The Classification of Diabetic Foot Based on Faster

In order to realize the clinical automatic identification of the Wagner level of diabetic foot and solve the clinical problem that a large number of professional doctors need to spend time to detect and identify the patient’s foot disease level, this paper proposes a Faster R-CNN based diabetic foot Wagner level classification detection model. The image data of the sick feet we collected comes from the top hospitals and the network and the professional doctor team carries out the data set annotation and inspection, and the data validity and scientificity are guaranteed. Our model uses a convolutional neural network which has the advantages of automatic feature extraction, strong generalization ability, and high recognition accuracy. The detection model is based on two model schemes of resnet101 network and VGG16 network respectively. Experiments show that the precision of the Faster R-CNN model based on resnet101 is higher than that of VGG16. In view of the uneven distribution of the number of foot pictures in some grades, we use the method of changing the contrast of the picture, adding random Gaussian noise, and flipping to enhance the data. The final result shows that the accuracy of our improved method is increased to 95.24% on the original basis. It’s good to provide effective solutions for the current clinical assisted primary screening of diabetic foot severity.