AI-Driven Prognostic Models in Melanoma

Melanoma care has progressed substantially with improved staging frameworks and the adoption of targeted therapies and immune checkpoint inhibitors, yet outcomes remain highly variable, even among patients with similar clinicopathological profiles. Prognostic uncertainty is especially consequential when decisions depend on identifying who is most likely to relapse, progress, or develop metastatic disease. At the same time, the growing volume and complexity of melanoma-relevant data has outpaced what traditional prognostic approaches can consistently integrate, creating a clear need for tools that enable more individualized risk stratification.AI and machine learning methods are increasingly proposed to address this gap, but many models remain difficult to interpret, under-validated, or developed on narrow cohorts that limit generalizability. Heterogeneity in endpoints, cohort definitions, follow-up, and reporting practices also makes it challenging to compare findings across studies or judge readiness for real-world use. This Research Topic therefore seeks to bring together rigorous, clinically relevant work that moves beyond proof-of-concept performance toward transparent, robust, and actionable AI-driven prognostic modelling in melanoma.Within this scope, relevant themes include (but are not limited to):• AI/ML prognostic modelling for survival, recurrence, metastasis risk, and clinically actionable risk stratification.• External validation strategies (multi-center, temporal, geographic validation) and reproducibility best practices.• Model calibration, uncertainty quantification, and clinical utility assessment (e.g., decision-curve approaches).• Multimodal prognostic models integrating clinical data with digital pathology, radiology/radiomics, and/or omics.• Explainability and interpretability approaches that support clinician trust and biological plausibility.• Fairness, bias assessment, and subgroup performance reporting across clinically relevant populations.• Prognostic modelling in the modern treatment era (immunotherapy/targeted therapy) and non-stationarity over time.• Systematic reviews, meta-analyses, and critical appraisals of AI prognostic model quality and reporting standards.• Translational studies evaluating pathways to implementation, including workflow integration and prospective evaluation plans.We welcome Original Research, Systematic Reviews, Meta-analyses, Methods articles, Brief Research Reports, and Perspectives that offer practical guidance for improving quality, validation, and adoption of melanoma prognostic AI. By bringing together robust evidence and thoughtful methodology, this collection aims to clarify what is ready for translation and what remains to be solved, helping the field move toward dependable prognostic tools that can genuinely support patient care.Please note: manuscripts consisting solely of bioinformatics, computational analysis, or predictions of public databases which are not accompanied by validation (independent clinical or patient cohort, or biological validation in vitro or in vivo, which are not based on public databases) are not suitable for publication in this journal. Topic editors: Diego I. Gallardo, Elena Codruta Cozma, Victor Desmond Mandel Source: https://www.frontiersin.org/research-topics/79880/ai-driven-prognostic-models-in-melanoma