Classification in DTI using shapes of white matter tracts

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Computational Neuroscience

Background:

• Diffusion Tensor Imaging (DTI) offers insights into brain white matter structure, including fiber bundle geometry and tissue properties.
• Quantitative DTI measures like tensor orientation and anisotropy characterize tissue properties.

Purpose of the Study:

• To evaluate the effectiveness of white matter tract shape representations derived from DTI data for classifying distinct population groups, specifically autistic versus control individuals.
• To explore the utility of geometric modeling of tracts for clinical group differentiation.

Main Methods:

• Extraction of white matter fiber bundles using regions of interest on aligned brain volumes.
• Analysis based entirely on the geometric modeling of extracted fiber tracts.
• Development of classifiers utilizing tract shape representations.

Main Results:

• Demonstrated that classifiers built using tract shape representations achieve reasonable prediction accuracies.
• Showcased the ability to build accurate classifiers without heavy reliance on spatial normalization techniques.
• Highlighted the efficiency of using invariant features for large sample size studies.

Conclusions:

• Geometric modeling of white matter tracts from DTI data is a viable approach for population group classification.
• Tract shape analysis provides an efficient alternative to voxel-based methods, reducing the need for extensive spatial normalization.
• This method facilitates robust classification with potential for large-scale neuroimaging studies.