The field of oncology is witnessing a significant shift towards the development of interpretable AI models that can provide accurate and transparent predictions for patient outcomes. Researchers are focusing on creating multimodal frameworks that integrate clinical variables and medical imaging data to automate survival analysis. These models aim to provide transparent risk estimation and stratification, enabling personalized treatment plans. Noteworthy papers in this area include: SHAPoint, which presents a task-agnostic framework for point-based risk scoring via Shapley values, offering superior flexibility and reduced reliance on manual preprocessing. Automated and Interpretable Survival Analysis from Multimodal Data, which proposes an interpretable multimodal AI framework to automate survival analysis, achieving a C-index of 0.838 and outperforming clinical and academic baseline approaches.
Interpretable AI for Precision Oncology
Sources
Automated and Interpretable Survival Analysis from Multimodal Data
Towards Understanding Feature Learning in Parameter Transfer
SHAPoint: Task-Agnostic, Efficient, and Interpretable Point-Based Risk Scoring via Shapley Values
A Second-Order Perspective on Pruning at Initialization and Knowledge Transfer
Evaluating Foundation Models with Pathological Concept Learning for Kidney Cancer
Intuitions of Machine Learning Researchers about Transfer Learning for Medical Image Classification