White Matter Bundle Reconstruction From Single-Shell Diffusion Magnetic Resonance Imaging: Test-Retest Reliability...

Area of Science:

• Neuroimaging
• Diffusion MRI
• White Matter Anatomy

Background:

• White matter (WM) bundle derivation typically requires high-resolution, multi-shell diffusion MRI (dMRI) acquisitions, limiting its use to research settings due to cost and time.
• Existing legacy dMRI datasets, often single-shelled with lower angular resolution, represent a vast resource for broader clinical applications and population studies.
• The reliability of WM bundles reconstructed from these legacy datasets is currently unclear.

Purpose of the Study:

• To evaluate the test-retest reliability of WM bundle reconstruction from lower-resolution, single-shell dMRI data.
• To compare the performance of three orientation distribution function (ODF) reconstruction methods: generalized q-sampling imaging (GQI), constrained spherical deconvolution (CSD), and single-shell three-tissue CSD (SS3T).
• To assess the potential of reconstructed WM bundles for predicting cognitive function.

Main Methods:

• Leveraged a large research dataset with 64-direction dMRI scans acquired as two independent 32-direction runs per subject.
• Reconstructed WM bundles using GQI, CSD, and SS3T methods from the 32-direction scans.
• Assessed bundle reliability using Dice coefficients and evaluated predictive performance for complex reasoning using extracted bundle features.

Main Results:

• The majority of WM bundles were reliably extracted from 32-direction, single-shell dMRI data across all tested ODF methods.
• Within-subject Dice coefficients were consistently higher than between-subject coefficients, indicating preserved individual anatomy.
• Stable prediction accuracies (r: 0.15-0.36) for complex reasoning were observed across test-retest data.
• SS3T demonstrated a favorable balance of sensitivity and specificity, high intra-class correlation, plausible bundles, and strong predictive performance compared to GQI and CSD.

Conclusions:

• Bundle-specific reconstruction methods can achieve robust performance even with lower angular resolution, single-shell dMRI data.
• ODF methods optimized for single-shell data, like SS3T, offer particular advantages for reconstructing WM bundles from legacy datasets.
• This work highlights the significant potential of utilizing existing dMRI data from healthcare and research settings to advance WM research and clinical applications.