Detection of Buruli ulcer disease: Preliminary results with dermoscopic images on smart handheld devices

Buruli ulcer (BU) disease is a devastating flesh-eating bacterial infection that each year affects thousands of people, but if detected early, it can be easily treated and cured. During the past several years, we have been developing an automated dermoscopy system for accurate analysis of skin lesions. In this paper, we describe our experience with BU detection in West Africa, and we focus on the classification and validation stages of the system. We analyzed a set of 58 skin lesions consisting of 16 confirmed BU and 42 non-BU cases. After segmentation, we extracted texture and color descriptors from all images and studied the accuracy of BU detection, using a bag-of-feature classification procedure, as well as the influence of different sampling strategies, patch size, codebook size, and kernel type. Our results show an overall 95.2% classification accuracy. These findings suggest that smart phones can be used as assistive diagnostic devices for routine skin screening in underserved areas, in general, and in developing countries, in particular, where healthcare infrastructure is limited.