Abstract:

The World Health Organization determined the annual number of deaths caused by burn is approximately 265,000. This number clearly reveals the importance of burn wound diagnosis. Determining the burn/normal skin region is the one of the most important parameters which are needed to be determined in the planning of burn wound treatment. In this study, fuzzy clustering method have been used to determine the burn / normal skin. We selected 40 images, from the burn wound image dataset of the burn unit of the Karadeniz Technical University Faculty of Medicine Farabi Hospital. Although Euclidean distance is the most commonly used distance metric in image clustering methods, we examined the effects of different distance metrics on the clustering of burn wounds, in this study. We have evaluated the clustering performance of Euclidean, Mahattan, Jaccard, Cosine, Chebyshev, Minkowski distance metrics.We measured the performance of the distance metrics in terms of PBMF, Partition Coefficient, Cohesion and Separation validity indexes. As a result, we found that the Cosine distance metric gives the best result with 3 clusters.

Keywords: