On Supporting Cancer Grading Based on Histological Slides Using a Limited Number of Features

Abstract

A computer assisted cancer diagnosis approach based on histopathological images is outlined in the current study. It mainly detects the lighter components that appear in each slide and computes simple morphological features for them, as well as basic statistic about these. This relatively short numerical data set is subsequently fed to a classifier for learning the correspondence between the features and their classes. The classification accuracy reaches outputs similar to other more complex methodologies and therefore encourages future investigation.