[Sportschosun Reporter Jang Jong-ho] A domestic research team has developed an AI-assisted diagnostic technology that can help detect nail melanoma at an early stage.
A research team led by Professor Byung Ho Oh of the Department of Dermatology at Severance Hospital, and Dr. Chu Yoo-seong of the Department of Precision Medicine at Wonju College of Medicine, developed a deep learning-based AI diagnostic support system that distinguishes nail melanoma from benign melanonychia. The team said on the 11th that the system showed diagnostic accuracy of more than 95%.
Melanonychia, which appears as a black line on the nail, is relatively common. However, in some cases, it can be an early sign of nail melanoma, a deadly skin cancer. Early nail melanoma can look very similar to benign lesions, making it difficult even for experienced dermatologists to tell them apart with the naked eye.
For an accurate diagnosis, a biopsy of the nail matrix is necessary. However, the procedure can leave permanent nail deformity or cosmetic and functional side effects, which makes it difficult to perform. For this reason, clinicians often delay biopsy and monitor the condition instead, limiting the possibility of early diagnosis in real-world practice.
Using clinical data on nail melanoma and benign melanonychia, the team developed a deep learning-based AI diagnostic support model. It collected clinical records from 294 patients, including 122 with nail melanoma and 172 with benign melanonychia. To reduce diagnostic confusion, pediatric melanonychia was excluded from AI training. The researchers also separated the data by patient so that photos of the same patient were not used for both training and evaluation, improving the model's reliability.
The AI model achieved high accuracy of more than 95% (AUROC 0.954) in distinguishing nail melanoma from benign melanonychia. It also maintained strong performance in external validation using patient data from another medical institution.
The research team also examined whether AI could help with diagnosis in actual clinical settings.
When dermatology specialists and residents diagnosed the same cases without AI support and then re-diagnosed them after reviewing the AI results, overall diagnostic accuracy improved from 70% to 80.8%. The largest improvement was seen among dermatology residents, and agreement among medical staff also improved.
Professor Byung Ho Oh said, "This study is significant because it suggests the potential of a diagnostic support technology that can be used in real clinical practice." He added, "We hope it will contribute to improving patient survival and reducing unnecessary biopsies through the early diagnosis of nail melanoma." The findings were published in the latest issue of the international journal Journal of the German Society of Dermatology (JDDG).
Reporter Jang Jong-ho bellho@sportschosun.com
This article has been translated by GripLabs Mingo AI.