Personalized Survival Prediction with Contextual Explanation Networks

TL;DR

This paper introduces a flexible recurrent network model that predicts individual cancer survival times and provides interpretable explanations based on patient attributes, enhancing both accuracy and transparency in personalized prognosis.

Contribution

The proposed model uniquely combines survival prediction with patient-specific explanations using simple linear regressions within a recurrent network framework.

Findings

Outperforms baseline models in survival prediction accuracy

Provides interpretable, patient- and time-specific explanations

Effective on multiple publicly available datasets

Abstract

Accurate and transparent prediction of cancer survival times on the level of individual patients can inform and improve patient care and treatment practices. In this paper, we design a model that concurrently learns to accurately predict patient-specific survival distributions and to explain its predictions in terms of patient attributes such as clinical tests or assessments. Our model is flexible and based on a recurrent network, can handle various modalities of data including temporal measurements, and yet constructs and uses simple explanations in the form of patient- and time-specific linear regression. For analysis, we use two publicly available datasets and show that our networks outperform a number of baselines in prediction while providing a way to inspect the reasons behind each prediction.