Interpretable machine learning for time-to-event prediction in medicine and healthcare
This work addresses the need for interpretable AI in medical applications, enabling model debugging for developers and biomarker discovery for physicians, though it is incremental in extending existing interpretability techniques to survival analysis.
The paper tackled the lack of interpretable methods for time-to-event prediction in healthcare by introducing time-dependent feature effects and global importance explanations, demonstrating their use in analyzing biases in a multi-modal dataset of 1235 X-ray images and evaluating cancer survival models across 11 TCGA datasets.
Time-to-event prediction, e.g. cancer survival analysis or hospital length of stay, is a highly prominent machine learning task in medical and healthcare applications. However, only a few interpretable machine learning methods comply with its challenges. To facilitate a comprehensive explanatory analysis of survival models, we formally introduce time-dependent feature effects and global feature importance explanations. We show how post-hoc interpretation methods allow for finding biases in AI systems predicting length of stay using a novel multi-modal dataset created from 1235 X-ray images with textual radiology reports annotated by human experts. Moreover, we evaluate cancer survival models beyond predictive performance to include the importance of multi-omics feature groups based on a large-scale benchmark comprising 11 datasets from The Cancer Genome Atlas (TCGA). Model developers can use the proposed methods to debug and improve machine learning algorithms, while physicians can discover disease biomarkers and assess their significance. We hope the contributed open data and code resources facilitate future work in the emerging research direction of explainable survival analysis.