Skin lesion evaluation from multispectral images

During evaluation of skin disease treatments, dermatologists have to clinically measure the pathology severity of each patient during treatment periods. Such a process is sensitive to intra- and inter- dermatologist diagnosis. To make this severity measurement more robust we propose to use image processing to quantify the pathology severity. We focus on a hyperpigmentation disorder called melasma. During a treatment period, multispectral images are taken on patients receiving the same treatment. After co-registration and classification steps, we propose algorithms to measure the intensity, the size and the homogeneity evolution of the pathological area. Obtained results are compared with a dermatologist diagnosis using statistical tests on a full clinical study.