AUTHOR=Zhang Jianglin , Qiu Yue , Peng Li , Zhou Qiuhong , Wang Zheng , Qi Min 

TITLE=A comprehensive review of methods based on deep learning for diabetes-related foot ulcers

JOURNAL=Frontiers in Endocrinology

VOLUME=Volume 13 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2022.945020

DOI=10.3389/fendo.2022.945020

ISSN=1664-2392

ABSTRACT=Background: Diabetes mellitus (DM) is a chronic disease with hyperglycemia, if not treated in time,it may lead to lower limb amputation.At an initial stage,the detection of DFU is very difficult.  Deep learning has demonstrated state-of-the-art performance in various fields and been used to analyze the images of diabetes-related foot ulcers (DFUs).
Objectives: This article reviews the current applications of deep learning to early detection of DFU to avoid limb amputation or infection.
Methods: Relevant literatures on deep learning in analysis for images of DFU published during the past 10 years were analyzed, including deep learning models in the classification, object detection and semantic segmentation for images of DFUs.
Results: The primary DFU uses of deep learning currently are related to using different algorithms. For classification tasks, the improved classification models were all based on convolutional neural networks. The model with parallel convolutional layers based on GoogLeNet and the ensemble model outperform the other models in classification accuracy. For object detection tasks, the models were based on architecture such as Faster R-CNN, YOLOv3, YOLOv5 or EfficientDet. The refinements on YOLOv3 models achieved an accuracy of 91.95% and the model with an adapted Faster R-CNN architecture achieved a mean average precision of 91.4%, which outperformed the other models. For semantic segmentation tasks, the models were based on architecture such as fully convolutional networks, U-Net, V-Net or SegNet. The model with U-Net outperformed the other models in accuracy of 94.96%. Taking segmentation tasks as an example, the models were based on architecture such as Mask-RCNN. The model with Mask-RCNN obtained precision:0.8632 and mAP: 0.5084
Conclusions: Although promise is shown by current research in ability of deep learning to improve patients’ life, further research is required to better understand the mechanisms about deep learning for DFUs.