A research team at RIKEN Biomedical Spatial Science Team, led by Jun Akatsuka (RIKEN;Nippon Medical School), Kotaro Tsutsumi (RIKEN; University of California), and Yoichiro Yamamoto (RIKEN; Tohoku University), has tackled the domain shift problem in cancer AI.
Domain shifts occur when AI encounters variations in data that differ from their training conditions. The present study addresses dual-domain shifts spanning both hospital settings and prostate cancer specimen types. Moving beyond purely data-driven approaches, the team introduces an intermediate reasoning step guided by expert–derived clinical knowledge. This clinically Informed intermediate reasoning enables more robust, generalizable, and clinically applicable cancer prognostication. The findings were published in npj Digital Medicine (January 9, 2026).
For more information, please see the following site.
npj Digital Medicine
