Abstract:

Breast cancer screening using the standard mammography exam currently exhibits a high false recall rate (11.6% for women in the U.S.). Only a low proportion (0.5%) of women who were recalled for additional workup were actually found to have breast cancer. As a result of the unnecessary stress and follow-up work from these false recalls, patients exhibit increased anxiety and potential increase in morbidity, as well as create an increased pressure on clinical workload and medical costs. There has been an interest in reducing these false recalls by better distinguishing negative and positive mammogram images. In our research, we focus on whether deep learning models can distinguish not only positive and negative mammograms, but also mammograms which were recalled and subsequently determined to be benign.

Distinguishing potentially recalled but biopsy benign images from both malignant and negative images represents a critical need and a practical approach to aid radiologists. Thus, the purpose of this study was to investigate end-to-end deep learning CNN models for automatic identification of nuanced imaging features to distinguish mammogram images belonging to negative, recalled-benign, and malignant cases aimed to improve clinical mammographic image interpretation and reduce unnecessary recalls.

BIO:

Sarah Aboutalib completed her Bachelors of Sciences in Cognitive Science in 2005 at the University of California, San Diego, and completed her Ph.D. in Computer Science at Carnegie Mellon University, graduating in 2011. Her thesis research was based in computer vision, specifically research in multiple-cue object recognition from video data. Her current interests include applying cognitive science, machine learning and deep learning techniques to non-profit and community beneficial contexts such as healthcare and education. Her most recent research was as a Postdoctoral Scholar at the University of Pittsburgh Biomedical and Informatics Department studying the application of CNN models to breast cancer mammography data.

Homepage: www.sarahaboutalib.com

Sponsored in part by:   Facebook Reality Labs Pittsburgh