Classifying human blood samples using characteristics of single molecules and cell structures on microscopy images

Abstract

In this paper we present a method for the definition of characteristics of single molecules as well as of cell structures on fluorescence microscopy images for classifying human disease states. Fluorescence microscopy is one of the most emerging fields in modern laboratory diagnostics and is used in various research areas, for instance in studies of protein-protein interactions, analyses of cell interactions, diagnostics, or drug distribution studies. We have developed a new combinatory workflow comprising image processing and machine learning techniques to define characteristics out of given fluorescence microscopy images and to classify given images of blood samples according to their level of protein expression (high or low), i.e. according to their disease state. This combinatory workflow is not adapted to a specific illness but is usable for all kinds of diseases that can be characterized using single molecule fluorescence microscopy.