A new deep learning system developed by an international research team detects melanoma with 94.5% accuracy by fusing dermoscopy images with metadata such as patient age, gender, and lesion location. This approach improves diagnostic accuracy, transparency, and access to early detection of skin cancer through smart healthcare technology. Credit: Professor Gwangill Jeon / Incheon National University, South Korea
Melanoma remains one of the most difficult skin cancers to diagnose because it closely resembles a harmless mole or lesion. Most artificial intelligence (AI) tools rely solely on dermoscopy images, but they often overlook important patient information (such as age, gender, and where the lesion is located on the body) that can improve diagnostic accuracy. This highlights the importance of multimodal fusion models that enable highly accurate diagnosis.
To fill this gap, Professor Gwang-gil Chung from the Department of Embedded Systems Engineering at Incheon National University in South Korea collaborated with the University of the West of England (UK), Anglia Ruskin University (UK), and the Royal Military College of Canada to create a deep learning model that integrates patient data and dermoscopy images.
The research will be published in a journal information fusion.
“Skin cancer, and melanoma in particular, is a disease where early detection is critical to determining survival rates,” says Professor Jeon.
“Melanoma is difficult to diagnose based on visual features alone, so we recognized the need for an AI convergence technology that could consider both image data and patient information.”
How was the AI model developed?
The team used the large-scale SIIM-ISIC melanoma dataset, which includes more than 33,000 dermoscopy images combined with clinical metadata, to train an AI model to recognize the subtle relationships between what appears on the skin and who the patient is. The model achieved an accuracy of 94.5% and an F1 score of 0.94, outperforming popular image-only models such as ResNet-50 and EfficientNet.
The researchers also performed feature importance analysis to increase the transparency and robustness of the system. Factors such as lesion size, patient age, and anatomical location were found to significantly contribute to accurate detection. These insights can help provide a roadmap for physicians to understand and trust the diagnoses performed by AI.
Potential impact on melanoma screening
Professor Jeon said: “This model was not designed solely for academic purposes; it could be used as a practical tool to transform melanoma screening in the real world. This research has direct application to the development of AI systems that analyze both skin lesion images and basic patient information to enable early detection of melanoma.”
In the future, this model could help power smartphone-based skin diagnostic applications, telemedicine systems, or AI-assisted tools in dermatology clinics to reduce misdiagnosis rates and improve access to treatment.
Professor Jeon explains, “This research represents a step forward toward personalized diagnosis and preventive medicine using AI convergence technology.”
This study highlights how multimodal AI can bridge the gap between machine learning and clinical decision-making, paving the way for more accurate, accessible, and reliable skin cancer diagnosis.
Detailed information:
Misbah Ahmad et al., Fusion of metadata and dermoscopy images for melanoma detection: deep learning and feature importance analysis, information fusion (2025). DOI: 10.1016/j.inffus.2025.103304
Provided by Incheon National University
quotation: Deep learning systems can transform skin cancer detection with near-perfect accuracy (November 14, 2025) Retrieved November 16, 2025 from https://medicalxpress.com/news/2025-11-deep-skin-cancer-accuracy.html
This document is subject to copyright. No part may be reproduced without written permission, except in fair dealing for personal study or research purposes. Content is provided for informational purposes only.
