A Deep Learning-Based Segmentation Strategy for Diabetic Foot Ulcers: Combining the Strengths of HarDNet-MSEG and SAM Models

Abstract

The primary objective of this study is to investigate and propose a reliable and accurate method for segmenting Diabetic Foot Ulcers (DFU) wounds. DFU is a prevalent complication among diabetic patients that can have severe consequences if not promptly addressed. However, the segmentation of DFU wounds poses a complex challenge due to variations in symptom color, size, and contrast, which can vary depending on the severity of the condition. Furthermore, challenges such as image noise, lighting and contrast variations, and labeling difficulties further complicate the taskTaking advantage of the rapid advancements in deep learning and its application to image segmentation, this study introduces a robust DFU segmentation model based on deep learning techniques. The proposed model aims to achieve accurate and precise segmentation of DFU wounds, addressing the aforementioned challenges..To assess the effectiveness of our segmentation strategy, we evaluated its performance using the public database of the 2022 DFU Segmentation Challenge. The results obtained demonstrate that our model achieves an average Dice coefficient of 83.44%, a substantial improvement compared to the average Dice coefficient of 72.87% achieved by other participants. These results serve as compelling evidence that our segmentation method successfully achieves high-precision segmentation of DFU wounds.