Assessment for pigmentation and vascularity of hypertrophic scar based on residual network

Hypertrophic scars are a type of pathological scar that can cause discomfort, pain, and itching. The assessment of hypertrophic scars depends largely on the clinician’s long-term experience. Computer-aided assessment can greatly improve the efficiency and accuracy of scar assessment. In this paper, we propose a deep neural network-based assessment method for the degree of scar pigmentation and vascularity using real scar images. Firstly, a large amount of hypertrophic scar images were collected to produce a dataset including patients of all ages and the distribution of lesions in different parts of the body. These images were then scored on pigmentation and vascularity using the Vancouver Scar Scale (VSS). Then, the residual network model was proposed to evaluate the degree of scar pigmentation and vascularity according to the hypertrophic scar dataset. The degree of the pigmentation and vascularity were assessed respectively by two residual network models. The experimental results demonstrated that our method provides a reliable tool for the application of a scar diagnosis system.