Improving trust and confidence in medical skin lesion diagnosis through explainable deep learning

Abstract

A key issue in critical contexts such as medical diagnosis is the interpretability of the deep learning models adopted in decision-making systems. Research in eXplainable Artificial Intelligence (XAI) is trying to solve this issue. However, often XAI approaches are only tested on generalist classifier and do not represent realistic problems such as those of medical diagnosis. In this paper, we aim at improving the trust and confidence of users towards automatic AI decision systems in the field of medical skin lesion diagnosis by customizing an existing XAI approach for explaining an AI model able to recognize different types of skin lesions. The explanation is generated through the use of synthetic exemplar and counter-exemplar images of skin lesions and our contribution offers the practitioner a way to highlight the crucial traits responsible for the classification decision. A validation survey with domain experts, beginners, and unskilled people shows that the use of explanations improves trust and confidence in the automatic decision system. Also, an analysis of the latent space adopted by the explainer unveils that some of the most frequent skin lesion classes are distinctly separated. This phenomenon may stem from the intrinsic characteristics of each class and may help resolve common misclassifications made by human experts.

Advancing Dermatological Diagnostics: Interpretable AI for Enhanced Skin Lesion Classification

“…The rapid ascent in ABELE's insertion curve signifies that its saliency map more accurately identifies the image segments most critical to the classifier's decisionmaking process. ABELE insertion curve grows much earlier with respect to LIME and LORE (from [17]).…”

“…This paper aims to build upon and refine the methodologies discussed in [3,[15][16][17], exploring the application of an explanation method in a genuine medical context, specifically for diagnosing skin lesions from images. Utilizing the labeled dataset from the ISIC 2019 (International Skin Imaging Collaboration) challenge (https://challenge.isic-archive.…”

“…Saliency maps for LIME (left), LORE (center), and ABELE (right). LIME and LORE highlight the macro-regions of the image that contribute positively (green) or negatively (red) to the prediction while ABELE provides a more fine-grained level of information with a divergent color scale, from relevant areas (dark orange) to marginally significant areas (green/cyan) (from[17]).…”

Towards Transparent Healthcare: Advancing Local Explanation Methods in Explainable Artificial Intelligence

“…However, the effectiveness of these models in clinical settings hinges on their interpretability and the transparency of their decision making, ensuring healthcare professionals can integrate AI insights confidently into patient care. In [42][43][44][45], a case study on skin lesion diagnosis using a ResNet classifier trained on the ISIC (International Skin Imaging Collaboration) dataset is presented. The classifier's decisions are explained using ABELE.…”