Improving Interpretability in Medical Imaging Diagnosis using Adversarial Training
This work offers an incremental improvement in interpretability for medical imaging diagnosis, specifically for clinicians using CNNs to detect skin cancer.
This paper explores how adversarial training impacts the interpretability of CNNs for skin cancer diagnosis. They found that adversarially trained CNNs produce significantly sharper and more visually coherent gradient-based saliency maps, which highlight diagnostically relevant regions like color variations within lesions.
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.