Hierarchical Clustering Methodologies for the Detection of Senescent Cells

Abstract

This Master Thesis project is related to a data set of microscope images whose puppose is to analyze senescence. Senescence is a dynamic process whereby cells stop to duplicate and change their morphology. The results here described are a first attempt to classify cells by their morphology, in order to find a method to detect the cluster of senescent cells. The microscope images are first of all preprocessed and the objects are properly segmented. Then some features are extracted for each object, namely "area", "circularity", "eccentricity" and "convexity defects". With these features, an Agglomerative Hierarchical Clustering is applied with different methods. It results that, on the basis of the extracted features, the cells in 24 hours can be classified in 4 clusters with diffent morphological characteristics.