MEL-IA AI System Integrates Into Hospitals to Classify Skin Lesions

Researchers at the University of Alicante and Sant Joan d'Alacant University Hospital have successfully deployed MEL-IA, an artificial intelligence system engineered to automate the classification of skin lesions directly within clinical workflows. Published in the Journal of Medical Systems, the technology addresses the growing global burden of skin cancer, which accounted for approximately 1.5 million new cases worldwide in 2024. By embedding diagnostic support into existing hospital infrastructure, MEL-IA aims to accelerate early detection and improve patient outcomes. Developed by the university's Bio-inspired Engineering and Health Informatics research group, the system operates through a unified digital pipeline. Clinicians use a dedicated mobile application to capture dermatoscopic imagery and log patient metadata, including age, sex, and anatomical location. The application transmits this information to a centralized AI model that performs multi-class classification across five primary lesion categories: melanoma, nevus, basal cell carcinoma, actinic keratosis, and benign keratosis. Unlike binary screening tools that only differentiate between malignant and benign growths, MEL-IA provides granular differential diagnosis to assist dermatologists in refining treatment pathways. Model training and validation relied on a curated dataset of more than 15,000 dermatoscopic images paired with clinical records. The resulting architecture delivers an overall classification accuracy of 86 percent. Notably, the system achieved 88 percent sensitivity for melanoma detection and 92 percent sensitivity for basal cell carcinoma, with peak performance recorded for nevi and basal cell carcinomas. During pilot operations at the Sant Joan d'Alacant University Hospital, the platform processed 980 dermatological evaluations with inference times consistently under one second. The system also maintains longitudinal patient records, aggregating sequential imaging, diagnostic outputs, and clinical notes to support long-term monitoring. MEL-IA is explicitly positioned as a clinical decision-support tool rather than an autonomous diagnostic agent. Its architecture prioritizes secure data exchange, compliant storage, and seamless interoperability with institutional health-information systems. The development team, comprising computer science faculty and hospital IT specialists, has designed the platform to augment rather than replace clinician judgment. Looking ahead, researchers plan to conduct prospective clinical trials to evaluate real-world efficacy alongside medical professionals. Additional development milestones include enabling direct smartphone-based image acquisition and expanding the classification framework to encompass a broader spectrum of dermatological conditions. The successful integration of MEL-IA into routine hospital operations marks a significant step toward scalable, AI-assisted dermatological care.