Master Thesis Defense: Shaghayegh Taheri Hosseinabadi

Speaker: Shaghayegh Taheri Hosseinabadi

Supervisors: Drs. T. D. Bui, T. Fevens

Examining Committee:
Drs. A. Krzyzak, C. Poullis, G. Butler (Chair)

Title:  Robust Nuclei Segmentation in Cytohistopathological Images Using Statistical Level Set Approach with Topology Preserving Constraint

Date: Wednesday, April 6, 2016

Time: 10:00 a.m.

Place: EV 11.119

ABSTRACT

Computerized assessments of cyto-histological specimens have drawn increased attention in the field of digital pathology as the result of developments in digital whole slide scanners and computer hardwares. Due to the essential role of nucleus in cellular functionality, automatic segmentation of cell nuclei is a fundamental prerequisite for all cyto-histological automated systems. In 2D projection images, nuclei commonly appear to overlap each other, and the separation of severely overlapping regions is one of the most challenging tasks in computer vision. In this thesis, we will present a novel segmentation technique which effectively addresses the problem of segmenting touching or overlapping cell nuclei in cyto-histological images. The proposed framework is mainly based upon a statistical level-set approach along with a topology preserving criteria that successfully carries out the task of segmentation and separation of nuclei at the same time. The proposed method is evaluated quantitatively and qualitatively on both H&E and fluorescent stained images and the results indicate that the method outperforms the conventional nuclei segmentation approaches, i.e. thresholding and watershed segmentation.