Infrared Thermography and Machine Learning for Skin Cancer Screening: A Benchmarking and Deployment Study

Abstract

Early detection of malignant skin lesions remains a critical challenge in dermatology, where diagnostic accuracy depends on visual inspection, dermoscopy, and invasive biopsy procedures. Infrared thermography offers a non-contact and radiation-free modality capable of capturing physiological heat patterns associated with abnormal tissue metabolism and vascular activity. When combined with machine learning, thermographic data enables automated screening pipelines that can assist clinicians in identifying suspicious lesions at scale. This study presents a comprehensive benchmarking and deployment-oriented evaluation of infrared thermography based skin cancer screening systems. Multiple machine learning strategies are assessed under realistic acquisition conditions, with emphasis on robustness, explainability, and operational feasibility. The work advances practical understanding of thermographic decision support systems and outlines pathways for safe clinical integration.