Wound Care: Wound Segmentation and Parameter Estimation

Wound progression monitoring is critical, as it involves repeated clinical examination over days; chronic wounds take around 3-6 months. In many cases, the wound healing period is unpredictable, due to several factors, including etiology, age, nutrition, patient comorbidity, medication, and environment, influence the healing cycle. Our system aims to automatize wound segmentation and quantification, that aids telehealth for wound care professionals. Here the AZH and Foot ulcer challenge datasets are merged, which summed up 1841 images. Further, augmenting the dataset led to 51,362 images: one of the large datasets till-date. We adopted the DeepLabV3+, with Resnet-50 encoder, resulting in a dice score–0.919, Mean IOU–0.920. Adding Squeeze and Excite with DeepLabV3+ resulted in a dice score of 0.923 and IOU-0.924.Wound quantification was performed by estimating the wound metrics such as area, perimeter, diameter, and major and minor axis length of an ellipse using morphological operations, connected components, and shape analysis.