Quantitative Analysis of Pathological Mitochondrial Morphology in Neuronal Cells in Confocal Laser Scanning Microscopy Images

The quantitative analysis of mitochondria morphology is of great interest on medical research of neurodegnerative diseases, and also for cancer disease, and Diabetis melitus. The state-of-the-art in mitochondrial analysis is based on length measurements of randomly selected individual mitochondria. As manual segmentation and evaluation for whole cell analysis is very time-consuming and error-prone, we present a framework for semi-automatic processing of these data which only requires a minimal amount of simple manual user interactions. The mitochondrial structures are segmented by Oriented Dierences of Boxes (ODoB) lters and normalized to the cell volume. Thus, the core ingredient of our system is an algorithm for approximating convex hulls of point sets, called the Regularized Geometric Hull (RGH). As the proposed framework produces promising results in experiments performed in cultured primary motor neurons which were successfully approved by medical experts, it will be subject to further investigation in clinical studies.