ExplainabilityTransparencyTrustworthinesscs.CV
Explainable Deep Image Classifiers for Skin Lesion Diagnosis
Carlo Metta, Andrea Beretta, Riccardo Guidotti, Yuan Yin, Patrick Gallinari, Salvatore Rinzivillo, Fosca Giannotti
November 22, 2021
arXiv.org
7 citations
Key Takeaways
-
Explainability builds trust in medical AI: A survey with domain experts, beginners, and untrained people demonstrated that providing explanations for AI classification decisions increased trust and confidence in the system's recommendations for skin lesion diagnosis.
-
Visual explanations can make medical AI more interpretable: The researchers developed an approach that generates synthetic example and counter-example images of skin lesions to show practitioners which visual features the AI model used to make its diagnosis, making the "black box" decision-making process more transparent.
-
XAI methods need real-world testing: The paper emphasizes that explainability research is often tested only on general classifiers, but this study demonstrates the importance of customizing XAI approaches for realistic medical diagnosis problems where interpretability is critical for clinical adoption.
-
AI explanations may reveal clinically useful patterns: Analysis of the model's internal representations showed that common skin lesion classes were distinctly separated in the AI's "latent space," suggesting the model learned meaningful distinctions that could potentially help resolve frequent misclassifications made by human experts.
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 analyze a case study on skin lesion images where we customize an existing XAI approach for explaining a deep learning model able to recognize different types of skin lesions. The explanation is formed by synthetic exemplar and counter-exemplar images of skin lesion and offers the practitioner a way to highlight the crucial traits responsible for the classification decision. A survey conducted with domain experts, beginners and unskilled people proof that the usage of explanations increases the 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 could derive from the intrinsic characteristics of each class and, hopefully, can provide support in the resolution of the most frequent misclassifications by human experts.