New DynSHAP framework enhances AI explainability for medical survival analysis

Researchers have developed DynSHAP, a new framework designed to provide explainability for dynamic survival analysis (DSA) models. Existing methods struggle with the longitudinal and irregular nature of patient data and the functional survival outputs common in DSA. DynSHAP extends SHAP (SHapley Additive exPlanations) by treating time-feature pairs as game players and introduces Temporal DynSHAP, which accounts for temporal dependencies in features to generate more accurate explanations. Initial testing on synthetic and real-world clinical datasets shows DynSHAP's ability to produce faithful attributions, enabling medical experts to understand which patient information influenced predictions and at what time. AI

IMPACT Enhances trust and adoption of AI in clinical settings by making complex survival analysis models more interpretable.

RANK_REASON The item describes a new research paper introducing a novel framework for explainability in a specific machine learning domain. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• CORE Recommender
• DagsHub
• DynSHAP
• Gotit.pub
• Hugging Face
• IArxiv Recommender
• Litmaps
• ScienceCast
• scite Smart Citations
• SHAP
• Temporal DynSHAP

• paper
• other

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