Corpus Callosum Subdivision Based on a Probabilistic Model of Inter-hemispheric Connectivity

Martin Andreas Styner, Ipek Oguz, Rachel Gimpel Smith, Carissa Cascio, Matthieu Jomier

Abstract

Traditional methods for subdividing the corpus callosum (CC) rely on rigid geometric schemes that often fail to account for individual anatomical variability and functional connectivity. This paper presents an automated, probabilistic approach that integrates statistical shape analysis with Diffusion Tensor Imaging (DTI) tractography. By modeling the probability of connectivity between specific cortical lobes and the CC, the method provides a biologically grounded subdivision. Preliminary results in a pediatric longitudinal study demonstrate the model's stability and its ability to detect regional growth patterns, such as a 26% increase in the anterior-frontal region between ages 2 and 4.

Introduction

The corpus callosum (CC) serves as the primary commissural pathway, facilitating inter-hemispheric communication. In neurodevelopmental research, subdividing the CC is crucial for identifying regional changes associated with growth or pathology, such as autism. The conventional "gold standard," the Witelson scheme, utilizes a simple bounding box to partition the CC. However, this geometric approach is limited by its inability to handle the non-convex nature of the brain and its lack of reproducibility across variable neuroanatomy.

Overview of the connectivity-based subdivision approach, showing the relationship between cortical lobes, DTI fibers, and the resulting CC partitions.

To address these limitations, we propose a method that defines CC subregions based on the actual destination of the fibers passing through them, rather than arbitrary spatial boundaries.

Methodology

The proposed pipeline combines structural MRI and DTI data through a multi-step process:

  1. Automated Segmentation: An Active Shape Model (ASM) based on Fourier descriptors was employed to automatically extract the 2D contour of the CC from mid-sagittal structural images.
  2. Tractography and Connectivity Mapping: DTI data were used to trace inter-hemispheric fibers connecting corresponding cortical lobes (frontal, parietal, and a combined occipital-temporal region).
  3. Probabilistic Modeling: Rather than assigning hard boundaries, we calculated the distance from each point on the CC contour to the various fiber sets. This allowed for the generation of probabilistic maps, representing the likelihood that a specific CC region connects to a particular lobe.
Probabilistic maps showing the distribution of connectivity likelihood for different cortical regions along the CC contour.
  1. Template Generation: An average probabilistic model was created from a training set of five pediatric subjects. This template can be mapped onto new subjects using geometric correspondence, enabling connectivity-based analysis even in datasets lacking high-quality DTI.

Results and Discussion

The probabilistic method demonstrated superior stability compared to geometric schemes. In a longitudinal study of three healthy children scanned at ages 2 and 4, the model identified significant regional growth. Notably, the anterior-frontal region exhibited the highest relative growth rate at 26%, despite the posterior-frontal region being larger in total area.

Anatomical Accuracy and Stability

By embracing the inherent uncertainty and overlap in fiber tracts through probabilistic modeling, the method avoids the sensitivity to small shape variations that plagues "hard" subdivision methods. This approach ensures that functional connectivity information is preserved in the resulting partitions.

Retrospective Utility

A key advantage of this model is its applicability to legacy datasets. Because the final subdivision is projected onto the CC contour, researchers can apply these connectivity-based insights to older structural MRI studies where DTI was not performed.

Conclusion

This study introduces a robust, automated framework for CC subdivision grounded in inter-hemispheric connectivity. While the small sample size and DTI resolution (necessitating the grouping of temporal and occipital lobes) are limitations, the preliminary findings highlight the dynamic nature of frontal lobe development in early childhood. Future work will focus on refining the model with higher-resolution data and larger clinical cohorts to explore regional CC variations in neurodevelopmental disorders.

Summary updated

Abstract

Statistical shape analysis has become of increasing interest to the neuroimaging community due to its potential to locate morphological changes. In this paper, we present the a novel combination of shape analysis and Diffusion Tensor Image (DTI) Tractography to the computation of a probabilistic, model based corpus callosum (CC) subdivision. The probabilistic subdivision is based on the distances of arc-length parameterized corpus callosum contour points to trans-callosal DTI fibers associated with an automatic lobe subdivision. Our proposed subdivision method is automatic and reproducible. Its results are more stable than the Witelson subdivision scheme or other commonly applied schemes based on the CC bounding box. We present the application of our subdivision method to a small scale study of regional CC area growth in healthy subjects from age 2 to 4 years.
BibTeX
@inproceedings{styner2005corpus,
  author = {Martin Andreas Styner and Ipek Oguz and Rachel Gimpel Smith and Carissa Cascio and Matthieu Jomier},
  title = {{Corpus Callosum Subdivision Based on a Probabilistic Model of Inter-hemispheric Connectivity}},
  booktitle = {International Conference on Medical Image Computing and Computer-Assisted Intervention},
  year = {2005},
  doi = {10.1007/11566489\_94},
}
Figure