XAI for Psoriasis and Psoriatic Arthritis: A Systematic Survey of Diagnostic, Prognostic, and Therapeutic Applications

Resumo

Explainable Artificial Intelligence (XAI) has become an essential pillar of trustworthy and transparent decision support systems in managing medically complex autoimmune diseases, such as psoriasis and psoriatic arthritis. While Artificial Intelligence (AI) and Machine Learning methods have reached high accuracy levels in diagnosing psoriasis, determining its severity, identifying biomarkers, and predicting treatment outcomes, their black-box nature has prevented them from entering everyday clinical practices on a large scale. This review aims at making a thorough and well-organized examination of the most recent progress that has been made using XAI, based approaches in psoriasis and other psoriatic disorders. We have performed a comprehensive literature search through the top-tier research databases such as Scopus, PubMed, IEEE Xplore, and Web of Science. The search was limited to only peer-reviewed articles published in the 2018, 2025 period. The papers are distributed across the four main clinical application fields, namely diagnostic imaging, disease severity quantification, biomarker discovery, and treatment response prediction. We also look in detail at the functionalities of the different explainability techniques that have been used in the studies, e.g., Gradient, weighted Class Activation Mapping (GradCAM), SHapley Additive exPlanations (SHAP), Local Interpretable Model, agnostic Explanations (LIME), attention mechanisms, and rule, based models, thereby underlining how they assist in removing the opacity surrounding AI models and thus gaining clinicians' confidence. Our survey has revealed that some of the major challenges standing in the way of successful real, life applications are data diversification, scarce external validation, bias, and absence of universal explainability criteria.