Color detection in dermoscopic images of pigmented skin lesions through computer vision techniques

This thesis offers an insight into skin cancer detection, focusing on the extraction of distinct features (color, namely) from potential melanoma lesions. The following document provides an outlook of melanoma analysis, as well as experimental results based on Matlab implementations. The relevance of the work carried out throughout this project resides in the specificity of the study: color is a key characteristic in melanoma inspection. It is usually linked to pattern analysis but seldom the sole object of research. Most lines of work in the field of skin cancer diagnosis associate color with other features such as texture, shape, asymmetry or pattern of the lesion. Studies cement this belief regarding the vital significance of color, as the number of colors in a lesion happens to be the most significant biomarker for determining malignancy. Different image processing techniques will be applied to build statistical models that shape the outcome of the prospective diagnosis. The purpose of the project is the development of an assisting tool able to detect the most prevalent colors in skin pigmented lesions, in order to give a probabilistic result. The strength of this idea lies in the resemblance to actual medical procedures; dermatologists examine color to diagnose melanoma. Simulating medical proceedings is a burgeoning trend in CAD systems because it renders the advancements in this field more likely to be accepted by the medical community. An additional motivation comes from real-life statistics: skin cancer is, by far, the most frequent type of cancer. Moreover, although melanoma is the least common form of skin cancer at only around 1% of all cases, the majority of deaths related to skin cancer are due to melanoma. Furthermore, the rate of melanoma occurrence is particularly high in Spain and has significantly increased in the last decade, hence the importance of reliable diagnosis that is not exclusively contingent on the specialist’s subjective judgment.Ingeniería de Sistemas Audiovisuale