New data suggest that preprocessed images improve efficacy, precision, and comprehensibility of machine learning (ML) models in image classification tasks of skin cancer. Tarek Khater and colleagues used classical ML algorithms to develop a machine learning-based model capable of accurately classifying skin cancer by utilizing extracted features from preprocessed images in the publicly available PH2 dataset. The results of the study show that the proposed model enhances the accuracy of identifying information in images, which improves ML classification performance and interpretability. In simulation results, utilizing XG-boost achieved a 94% accuracy, effectively distinguishing be- tween nonmelanoma and melanoma skin cancer. Leveraging model-agnostic methods such as partial dependence plot, permutation importance, and SHAP, explainable AI provides insights for providers. According to the researchers, the results highlight asymmetry and pigment network features as the most critical factors in skin cancer classification, under- scoring their significance in distinguishing be- tween different skin cancer types.
