Semi-automated CT image segmentation program for the easy calculation of cerebrospinal fluid (csf) and brain tissue

Open Access
Macdonald, Michael J
Graduate Program:
Mechanical Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
April 29, 2014
Committee Members:
  • Steven Schiff, Thesis Advisor
  • Dr Jack Langelaan, Thesis Advisor
  • Stephen Jacob Piazza, Thesis Advisor
  • Medical image segmentation
  • CT imaging
  • Particle filter
  • Pixel classification
  • Edge tracing
  • Hydrocephalus
There is a lack of easy, efficient, and affordable software methods for performing volumetric segmentation of computed tomography (CT) images. This program combines an automated pixel classification algorithm with a semi-automated edge-tracking algorithm to quickly and accurately segment entire CT image stacks of the brain. The edge-tracking algorithm incorporates a novel particle-filter method along with an automated seed point selection algorithm. The classification method establishes pixel class likelihoods through the use of intensity thresholds created by fitting a double Gaussian curve to the image histogram. The current software program was created in Matlab and is being used to analyze patient CT scans as part of an on-going NIH phase III surgical trial being conducted in Uganda. This software program will provide a valuable tool to clinicians in developing countries as a means to quickly analyze brain tissue and cerebrospinal fluid volumes of patients seeking care.