Clustering of Dielectric and Colour Profiles of an Ex-vivo Burnt Human Skin Sample

Abstract

This work aims to introduce two techniques to characterize human burnt skin based on permittivity measurement and image processing. The first method is the sectorized measurement of permittivity (dielectric profiling, DEP) by using an open-end coaxial probe technique. The second method is the analysis of color variation in the burnt skin sample through image processing. Statistical analysis is done using tools such as Analysis of Variance (ANOVA), k-means, in order to classify, evaluate the data. As part of the classification, the experimental data are clustered into five groups based on the distribution of means (dielectric profiles) and centroids (color profiles). The color image is converted into a gray image and resized to a one-dimensional array. Furthermore, the analysis is done based on the intensity range, various centroid values, and silhouette analysis. The clustering results we obtained with these two methods can be used for comparing the dielectric characteristics with the color variation of the burnt human skin to assess burn degree.