Back to search

Aristotle University Of Thessaloniki (AUTH)

Μορφολογική ανάλυση και αυτόματη κατάτμηση βιολογικών δομών από ιατρικές απεικονίσεις τους

Abstract

dc:description

The objective of this dissertation is the development of automatic techniques towards the reliable segmentation of hippocampus from brain magnetic resonance images. Over the last years the relevant research has revealed the implication of hippocampus structural changes in many brain disorders, and that assessing its morphology should be adopted in clinical practice, as a new form of biomarker to help disease diagnosis. The aforementioned findings are based on comparisons between healthy and diseased subjects, using manually defined hippocampus segmentations from small number of subjects, due to the implicated cost and the inter-rater variability. In order to validate and eventually establish the use of hippocampus morphology as a biomarker in clinical practice, it is required that there exist automatic methods that offer reproducible results and apply them for large-scale experiments. This constitutes the motivation of this dissertation, which focuses on the use of active contour models and incorporating to them two forms of prior knowledge about hippocampus, in an effort to model the human experts' knowledge. The first form of prior knowledge is about the spatial distribution of hippocampus, while the second one is about the varying boundary properties of hippocampus. Given the boundary properties, it can be decided which type of information or which blending of multiple types of information is the optimum one to drive the segmentation for each specific boundary part. The optimum local blending of the different types of information is defined through a local interaction map for the energy terms used within a hybrid active contour model, which is formulated based on the prior knowledge about the varying boundary properties. Two different ways for formulating the local interaction map are proposed within this dissertation. The proposed methodology has been designed to work on top of the multi-atlas framework, taking advantage of its reliable outcomes. The evaluation of the proposed methodology has been performed using three different MRI datasets, which provide manually defined hippocampus segmentations, verifying its improvements on top of the multi atlas output and showing its superior performance against the existing techniques.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2014

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Zarpalas, Dimitrios
  • Ζαρπαλάς, Δημήτριος

Subjects

dc:subject × 16

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/34699

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Zarpalas, Dimitrios; Ζαρπαλάς, Δημήτριος. Μορφολογική ανάλυση και αυτόματη κατάτμηση βιολογικών δομών από ιατρικές απεικονίσεις τους. Aristotle University Of Thessaloniki (AUTH), 2014. http://hdl.handle.net/10442/hedi/34699