Characterization of neurite and soma organization in the brain and spinal cord with diffusion MRI

Kurt G. Schilling, Marco Palombo, Atlee Witt, Kristin P. O’Grady, Marco Pizzolato, Bennett A. Landman, Seth A. Smith

TL;DR

This paper presents a novel, clinically feasible diffusion MRI (dMRI) protocol that can simultaneously image the microstructure of both the human brain and spinal cord. By applying advanced dMRI models (DTI, SMI, and SANDI), the researchers demonstrate that it's possible to characterize neurite and soma organization across the entire central nervous system, opening doors for studying neurological conditions that affect both regions.

What Problem Does This Paper Address?

The central nervous system (CNS) is a highly interconnected network, with the brain and spinal cord working in tandem to control our functions. However, most neuroimaging studies tend to focus on either the brain or the spinal cord in isolation. This separation limits our ability to understand how diseases or injuries might affect the entire CNS. Diffusion MRI (dMRI) is a powerful, non-invasive tool for probing tissue microstructure, but applying advanced dMRI models, particularly those that can differentiate between neuronal cell bodies (soma) and their extensions (neurites), has historically been challenging, especially for the spinal cord due to its smaller size and unique anatomy. This paper aims to bridge this gap by developing and validating a unified dMRI protocol that can capture detailed microstructural information from both the brain and spinal cord simultaneously.

Methods

The researchers developed a clinically feasible dMRI protocol on a 3T scanner. This protocol utilized a multi-shell acquisition (meaning data was collected at multiple diffusion gradient strengths, or b-values) to enable the application of three different dMRI models:

  • Diffusion Tensor Imaging (DTI): A standard technique that models water diffusion as a 3D Gaussian distribution, providing measures like fractional anisotropy (FA) and mean diffusivity (MD).
  • Standard Model Imaging (SMI): An extension of DTI that models axons as "sticks" within an extra-axonal space, allowing for the estimation of neurite fraction (f) and orientation dispersion. This was the first time SMI was evaluated in the spinal cord.
  • Soma and Neurite Density Imaging (SANDI): A more advanced model that further distinguishes between water diffusion within neurites and within neuronal cell bodies (soma), providing estimates of soma fraction (fsoma) and soma radii (Rsoma). This was also the first time SANDI was applied to the spinal cord and simultaneously to both brain and cord.

The study involved eleven healthy participants, with six undergoing repeat scans to assess reproducibility. Sophisticated preprocessing pipelines were used for both brain and spinal cord data, including denoising, motion correction, and artifact removal. The researchers then evaluated image quality, contrast between tissue types, and the reproducibility of the derived metrics using statistical measures like Pearson correlation, test-retest variability (TRV), and intraclass correlation coefficient (ICC).

Datasets

The study utilized data from 11 healthy human participants. Six of these participants underwent a second scan (rescan) to evaluate the test-retest reliability of the imaging protocol and derived metrics. No new benchmarks were introduced; rather, the study aimed to demonstrate the feasibility of existing advanced dMRI models on a harmonized brain and spinal cord acquisition.

Key Insights and Ideas

  • Unified Neuroaxis Imaging: The most significant contribution is the demonstration that advanced microstructural imaging (SMI and SANDI) can be performed consistently across both the brain and spinal cord using a single, clinically feasible protocol. This opens up the possibility of studying diseases that affect the entire CNS as a connected system.
  • Spinal Cord Microstructure Revealed: This work represents the first application of SMI and SANDI models to the in vivo human spinal cord, providing novel insights into its microstructural organization.
  • Feasible High-Quality Acquisition: The protocol achieved high image quality even at high b-values, crucial for resolving fine microstructural details, with visible contrast between white and gray matter in both regions.
  • Reproducibility Challenges in the Cord: While brain metrics showed good reproducibility, the spinal cord exhibited slightly lower reproducibility and contrast. This is attributed to factors like smaller voxel sizes, partial volume effects, and challenges in image preprocessing for the cord.

Key Results

  • Image Quality: The dMRI protocol produced high-quality images with visible contrast between white and gray matter in both the brain and spinal cord, even at high b-values (up to 6000 s/mm²).
  • Contrast: SMI and SANDI models revealed biologically plausible contrasts for neurite and soma organization in the brain, and similar, though less pronounced, contrasts were observed in the spinal cord.
  • Reproducibility: SMI and SANDI showed comparable reproducibility to DTI in the brain. In the spinal cord, reproducibility was generally reduced compared to the brain but remained moderate for most measures.
  • Soma and Neurite Signatures: The SANDI model successfully identified increased soma size and fraction in gray matter regions of the brain, consistent with known neuroanatomy. Similar trends were observed in the spinal cord, though with more pronounced partial volume effects.

Limitations and Future Work

  • Limited Sample Size: The study had a small sample size (N=11), which limits the ability to establish normative values or fully validate the derived indices. Future work with larger cohorts is needed.
  • Spinal Cord Challenges: Partial volume effects, smaller voxel sizes, and motion artifacts present greater challenges for spinal cord imaging, leading to slightly reduced reproducibility and contrast compared to the brain.
  • Exchange Bias: Long diffusion times in the protocol can introduce biases due to water exchange between compartments, particularly in gray matter. The models used do not fully account for this time dependence.
  • Future Directions: The authors suggest investigating pathway-specific analyses of projection tracts, exploring alternative diffusion weightings (e.g., planar, spherical) to improve model fitting, and further optimizing spinal cord image processing. Crucially, they highlight the need to investigate the clinical utility of these advanced microstructural measures for studying neurological disorders like multiple sclerosis, where combined brain and cord pathology is common.

Summary updated

Abstract

The central nervous system (CNS), comprising both the brain and spinal cord, is a complex network of white and gray matter responsible for sensory, motor, and cognitive functions. Advanced diffusion MRI (dMRI) techniques offer a promising mechanism to non-invasively characterize CNS architecture, however, most studies focus on the brain or spinal cord in isolation. Here, we implemented a clinically feasible dMRI protocol on a 3T scanner to simultaneously characterize neurite and soma microstructure of both the brain and spinal cord. The protocol enabled the use of Diffusion Tensor Imaging (DTI), Standard Model Imaging (SMI), and Soma and Neurite Density Imaging (SANDI), representing the first time SMI and SANDI have been evaluated in the cord, and in the cord and brain simultaneously. Our results demonstrate high image quality even at high diffusion weightings, reproducibility of SMI- and SANDI-derived metrics similar to those of DTI with few exceptions, and biologically feasible contrasts between and within white and gray matter regions. Reproducibility and contrasts were decreased in the cord compared with that of the brain, revealing challenges due to partial volume effects and image preprocessing. This study establishes a harmonized approach for brain and cord microstructural imaging, and the opportunity to study CNS pathologies and biomarkers of structural integrity across the neuroaxis.
BibTeX
@inproceedings{schilling2025characterization,
  author = {Kurt G. Schilling and Marco Palombo and Atlee Witt and Kristin P. O’Grady and Marco Pizzolato and Bennett A. Landman and Seth A. Smith},
  title = {{Characterization of neurite and soma organization in the brain and spinal cord with diffusion MRI}},
  booktitle = {Imaging},
  year = {2025},
  doi = {10.1162/imag.a.111},
}
Figure