Brain MRI studies has reported inconsistent results regarding the abnormalities of brain structures in autism. The inconsistency may be due to factors such as the sample size, subject age and gender. However, heterogeneity within the autism diagnosis can significantly obscure the genetic basis of the disorder. Recently, we proposed a new definition of autism subgroups, which divided autism into essential autism and complex autism. Essential autism comprises the majority of patients and includes the more heritable segment of autism compared to complex autism, and a higher male to female ratio. Complex autism, on the contrary, shows more heterogeneous features such as lower IQs, more seizures, and poorer outcomes, etc.
Objectives: The aim of this study was to identify shape differences of the corpus callosum, the major commissural pathway between the brain hemispheres, between patients with essential autism and the controls using anatomical landmarks extracted from the corpus callosum (CC). We limit our subjects to essential autism group in order to rule out as much as within-group heterogeneity which can lead to inaccurate statistical findings in the brain morphology analysis.
Euclidean Distance Matrix Analysis (EDMA) and Thin-plate spline (TPS) are two traditional approaches to analyze the landmarks. EDMA uses landmark coordinate data to calculate all pairwise distances between landmarks. It is invariant to the coordinate system, which makes it biologically and statistically advantageous. TPS has been widely used to compare landmark configurations. The fundamental principle of TPS is the comparison of two different shapes by deforming one shape to the other. The deformation requires bending energy, which can be analyzed in terms of partial warp (PW) scores. Analysis of covariance is performed to compare the group mean PW scores.
Although the above landmark analyses can reveal some shape information, local shape morphologies at non-landmark locations cannot be detected. In this paper, we examine the abnormalities of the CC in essential autism by integrating the traditional landmark analyses with a landmark guided local shape comparison. A configuration of landmarks is identified in brain MRI mid-sagittal sections based on a predefined criterion. In the traditional analyses, we perform the aforementioned EDMA and TPS procedures. In the landmark guided shape comparison, we aim at finding the morphology at every location on the shape. Point correspondence is established locally based on the landmark correspondences, and statistical methods are used to compare two groups of shapes (essential autism vs. control) at every location. Each of the three analyses examines the shape morphology from a different perspective.
Brain MRIs of 30 essential autistic patients and 24 controls are studied. The results revealed significant form difference in the landmarks between patients and controls. The corresponding pairwise distances that caused the form difference were found in anterior and posterior of the CC. No influence of the size or the diagnosis on the shape deformation was found, but a reduced centroid size of the CC in patients was found.
We proposed a new landmark based shape analysis method and the result is consistent with previous studies.