Statistical 3D Shape Analysis of Lateral Ventricles in Autism

Background:

Autism is a heterogeneous neurodevelopmental disorder of unknown etiology marked by social, emotional and communication impairments. The ventricular system is a structure in the center of the brain filled with fluid. Enlargement of the ventricles in autism has been reported in previous works. However, the magnitude of the differences revealed by volume analysis is limited, because it does not consider how shapes overlap each other. Shape analysis can precisely locate morphological changes in brain structures which cannot be reflected in volume measurements, thus it becomes more and more popular in neuroimaging community. Very few studies have been done with regard to the shape morphology of the ventricles in autism.

Objectives:

The aim of this paper is to study the shape differences of the lateral ventricles between autistic children and normal controls. 3D models of ventricle shapes are compared to reveal the shape difference at every surface location.

Methods:

Children with autism were recruited from the Thompson Center for Autism and Neurodevelopmental Disorders. Control subjects with matching gender, age and ancestry were recruited from the community. Student t-test was used to compare the ages and χ² test was used to compare the gender ratios. There was no significant difference between the two groups in terms of age, gender and race. This study was approved by the Health Sciences Institutional Review Board. The parents or legal guardians of all subjects provided written consent for participation in this study.

The 3D ventricle models were obtained using the semiautomatic segmentation software that has been developed in our previous work. Six anatomical landmarks were identified. Pseudo-landmarks on the surface were then interpolated using the method in our previous work. Generalized Procrustes Analysis is performed on all the 3D shapes in order to eliminate the shape differences caused by translation, rotation and scaling. Difference between groups at every surface location was tested using Hotelling T² two-sample metric. Since comparisons are made at thousands of surface points, we adopted False Discovery Rate estimation (FDR) for p-value correction.

Results:

From the overlaid average structures of the two groups, we find that the posterior horn of both ventricles in autistic children are bent more inward, the anterior horn of the left ventricle is bent more downward and the anterior horn of the right ventricle is bent more outward in the autistic children. A two-tailed alpha level of 0.05 is chosen as the significance threshold for the raw p-values. The raw significance maps of both ventricles show significant shape difference in part of the anterior and posterior horns between autistic children and controls, but the corrected significance maps show no shape difference between the two groups.

Conclusions:

This paper studies the shape differences of the lateral ventricles between autistic children and controls. The overlaid mean structures show shape difference in the anterior and posterior horns, but statistical testings do not show any significance in those differences. This may be due to small sample size and further experiment needs to be done when more data are available.