The British Machine Vision Association and Society for Pattern Recognition 

BibTeX entry

  AUTHOR={Kolawole Babalola},
  TITLE={Three-Dimensional Morphometric Analysis of Brain Ventricles
    Using Statistical Shape Models},
  SCHOOL={University of Manchester},


It is widely accepted that schizophrenia is accompanied by an increase in the volume of the lateral ventricles of the brain. It is of interest to investigate if the volumetric changes are accompanied by specific localised shape changes. If these can be shown they will aid in understanding what is a complex disease. Studies based on volumetric measurements have been carried out to further characterise the observed differences. However, volume is not a powerful shape descriptor - as ventricles with the same volume can have quite different shapes. Additionally, because of the intrinsic natural variation of biological structures statistical methods which allow the separation of variability due to disease from natural variability are of interest. Statistical shape models (SSMs) are a tool from computer vision that capture the inherent variation in the shape of a specified class of object based on statistics learnt from a sample representative of the population of interest. Our choice of SSM for carrying out morphometric analysis is the point distribution model (PDM). This type of model has been widely used as a segmentation tool in 2D and application in 3D is a subject of active research. The aim of this thesis is to use a 3D PDM to quantify localised changes in the shape of the lateral ventricles associated with schizophrenia. Starting with magnetic resonance (MR) images of 30 control subjects and 39 age and sex matched schizophrenics, we describe the pre-processing of the MR images to improve their quality prior to segmentation. A semi-automatic approach to the segmentation of the lateral ventricles was adopted. We give details of an automatic approach to the construction of the 3D PDM using crest points as curvature-based landmarks on the ventricles. We introduce a novel application of the transportation algorithm to solve the problem of defining correspondences of crest points. Issues regarding the comparison of 3D PDMs are also discussed. Applying discriminant analysis to the most important shape parameters obtained from the PDM, the means of the schizophrenic and control groups are significantly different (p 10(-13)). The shape changes observed were localised to three regions : the temporal horn (its tip near the amygdala, and along its body near the parahippocampal fissure), the central part of the lateral ventricles around the corpus callosum, and the tip of the anterior horn in the region of the frontal lobe. The differences in the temporal region and anterior horns are in regions close to structures thought to be implicated in schizophrenia. Investigations of asymmetry between left and right ventricle pairs was also undertaken. This showed that a degree of asymmetry existed in both schizophrenics and controls, and that the nature of asymmetry was different in the two groups.