An Automatic Method of Chronic Wounds Segmentation in Multimodal Images

Chronic wounds are common diseases in aging society. Automatic method of images segmentation is required to effectively and objectively monitor the healing process. The segmentation method proposed in the paper employs Histograms of Oriented Gradients, Weighted Fuzzy C-Means Clustering, Edge Detection, Gradient Vector Flow and Active Contour techniques. The method gives high compliance with manual outlines performed by two experts. Mean Dice Index for 11 cases was 0.84. Obtained results indicate the possibility of automation of diagnosis and monitoring processes. An infrared image reveals the parts of the wound under the skin which are invisible for commonly used cameras and it might give valuable information for physicians in assortment of treatment.