
    • Use of the American College of Radiology BI-RADS to report on the mammographic evaluation of women with signs and symptoms of breast disease
    • Radiology, 222 (2002), pp. 536–542
    • [17]
    • PF Griner, RJ Mayewski, AI Mushlin, P Greenland
    • Selection and interpretation of diagnostic tests and procedures
    • Ann Intern Med, 94 (1981), pp. 555–600
    • [18]
    • JA Hanley, BJ McNeil
    • The meaning and use of the area under a receiver operating characteristic (ROC) curve
    • Radiology, 143 (1982), pp. 29–36
    • [19]
    • CE Metz
    • Basic principles of ROC analysis
    • Semin Nucl Med, 1978 (1978), pp. 283–298
    • [20]
    • Centers for Medicare, Medicaid
    • Contracted intermediary carrier fee schedule
    • First Coast Service Options, Inc., St. Augustine, FL (2004)
    • [21]
    • T Kuroishi, S Tominaga, T Morimoto et al.
    • Tumor growth rate and prognosis of breast cancer mainly detected by mass screening
    • Jpn J Cancer Res, 81 (1990), pp. 454–462
    • [22]
    • L Heuser, JS Spratt, HC Polk
    • Growth rate of primary breast cancer
    • Cancer, 43 (1979), pp. 1888–1894
    • [23]
    • JS Michaelson, E Halpern, DB Kopans
    • Breast cancer computer simulation method for estimating optimal intervals for screening
    • Radiology, 212 (1999), pp. 551–560
Corresponding author. Breast Cancer Research Institute, Nova Southeastern University College of Medicine, 5732 Canton Cove, Winter Springs, FL 32708, USA. Tel.: 1 407 699 7787.

Copyright © 2013 Elsevier Inc. All rights reserved.

Read Full Post »