Seoul St. Mary's Hospital Develops AI System Suggesting Optimal Biopsy Sites for Melanoma

Professor Han Juhee's Dermatology Research Team
Up to 98% Accuracy... Aiding Early Diagnosis

Professor Han Juhee (left) and resident Park Jiho of the Department of Dermatology at Seoul St. Mary's Hospital.

[Asia Economy Reporter Lee Gwan-joo] A domestic research team has developed an AI-based biopsy site recommendation system to assist in the early diagnosis of melanoma, a highly fatal skin cancer.

Professor Han Joo-hee and resident Park Ji-ho from the Department of Dermatology at The Catholic University of Korea Seoul St. Mary's Hospital announced on the 5th that they designed a model suggesting punch biopsy sites based on images generated by a GAN (Generative Adversarial Network), applying a non-invasive, augmented approach compared to the conventional biopsy method commonly used for melanoma diagnosis.

The research team trained machine learning classifiers by combining dermoscopic images of melanoma and benign nevi from Seoul St. Mary's Hospital with public data (HAM10000 melanoma dataset). As a result, the classifier achieved an accuracy of 91.05%, sensitivity of 49.18%, specificity of 98.16%, and an F1 score (harmonic mean of precision and recall) of 65.53%.

Furthermore, when comparing the biopsy site recommendations of dermatology specialists with those suggested by the AI model, the agreement ranged from 58% to as high as 98%, depending on the applied model.

Professor Han stated, "If the biopsy site for skin cancer is inaccurate, diagnosis can be delayed. Unlike other relatively slow-progressing skin cancers, melanoma can be fatal if treatment timing is missed, as it metastasizes to other organs such as lymph nodes, brain, and bones. If the AI model is further improved through additional research, it could assist decision-making to enable early and accurate diagnosis of melanoma, ultimately improving its prognosis."

This study was published in the August issue of the international journal Journal of the European Academy of Dermatology and Venereology.

This content was produced with the assistance of AI translation services.