Detection and segmentation of mitochondria from electron microscope tomography images


Tezin Türü: Doktora

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Enformatik Enstitüsü, Sağlık Bilişimi Anabilim Dalı, Türkiye

Tezin Onay Tarihi: 2016

Öğrenci: FARİS SERDAR TAŞEL

Danışman: ÜNAL ERKAN MUMCUOĞLU

Özet:

Recent studies exhibit that mitochondria have a significant role in cellular functions that are associated to the diseases of aging caused by neuron degeneration. These studies accentuate that the peripheral membrane and crista morphology of a mitochondrion deserves attention in order to reveal the relation between mitochondrial function and its physical structure. The analysis of the inner structures of mitochondria is carried out by electron microscope tomography (EMT) which provides detailed visualization of large volumes. In order to accelerate the studies that investigate the correlation between mitochondrial structure and its function, computerized segmentation of mitochondria with minimum manual effort is required. As a preliminary study, 2D detection and segmentation of mitochondria from transmission electron microcopy (TEM) images were performed on a limited dataset. An ellipse fitting algorithm utilizing double membrane features followed by a balloon snake extraction and a livewire-based automatic segmentation refinement process was applied. By considering the deficiencies of the initial attempt, a curve fitting approach was adopted and tested on a several datasets. For this purpose, a membrane extraction process was performed by utilizing a parabolic arc model. Then, active contour model based on curve energy was used to outline candidate mitochondrial regions. The final segmentation data were obtained by a validator function. Additionally, 3D extension of the algorithms were studied and provided. The proposed method achieved an Fscore performance of 0.84 on average. Average Dice similarity coefficient and median boundary error were measured as 0.87 and 14 nm respectively.