Sáez L, Rabanaque S, Casas J, Marqués F, Salembier P. Closing the Loop: Continuous AI Model Improvement Through Pathologist-Guided Feedback. In 21st European Congress on Digital Pathology. Barcelona: The European Society of Digital and Integrative Pathology; 2025.  (92.01 KB)

Abstract

AI integration in digital pathology requires continuous refinement, not just initial deployment. Annotating large datasets is time-consuming, and training an optimal model upfront is impractical. As part of the DigiPatICS project, our approach enables ongoing improvement of deployed algorithms through collaboration between pathologists and data scientists in a structured feedback loop with minimal workflow disruption. This accelerates database growth with task-specific annotations. Moreover, involving different end users reduces bias and enhances diversity and generalizability.

Projects

DigiPatICS: Projecte d'optimització del diagnòstic anatomopatòleg en xarxa als hospitals de l'Institut Català de la Salut a través de la digitalització i eines d'intel·ligència Artificial