Interpreting Medical Image Classifiers by Optimization Based Counterfactual Impact Analysis
This work addresses the need for robust interpretation in automated decision support systems for medical imaging, which is incremental as it improves upon existing saliency mapping techniques.
The authors tackled the problem of interpreting medical image classifiers by developing a model-agnostic saliency mapping framework that replaces heuristic methods with a neighborhood-conditioned inpainting approach to avoid anatomically implausible artifacts, resulting in quantitatively and qualitatively more precise localization than existing state-of-the-art methods on public mammography data.
Clinical applicability of automated decision support systems depends on a robust, well-understood classification interpretation. Artificial neural networks while achieving class-leading scores fall short in this regard. Therefore, numerous approaches have been proposed that map a salient region of an image to a diagnostic classification. Utilizing heuristic methodology, like blurring and noise, they tend to produce diffuse, sometimes misleading results, hindering their general adoption. In this work we overcome these issues by presenting a model agnostic saliency mapping framework tailored to medical imaging. We replace heuristic techniques with a strong neighborhood conditioned inpainting approach, which avoids anatomically implausible artefacts. We formulate saliency attribution as a map-quality optimization task, enforcing constrained and focused attributions. Experiments on public mammography data show quantitatively and qualitatively more precise localization and clearer conveying results than existing state-of-the-art methods.