AI-Powered System Enhances Diagnosis of Facial Pigmented Lesions

A groundbreaking diagnostic system utilizing artificial intelligence (AI) has been developed to accurately classify various facial pigmented lesions, aiding in precise laser treatment planning. This innovative approach employs deep learning models, specifically InceptionResNetV2 and DenseNet121, to differentiate among five common lesion types: melasma, ephelides, acquired dermal melanocytosis, solar lentigo, and lentigo maligna/malignant melanoma.

The system was trained and validated using 432 clinical images, with its diagnostic performance compared to nine expert dermatologists and eleven non-expert dermatologists. Results showed that the AI models achieved diagnostic accuracies of approximately 87% and 86%, respectively, surpassing the median accuracy of human experts (80%) and non-experts (63%). Notably, both models demonstrated 100% sensitivity in identifying lentigo maligna and its melanoma variant, highlighting their potential as effective diagnostic support tools.

This advancement is particularly significant given the challenge of visually similar facial pigmented lesions, which complicate differential diagnosis. Accurate classification is critical because treatment strategies vary considerably among these conditions. For example, inappropriate laser therapy can worsen melasma, while delayed diagnosis of malignant lesions could lead to severe health consequences.

The research, led by Haruyo Yamamoto, Chisa Nakashima, and Atsushi Otsuka from Kindai University Faculty of Medicine's Department of Dermatology, aims to improve diagnostic precision and facilitate better treatment decisions. This AI-based system promises to support dermatologists by providing reliable, quick assessments, ultimately enhancing patient care in dermatology.

Source: https://medicalxpress.com/news/2025-07-ai-based-classification-facial-pigmented.html