Abstract PR-04: Effect of breast cancer chemoprevention on a convolutional neural network-based mammographic evaluation using a mammographic dataset of women with atypical hyperplasia, lobular or ductal carcinoma in situ

Abstract

Background: Chemoprevention with anti-estrogens such as tamoxifen has been shown to lower mammographic density (MD), a strong predictor of breast cancer risk. However, measurement of MD is limited by variability in radiologists’ interpretations. We developed a novel, fully-automated convolutional neural network (CNN)-derived mammographic evaluation that is a more accurate predictor of breast cancer risk than MD. We evaluated whether chemoprevention with anti-estrogens is associated with a significant change in CNN breast cancer risk among women with atypical hyperplasia (AH), lobular or ductal carcinoma in situ (LCIS/DCIS). Methods: We conducted a retrospective cohort study using serial mammograms from women diagnosed with AH, LCIS, or DCIS at Columbia University Irving Medical Center (CUIMC) between 2007 and 2015. We collected mammograms at baseline (at diagnosis of AH/LCIS/DCIS or prior to initiation of chemoprevention) and 3-5 years follow-up. We extracted information from the electronic health record on age, race/ethnicity, menopausal status, body mass index (BMI), and chemoprevention uptake (yes/no). Briefly, each mammogram was normalized as a map of z-scores and resized to an input image size of 256 × 256. Then a contracting and expanding fully convolutional CNN architecture was composed entirely of 3 × 3 convolutions, a total of four strided convolutions instead of pooling layers, and symmetric residual connections. L2 regularization and augmentation methods were implemented to prevent over-fitting. CNN risk score was expressed as a continuous variable (0-1). We used 2-sample t-test to compare change in CNN risk score from baseline to follow-up among women who took chemoprevention compared to those who did not. We conducted multivariable linear regression adjusting for known breast cancer risk factors (age, BMI, menopausal status, race/ethnicity) to determine whether receipt of chemoprevention was associated with change in CNN risk score. Results: Among 728 evaluable women, mean age was 60.4 years (SD, 11.1), 70.4% were postmenopausal, and 248 (34.1%) received chemoprevention with anti-estrogens while 480 (65.9%) did not. Women who received chemoprevention compared to those who did not had a greater mean change in CNN risk score from baseline to 3-5 years of follow-up, -0.069 (SD, 0.278) and -0.019 (SD, 0.244), respectively (p=0.014). In multivariate analysis, women who received chemoprevention compared to those who did not had a 0.038 point greater decrease in CNN risk score (p=0.085, 95% confidence interval [CI]= -0.081, +0.005). Conclusions: We demonstrated that our CNN-based mammographic evaluation is modifiable with anti-estrogen therapy among high-risk women. Future studies should determine whether changes in CNN risk score are associated with the development of breast cancer, in order to further evaluate the CNN mammographic evaluation as a potential pharmacodynamic biomarker of response to breast cancer chemoprevention. Citation Format: Julia E. McGuinness, Vicky Ro, Aishwarya Anuraj, Haley Manley, Simukayi Mutasa, Richard Ha, Katherine D. Crew. Effect of breast cancer chemoprevention on a convolutional neural network-based mammographic evaluation using a mammographic dataset of women with atypical hyperplasia, lobular or ductal carcinoma in situ [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PR-04.