New framework DICE enhances AI pathology model reliability with uncertainty estimation

Researchers have developed a new framework called DICE to improve the reliability of pathology foundation models (PFMs) for whole-slide image analysis. This framework ensembles multiple frozen PFMs and uses their disagreement to estimate prediction uncertainty. By aligning these ensemble members through deep mutual learning, DICE provides more meaningful confidence estimates and can even localize abnormalities without explicit supervision. Evaluations on three benchmarks show that DICE accurately flags unreliable predictions in various settings and matches or surpasses state-of-the-art methods in classification, calibration, and localization, moving PFMs closer to clinical decision support. AI

IMPACT Enhances trust and clinical applicability of AI in pathology by providing reliable uncertainty estimates for model predictions.

RANK_REASON The cluster contains a research paper detailing a new framework for AI model uncertainty estimation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• pathology foundation models

• paper
• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →