Speaker: Emre Fişne

Abstract: Breast cancer is a heterogeneous disease whose accurate characterization requires the integration of complementary imaging and biological information. This seminar presents a multimodal artificial intelligence framework for breast cancer subtyping by integrating mammography, ultrasound, and magnetic resonance imaging with molecular biomarker information. The proposed approach aims to learn complementary representations across imaging modalities while incorporating molecular characteristics associated with clinically relevant breast cancer subtypes. In addition to subtype prediction, the framework considers radiological features and BI-RADS assessment within a unified learning setting. The seminar will discuss the motivation, multimodal architecture, data preparation and training strategy, as well as the potential of integrating imaging and molecular information to support more comprehensive and clinically meaningful breast cancer characterization.