Identification of melanoma cells: a method based in mean variance of signatures via spectral densities

摘要

In this paper a new methodology to detect and differentiate melanoma cells from normal cells through 1D-signatures averaged variances calculated with a binary mask is presented. The sample images were obtained from histological sections of mice melanoma tumor of 4 μm in thickness and contrasted with normal cells. The results show that melanoma cells present a well-defined range of averaged variances values obtained from the signatures in the four conditions used.