Journal of wound care · 2019
Convolutional neural networks for wound detection: the role of artificial intelligence in wound care.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
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.
Ohura Norihiko, Mitsuno Ryota, Sakisaka Masanobu, Terabe Yuta, Morishige Yuki, Uchiyama Atsushi, Okoshi Takumi, Shinji Iizaka, Takushima Akihiko
Аннотация
OBJECTIVE: Telemedicine is an essential support system for clinical settings outside the hospital. Recently, the importance of the model for assessment of telemedicine (MAST) has been emphasised. The development of an eHealth-supported wound assessment system using artificial intelligence is awaited. This study explored whether or not wound segmentation of a diabetic foot ulcer (DFU) and a venous leg ulcer (VLU) by a convolutional neural network (CNN) was possible after being educated using sacral pressure ulcer (PU) data sets, and which CNN architecture was superior at segmentation.
METHODS: CNNs with different algorithms and architectures were prepared. The four architectures were SegNet, LinkNet, U-Net and U-Net with the VGG16 Encoder Pre-Trained on ImageNet (Unet_VGG16). Each CNN learned the supervised data of sacral pressure ulcers (PUs).
RESULTS: Among the four architectures, the best results were obtained with U-Net. U-Net demonstrated the second-highest accuracy in terms of the area under the curve (0.997) and a high specificity (0.943) and sensitivity (0.993), with the highest values obtained with Unet_VGG16. U-Net was also considered to be the most practical architecture and superior to the others in that the segmentation speed was faster than that of Unet_VGG16.
CONCLUSION: The U-Net CNN constructed using appropriately supervised data was capable of segmentation with high accuracy. These findings suggest that eHealth wound assessment using CNNs will be of practical use in the future.
METHODS: CNNs with different algorithms and architectures were prepared. The four architectures were SegNet, LinkNet, U-Net and U-Net with the VGG16 Encoder Pre-Trained on ImageNet (Unet_VGG16). Each CNN learned the supervised data of sacral pressure ulcers (PUs).
RESULTS: Among the four architectures, the best results were obtained with U-Net. U-Net demonstrated the second-highest accuracy in terms of the area under the curve (0.997) and a high specificity (0.943) and sensitivity (0.993), with the highest values obtained with Unet_VGG16. U-Net was also considered to be the most practical architecture and superior to the others in that the segmentation speed was faster than that of Unet_VGG16.
CONCLUSION: The U-Net CNN constructed using appropriately supervised data was capable of segmentation with high accuracy. These findings suggest that eHealth wound assessment using CNNs will be of practical use in the future.