A computer-aided diagnosis system for vitiligo assessment: A segmentation algorithm

Abstract

Vitiligo is a condition where depigmentation occurs in parts of the skin. Although vitiligo affects only 0.5% to 1% of the world's population, vitiligo may impact patients' quality of life greatly. The success of vitiligo treatment is influenced by the accuracy of vitiligo assessment method used for monitoring the treatment progress. A popular method for assessing vitiligo is the Vitiligo Area Scoring Index (VASI), where VASI score is calculated based on the visually observed degree of depigmentation in the skin and the area of depigmented skin. This approach has two limitations. First, it has inter-examiner variability, and second, its accuracy can be very low. While the first limitation can be addressed through training, the second limitation is more difficult to address. With visual observation, positive but small progress may be unnoticed since it usually takes some time to notice skin color changes manually. To overcome this limitation, we propose in this paper a vitiligo lesion segmentation algorithm as a part of a computer-aided diagnosis system for vitiligo assessment. The algorithm reads color skin image as input and makes use of the Fuzzy C-Means clustering algorithm and YCbCr and RGB color spaces to separate pigmented and depigmented skin regions in the image. The corresponding degree of depigmentation is then estimated. The algorithm was evaluated on low resolution Internet images of skin with vitiligo. The results are encouraging, indicating that the proposed algorithm has a potential for use in clinical and teledermatology applications.