Narrative-based explainable AI for precision oncology survival prediction.

SurvXAI integrates Random Survival Forest survival prediction with SurvSHAP and locally deployed LLM narrative generation to produce coherent, clinically plausible explanations.

• Why it matters: Explainable survival prediction in precision oncology is hard because interpretable models can sacrifice accuracy, while newer ML methods can be hard for clinicians to understand.
• What they did: The pipeline trains a Random Survival Forest, derives patient-specific feature contributions via SurvSHAP, and uses locally deployed LLMs to convert numeric explanations into natural-language narratives, validated on six oncological datasets.
• The result: In a blind evaluation with seven human experts, larger LLMs yielded more coherent explanations with accurate feature ranking and clinically plausible reasoning, supporting trustworthy personalized interpretation.

AI summary of the abstract · may contain errors · not medical advice · read the paper.