Generalised coherent point drift for group-wise multi-dimensional analysis of diffusion brain MRI data.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Diffusion Tensor Imaging (DTI) provides insights into white matter microstructure.
• Analyzing DTI-derived metrics like fractional anisotropy (FA) and fiber orientation across subjects is crucial for understanding neurological conditions.
• Existing methods like TBSS and VBM have limitations in group-wise analysis of complex DTI data.

Purpose of the Study:

• To develop a probabilistic framework for registering generalized point sets with multiple voxel-wise data features.
• To apply this framework for joint registration and clustering of DTI-derived data across multiple subjects.
• To facilitate inter-group comparisons of FA and fiber orientation in specific white matter regions.

Main Methods:

• A hybrid Student's t-Watson-Gaussian mixture model-based non-rigid registration framework was formulated.
• The approach jointly estimates non-rigid transformations for registering an unbiased mean template to white matter regions of interest (ROIs).
• It approximates the joint distribution of voxel spatial positions, principal diffusion axes, and FA values.

Main Results:

• The framework successfully registered DTI-derived data from multiple subjects.
• It enabled joint registration and clustering of voxel-wise DTI data.
• The analysis facilitated inter-group comparisons of FA and fiber orientation in the corpus callosum and cingulum between healthy controls, MCI, and AD patients.

Conclusions:

• The proposed probabilistic framework offers a novel approach for analyzing multi-feature voxel-wise data, particularly DTI.
• It overcomes limitations of conventional methods by enabling group-wise comparisons of complex DTI metrics.
• This facilitates a deeper understanding of white matter alterations in neurological disorders like Alzheimer's disease.