Abstract P3-04-11: Longitudinal Changes of Contralateral Breast Mammographic Artificial Intelligence Algorithms Score in Ductal Carcinoma In Situ patient with Tamoxifen

Abstract Background Mammographic Artificial intelligence (AI) algorithms (Lunit insight MMG) draws attention as a diagnostic support tool for breast cancer detection. Lunit insight MMG provides a location suspected of breast cancer with a heatmap and a score reflecting the probability of the presence of suspicious areas. We investigated whether the Lunit insight MMG score is relevant for predicting the response to adjuvant tamoxifen. Methods Patients diagnosed with DCIS and underwent treatment at Seoul National University Hospital in 2010 were retrospectively enrolled. Clinical characteristics, tamoxifen use, survival data, and mammography images were extracted from the electronic medical records, and Lunit insight MMG scores were calculated retrospectively. We classified two groups according to tamoxifen treatment and compared the score change of contralateral breast from baseline to 5 years after surgery for DCIS. Change categories of Lunit insight MMG score included maintaining high risk, maintaining low risk, increasing from low risk to high risk, and decreasing from high risk to low risk. Results Of 100 patients, 50 (50%) had undergone tamoxifen treatment (group 1) and 50 (50%) had not (group 2). The median age of the patients was 48.4 years for group 1 and 51.5 years for group 2 (p=0.172). The median follow-up duration was 8.7 years for the whole cohort. Using Lunit insight MMG score, more patients in group 1 decreased in contralateral breast cancer (CBC) risk compared with group 2 (6.0% vs. 2.0%; P=.008). No patients in group 1 had an increase in CBC risk while 9 patients increased in group 2 (0% vs. 18.0%; P =.008). There was no ipsilateral breast cancer recurrence for the whole cohort, and two patients experienced contralateral invasive breast cancer in group 2. In two patients with CBC, the Lunit insight MMG score increased five years after surgery, one year and three years before the CBC diagnosis. Conclusions Longitudinal Changes of Mammographic AI algorithms Score may be a predictive surrogate marker for response to tamoxifen therapy in hormone receptor-positive DCIS. Citation Format: Changjin Lim, Eun Kyung Park, Hong-Kyu Kim, Eunhye Kang, Ji-Jung Jung, Hyunsu Yeoh, Jang-il Kim, Jung Whan Chun, Han-Byoel Lee, Hyeong-Gon Moon, Wonshik Han. Longitudinal Changes of Contralateral Breast Mammographic Artificial Intelligence Algorithms Score in Ductal Carcinoma In Situ patient with Tamoxifen [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P3-04-11.