3D image processing for mitochondria morphology variation analysis

Abstract

Mitochondria play an important role in producing energy of cell, aging and cell death. Previous investigations suggest that mitochondrial dysfunctions are associated with neurodegenerative, and the peripheral blood cells of those patients have validation of mitochondrial morphology. This study presents an analysis system of automated 3D mitochondrial morphology of lymphocytes in 3D images to reveal morphological features of mitochondria, and this system is specific for analyzing mice T lymphocyte precursors in thymus and mature lymphocytes in spleen. Mitochondria in original confocal and deconvoluted wild-field micrographs are segmented by Support Vector Machine, and then image features are extracted for morphological subtype clustering. The distribution of subtypes in individual lymphocytes is used to classify precursors and mature lymphocytes by decision tree. There are more short tubules in spleen lymphocyte, and more globules in thymus lymphocyte. The accuracy of cell classified from confocal and wild-field images in this system are both over 96% for training set, and for testing set, the accuracy for confocal and wild-field micrographs are 79% and 71% respectively. In conclusions, this new system can be used for investigations of lymphocyte development and drug discovery for inflammation and immunity.