Breadcrumb

 
 

Performance analysis of automatic techniques for tissue classification in magnetic resonance images of the human brain

Title:

Performance analysis of automatic techniques for tissue classification in magnetic resonance images of the human brain

Kollokian, Vasken (1996) Performance analysis of automatic techniques for tissue classification in magnetic resonance images of the human brain. Masters thesis, Concordia University.

[img]
Preview
PDF
4Mb

Abstract

Classification of Magnetic Resonance (MR) images of the human brain into anatomically meaningful tissue labels is an important processing step in many research and clinical studies in neurology. The medical imaging research community is presented with a wide choice of classification algorithms from artificial intelligence and pattern recognition. This thesis describes the development of a controlled test environment, where different classification algorithms were implemented and their performance evaluated in a brain imaging context. Furthermore, a mechanism for automating supervised classification algorithms is proposed through the use of a priori knowledge of neuro-anatomy, presented in the form of brain tissue probability maps. The results obtained through the automated methods compared favorably to those obtained through human supervision. The performance of five supervised (Artificial Neural Networks, Bayesian, k-Nearest Neighbors, C4.5 decision tree, Minimum Distance) and two unsupervised (Hard C Means, Fuzzy C Means) classification algorithms is compared under varying conditions of MR imaging artifacts. The Artificial Neural networks classifier was observed to be the best overall performer

Divisions:Concordia University > Faculty of Engineering and Computer Science > Computer Science and Software Engineering
Item Type:Thesis (Masters)
Authors:Kollokian, Vasken
Pagination:xx, 128 leaves ; 29 cm.
Institution:Concordia University
Degree Name:Theses (M.Comp.Sc.)
Program:Dept. of Computer Science
Date:1996
Thesis Supervisor(s):Shingal, Rajjan
ID Code:192
Deposited By:Concordia University Libraries
Deposited On:27 Aug 2009 13:10
Last Modified:08 Dec 2010 10:13
Related URLs:
All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Document Downloads

More statistics for this item...

Concordia University - Footer