Automatic segmentation of cristae membranes in 3d electron microscopy tomography images using artificial neural networks

---

Tezin Türü: Yüksek Lisans

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

Tezin Onay Tarihi: 2016

Öğrenci: MERİH ALPHAN KARADENİZ

Danışman: ÜNAL ERKAN MUMCUOĞLU

Özet:

Electron Microscopy Tomography (EMT) technique produces 3D images of cells comprising hundreds of slices of high resolution frames. Segmentation of membranes in these images are necessary in order to reveal the relations between the structural components of the cell and its behaviour. The physical shape of the crista which is a membrane of the mitochondria has been hypostatized for being an early indicator for many diseases or mitochondrial dysfunctions. Automatic segmentation of cristae in EMT images are necessary since it needs a huge human effort to manually segment these membranes. In this study, a method for automatic and robust segmentation of the crista membrane in mitochondria is proposed. The method incorporates a pre-processing stage in which a bilateral image smoothing is applied for noise removal while preserving the crista membrane boundaries. The cristae membranes are first detected by an artificial neural network (ANN) trained on cropped mitochondria images from three different data sets. When a portion of the membrane boundary is almost or totally invisible, ANN may produce disconnected segmentation. In order to overcome this issue and increase the final performance by means of detecting the barely invisible membrane boundaries and decreasing false alarms, a boundary growing method called ‘directional growing’ is proposed. The method is tested with examples from four different data sets and numerical and visual analysis of the results are conducted