Quantification of Diabetic Foot Ulcer Based on Ulcer Segmentation

The Diabetic Foot (DF) is a common and severe complication of diabetes. Many people suffer from diabetic foot ulcers (DFU) caused by diabetic feet. The quantification of DFU in clinical practice is of great help in helping doctors diagnose and formulate treatment plans.This paper proposes two DFU quantification methods based on deep learning, rough quantification, and fine quantification. Rough quantification is based on a quantitative indicator. We use an object detection network to locate the indicator. Fine quantification takes surrounding skin tissue(SST) as an indicator. We use a segmentation network to segment DFU and SST to quantity ulcer healing. Our method is evaluated upon our dataset collected by Shanghai Municipal Eighth People’s Hospital and the FUSC2021 dataset. We simultaneously followed multiple patients in clinical practice to assess their DFU healing by our quantification method. We evaluate the generalization ability of our rough quantification segmentation model, achieving Dice of 90.256%, IoU of 82.243%, and rough quantification object detection model achieving Recall of 65.00%, Precision of 92.86%, Fl-score of 76.00%, and AP 85.56%. Also, our fine quantification segmentation model performs Dice in the SST area of 85.46%, IoU in the SST area of 74.61%, and Dice in the Ulcer area 87.31%, IoU in the Ulcer area of 77.48%, and inference time of 0.09s. We selected several cases in clinical practice and applied our two quantification methods to the recovery of DFU in patients with satisfactory results.