Improving Interpretability in Medical Imaging Diagnosis using Adversarial Training

TL;DR

This paper demonstrates that adversarial training enhances the interpretability of CNNs in skin cancer diagnosis by producing sharper, more coherent saliency maps that better highlight relevant lesion features.

Contribution

It introduces the use of adversarial training to improve CNN interpretability in medical imaging, showing sharper saliency maps and highlighting key lesion characteristics.

Findings

Adversarially trained CNNs produce sharper saliency maps.

Saliency maps highlight regions with significant color variation.

Fine-tuning enhances saliency map quality.

Abstract

We investigate the influence of adversarial training on the interpretability of convolutional neural networks (CNNs), specifically applied to diagnosing skin cancer. We show that gradient-based saliency maps of adversarially trained CNNs are significantly sharper and more visually coherent than those of standardly trained CNNs. Furthermore, we show that adversarially trained networks highlight regions with significant color variation within the lesion, a common characteristic of melanoma. We find that fine-tuning a robust network with a small learning rate further improves saliency maps' sharpness. Lastly, we provide preliminary work suggesting that robustifying the first layers to extract robust low-level features leads to visually coherent explanations.