MergeSurv: Merging-Based Continual Learning for Survival Analysis on Whole-Slide Images

MergeSurv is a continual learning framework for survival analysis using Whole Slide Images (WSIs). It fine-tunes models for sequential cancer tasks and merges their parameters to prevent catastrophic forgetting. This allows the model to learn from many datasets while maintaining high performance, leading to better risk prediction.

Open the full episode

More from AI Radio