Tissue Differentiation Based on Classification of Morphometric Features of Nuclei

Abstract

The aim of the article is to analyze the shape of nuclei of various tissues and to assess the tumor differentiation based on morphometric measurements. For this purpose, an experiment was conducted, the results of which determine whether it is possible to determine a tissue’s type based on the mentioned features. The measurements were performed on a publicly available data set containing 1,356 hematoxylin- and eosin-stained images with nucleus segmentations for 14 different human tissues. Morphometric analysis of cell nuclei using ImageJ software took 17 parameters into account. Classification of the obtained results was performed in Matlab R2018b software using the SVM and t-SNE algorithms, which showed that some cancers can be distinguished with an accuracy close to 90% (lung squamous cell cancer vs others; breast cancer vs cervical cancer).