Image Classification with Consistent Supporting Evidence

TL;DR

This paper introduces methods to ensure machine learning models in healthcare provide consistent supporting evidence for their predictions, enhancing trust and interpretability without sacrificing performance.

Contribution

It proposes measures and regularizers to promote evidence consistency in models, demonstrated on edema severity grading from chest radiographs.

Findings

Consistent models achieve competitive accuracy.

Supporting evidence improves interpretability.

Regularizers effectively promote evidence consistency.

Abstract

Adoption of machine learning models in healthcare requires end users' trust in the system. Models that provide additional supportive evidence for their predictions promise to facilitate adoption. We define consistent evidence to be both compatible and sufficient with respect to model predictions. We propose measures of model inconsistency and regularizers that promote more consistent evidence. We demonstrate our ideas in the context of edema severity grading from chest radiographs. We demonstrate empirically that consistent models provide competitive performance while supporting interpretation.