Predicting psoriasis severity using machine learning: A systematic review.

Abstract

In dermatology, the applications of machine learning (ML), an artificial intelligence (AI) subset that enables machines to learn from experience, have progressed past the diagnosis and classification of skin lesions. A lack of systematic reviews exists to explore the role of ML in predicting the severity of psoriasis. This systematic review aims to identify and summarize the existing literature on predicting psoriasis severity using ML algorithms and identify gaps in current clinical applications of these tools. OVID Embase, OVID MEDLINE, ACM Digital Library, Scopus, and IEEE Xplore were searched from inception to August, 2024. A total of 30 articles met our inclusion criteria and were included in this review. One article used serum biomarkers, while the remaining 29 used image-based models. The most common severity assessment score employed by these ML models was the Psoriasis Area Severity Index score, followed by Body Surface Area, with fifteen and five articles, respectively. The small size and heterogeneity of the existing literature are the primary limitations of this review. Progress in assessing skin lesion severity through ML in dermatology has advanced, but prospective clinical applications remain limited. ML and AI promise to improve psoriasis management, especially in non-image-based applications requiring further exploration. Large-scale prospective trials using diverse image datasets are necessary to evaluate and predict the clinical value of these predictive AI models.