Photodiagnosis and photodynamic therapy · 2025
Ultraviolet-induced fluorescence mapping of facial porphyrin and sebum using deep-learning segmentation.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Jung Geunho, Yoo Sangwook
Аннотация
BACKGROUND: Porphyrins and sebum, which fluoresce under ultraviolet (UV) light, are important features closely linked to acne development. However, traditional detection methods are often unreliable as they are sensitive to variations in lighting conditions and skin tone.
OBJECTIVES: While deep learning-based segmentation has shown promise, comparative studies for this specific application remain limited. This study aims to address this gap by evaluating and comparing two distinct deep learning approaches for the automated segmentation of UV-induced fluorescence.
METHODS: We evaluated two deep learning approaches, a conventional segmentation network (UNet) and a generative adversarial network (pix2pix), using 294 facial fluorescence images from 49 subjects. The models were compared in both single-class and two-class segmentation settings to comprehensively evaluate their performance.
RESULTS: UNet demonstrated more stable performance than pix2pix, especially in cases with weak or absent fluorescence, while pix2pix often produced false positives in non-fluorescent regions. Quantitative analysis using multiple metrics, including Intersection over Union (IoU) and Dice coefficient, confirmed UNet's superior performance over pix2pix in both single- and two-class tasks. In the two-class setting, UNet achieved an IoU of 0.2106 for porphyrins and 0.1024 for sebum.
CONCLUSION: This study demonstrates the feasibility of using deep learning to detect porphyrins and sebum in facial fluorescence images, with UNet showing more robust performance than pix2pix. While the single-class model's overall IoU score was slightly higher, the two-class approach offered enhanced interpretability. Future work should expand dataset diversity and validate the models with clinical-grade imaging systems.
OBJECTIVES: While deep learning-based segmentation has shown promise, comparative studies for this specific application remain limited. This study aims to address this gap by evaluating and comparing two distinct deep learning approaches for the automated segmentation of UV-induced fluorescence.
METHODS: We evaluated two deep learning approaches, a conventional segmentation network (UNet) and a generative adversarial network (pix2pix), using 294 facial fluorescence images from 49 subjects. The models were compared in both single-class and two-class segmentation settings to comprehensively evaluate their performance.
RESULTS: UNet demonstrated more stable performance than pix2pix, especially in cases with weak or absent fluorescence, while pix2pix often produced false positives in non-fluorescent regions. Quantitative analysis using multiple metrics, including Intersection over Union (IoU) and Dice coefficient, confirmed UNet's superior performance over pix2pix in both single- and two-class tasks. In the two-class setting, UNet achieved an IoU of 0.2106 for porphyrins and 0.1024 for sebum.
CONCLUSION: This study demonstrates the feasibility of using deep learning to detect porphyrins and sebum in facial fluorescence images, with UNet showing more robust performance than pix2pix. While the single-class model's overall IoU score was slightly higher, the two-class approach offered enhanced interpretability. Future work should expand dataset diversity and validate the models with clinical-grade imaging systems.