Neuro-Fuzzy Approach to the Segmentation of Psoriasis Images

In this paper, an automatic method for psoriasis image segmentation using neuro-fuzzy techniques is proposed. It can be used in a therapy evaluation system. Since the psoriasis is a chronic disease, it is important to track the condition of the patient to select a proper treatment. In our design, the psoriasis images are segmented into normal skin regions and abnormal regions automatically. The areas of each kind of regions of a patient at different points of time can then be estimated. This information can be used to give a quantitative measure of the progress of the treatment. The provided information can avoid the variation of the human factor in the evaluation procedure and can offer a objective index for the doctor to select the most suitable treatment for the patient. The essential techniques required include feature extraction and image segmentation (classification) methods. The two-dimensional histogram of the hue and saturation components of the color image and the fuzzy texture spectrum of the gray-level image are used as the feature vectors to locate the homogeneous regions. Then these regions are used to train the neuro-fuzzy classifier to obtain a more accurate segmentation. After the image is segmented into normal and psoriasis regions, the area of psoriasis regions can be easily calculated to obtain the information required for the therapy evaluation system. For comparison, a color clustering algorithm which was used to segment digitized dermatoscopic images is also implemented. In the experiments, the proposed approach provides better performance.