Exemplars and Counterexemplars Explanations for Image Classifiers, Targeting Skin Lesion Labeling

TL;DR

This paper presents a framework for explainable AI in skin lesion diagnosis, providing practitioners with exemplars and counterexemplars generated by an autoencoder to interpret deep neural network decisions.

Contribution

It introduces a novel explanation module that offers exemplars and counterexemplars for medical image classification, enhancing human-AI interaction in sensitive domains.

Findings

Effective generation of exemplars and counterexemplars for skin lesion classification

Improved interpretability of deep neural network decisions in medical diagnosis

Potential to assist practitioners in understanding AI-based diagnoses

Abstract

Explainable AI consists in developing mechanisms allowing for an interaction between decision systems and humans by making the decisions of the formers understandable. This is particularly important in sensitive contexts like in the medical domain. We propose a use case study, for skin lesion diagnosis, illustrating how it is possible to provide the practitioner with explanations on the decisions of a state of the art deep neural network classifier trained to characterize skin lesions from examples. Our framework consists of a trained classifier onto which an explanation module operates. The latter is able to offer the practitioner exemplars and counterexemplars for the classification diagnosis thus allowing the physician to interact with the automatic diagnosis system. The exemplars are generated via an adversarial autoencoder. We illustrate the behavior of the system on representative examples.