investigator_user investigator user funding collaborators pending menu bell message arrow_up arrow_down filter layers globe marker add arrow close download edit facebook info linkedin minus plus save share search sort twitter remove user-plus user-minus
  • Project leads
  • Collaborators

Computational pathology to predict breast cancer risk in benign breast disease

Andrew H Beck

0 Collaborator(s)

Funding source

National Institutes of Health (NIH)
Benign breast disease (BBD) is diagnosed when a woman undergoes a breast biopsy for an abnormality found through physical breast exam or screening mammogram and pathological analysis of the biopsy shows no evidence of malignancy. Approximately 80% of breast biopsies reveal a benign lesion. The identification of atypia in BBD is a well-established, strong risk factor for future breast cancer; however, the diagnosis of atypia in BBD is one of the most challenging areas of diagnostic pathology, and it has proven difficult to create standardized objective criteria for the diagnosis of atypical lesion in BBD. Lobular involution has recently been shown to be significantly associated with breast cancer risk; however, there are currently no clinically available tools to quantitate lobular involution, and consequently, this feature is not currently incorporated into pathology reports. Stromal characteristics are known to play a crucial role in all stages of breast carcinogenesis; however, the association of quantitative stromal characteristics and breast cancer risk has never been evaluated. In this two year project, we will extend our previous work based in invasive cancer to develop a computational pathology program for the quantitative assessment of both established and novel morphological features in normal breast and benign breast disease lesions (Aim 1). To achieve this aim, we will use the Nurses' Health Study (NHS) Incident BBD cohort, which contains histological slides from a total of 1758 NHS participants with BBD. All cases have been previously reviewed and annotated by expert breast pathologists. These annotations will be used extensively in both the design and evaluation of the computational pathology platform. In Aim 2 of our study, we will examine associations between computational pathology (C- Path) features with future breast cancer risk. To achieve this aim, we will use the NHS BBD Breast Cancer Nested Case Control cohort, which consists of 613 women with BBD who went on to develop breast cancer matched to 2407 women who did not. Using this unique cohort, we will perform analyses to determine the association of established and novel C-Path derived morphological features with cancer risk and to determine the added value of utilizing C-Path to predict future cancer risk. The overriding goal of our project is to develop a new computational system for the objective, quantitative assessment of both established and novel morphologic characteristics of breast tissue in women with BBD. We aim to use this system to gain biological insight into morphologic factors associated with breast cancer risk and to improve the performance of breast cancer risk prediction models. If successful, our project will result in the development of a clinically applicable tool that will provide objective quantitative assessments of histopathological features in nonmalignant breast tissue to inform breast cancer risk prediction models and to guide clinical decisions. This development could represent a paradigm shift in how normal breast and benign breast disease pathology is measured and used in both clinical practice and translational breast cancer research.

Related projects