Statistical Analysis of the Relationship between Spots and Structures in Microscopy Images

Abstract

Fluorescence microscopy image analysis plays an important role in biomedical diagnostics and is an essential approach for researching and investigating the development and state of various diseases. In this paper we describe an approach for analyzing nanoscale microscopy images in which spots and background structures are identified and their relationship is quantified. A spatial analysis approach is used for identifying spots, then clustering of these spots is performed and those clusters are characterized using a series of here defined features. These cluster characteristics are used for comparing images via statistical hypothesis tests (using the Kolmogorov-Smirnov test for the equality of probability distributions). Moreover, to achieve a better distinction we additionally define features that quantify the relationship of clusters of spots and background structures. In the empirical section we demonstrate the use of this approach in the analysis of microscopy images of brain structures of patients potentially suffering from a neural disease (e.g., depression or schizophrenia). Using the here presented approach we will be able to investigate the development and state of various diseases in a better way and help to find more systematic medication of diseases in the future.

The work described in this paper was done within the FIT-IT project “NanoDetect: A Bioinformatics Image Processing Framework for Automated Analysis of Cellular Macro and Nano Structures” (project number 835918) sponsored by the Austrian Research Promotion Agency (FFG).