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.
@article{arxiv.2111.07048,
title = {Image Classification with Consistent Supporting Evidence},
author = {Peiqi Wang and Ruizhi Liao and Daniel Moyer and Seth Berkowitz and Steven Horng and Polina Golland},
journal= {arXiv preprint arXiv:2111.07048},
year = {2021}
}
Comments
13 pages, 6 figures, proceedings of the Machine Learning for Health NeurIPS Workshop, 2021