Quantitative mapping of the per‐axon diffusion coefficients in brain white matter

Enrico Kaden, Frithjof Kruggel, Daniel C. AlexanderMagnetic 2015 · DOI

TL;DR

This paper introduces a novel "Spherical Mean Technique" (SMT) to quantitatively map diffusion properties within individual axons in white matter, disentangling them from the complex arrangement of fibers within a voxel. This breakthrough allows for a more accurate understanding of brain microstructure, potentially leading to improved diagnosis and monitoring of neurological conditions.

What Problem Does This Paper Address?

Diffusion MRI is a powerful tool for non-invasively probing the microscopic structure of brain white matter. However, a major challenge is that the measured signal is an average over many axons within a voxel, which can have complex and varying orientations. Existing methods often struggle to separate the true diffusion properties of individual axons (like their caliber and myelination) from the effects of fiber dispersion and crossing. This makes it difficult to accurately assess the underlying tissue microstructure and can limit the sensitivity of MRI to subtle changes associated with neurological diseases. Previous attempts to address this often relied on complex, non-standard MRI sequences or made simplifying assumptions about fiber organization that don't hold true in many brain regions.

Methods

The core innovation is the Spherical Mean Technique (SMT). The key insight is that for a fixed diffusion weighting (b-value), the average diffusion signal across all possible gradient directions is independent of the intravoxel fiber orientation distribution. This "spherical mean" of the signal is solely determined by the diffusion properties of individual axons.

The authors model the diffusion signal of an individual axon using a second-order approximation, which yields two key parameters: $\lambda$ (the effective diffusion coefficient parallel to the axon) and $\lambda_1$ (the effective diffusion coefficient perpendicular to the axon). These are referred to as the "per-axon diffusion coefficients."

SMT involves:

  1. Acquiring diffusion MRI data with multiple b-values and a sufficient number of gradient directions.
  2. Calculating the spherical mean of the diffusion signal for each b-value.
  3. Using a constrained least-squares approach to fit the spherical mean signal to the theoretical model, thereby estimating $\lambda$ and $\lambda_1$.

This method is advantageous because it uses standard, clinically available pulse sequences and does not require prior knowledge of the fiber orientation distribution.

Datasets

The study utilized data from the Human Connectome Project (HCP) Lifespan pilot data (Phase 1a). This dataset was acquired on a Siemens 3 T Skyra MRI scanner with a 32-channel head coil. The diffusion data consisted of two b-shells (approximately 1000 and 2500 s/mm²) with 76 and 75 gradient directions, respectively, which were uniformly distributed on the sphere. The data was preprocessed using the HCP's Minimal Preprocessing Pipeline.

Key Insights and Ideas

  • Decoupling Microstructure from Architecture: The most significant insight is the ability to isolate the diffusion properties of individual axons from the complex spatial arrangement of fibers within a voxel. This is crucial for accurately characterizing tissue microstructure.
  • "Per-Axon" Metrics: The introduction of $\lambda$ and $\lambda_1$ provides direct measures of diffusion along and across axons, offering a more fundamental understanding of white matter than traditional metrics like fractional anisotropy (FA).
  • Clinical Feasibility: SMT's reliance on standard MRI sequences and its robustness to varying gradient direction sampling make it highly applicable in clinical settings, potentially enabling earlier and more sensitive detection of neurological changes.
  • Unveiling Tissue Heterogeneity: The resulting maps of $\lambda$ and $\lambda_1$ reveal significant variations across different white matter regions, suggesting that these parameters are sensitive to local microstructural differences (e.g., axon caliber, myelination, interaxonal space).

Key Results

  • Quantitative Maps: The study successfully generated quantitative maps of the per-axon diffusion coefficients ($\lambda$ and $\lambda_1$) in vivo for a healthy volunteer.
  • Robustness to Gradient Sampling: Simulations demonstrated that SMT can provide reasonably accurate estimates of $\lambda$ and $\lambda_1$ even with a reduced number of gradient directions (e.g., 25 directions per b-shell), suggesting clinical feasibility with shorter scan times.
  • Discrimination from Fiber Dispersion: The paper highlights that the per-axon anisotropy ($\lambda/\lambda_1$) is significantly higher in white matter than the fractional anisotropy derived from the standard tensor model, underscoring SMT's ability to factor out the effects of fiber dispersion.
  • Identification of Tissue Clusters: Analysis of the per-axon diffusion coefficients revealed distinct clusters, correlating with different tissue types (white matter, gray matter, CSF, and partial volume effects), demonstrating SMT's potential for tissue segmentation.

Limitations and Future Work

  • Second-Order Approximation: The current implementation uses a second-order approximation for the impulse response function. While a good compromise, more complex response functions could potentially capture finer microstructural details but would require more experimental data.
  • Noise Sensitivity: While SMT is robust, the accuracy of the estimates is still influenced by signal-to-noise ratio (SNR). The paper discusses methods to mitigate Rician noise bias.
  • Future Work: The authors suggest extending SMT to more complex response functions to better characterize axon microanatomy. They also propose exploring optimal experimental designs for maximizing sensitivity to target parameters within clinical time constraints and investigating the elimination of partial volume effects from CSF contamination. The potential for SMT in discriminating between different tissue types, such as tumors, is also highlighted.

Summary updated

Abstract

The estimation of the per-axon diffusion coefficients is essential for the accurate recovery of the fiber orientation distribution. In addition, the spherical mean technique enables us to discriminate microscopic tissue features from fiber dispersion, which potentially improves the sensitivity and/or specificity to various neurological conditions. Magnetic Resonance in Medicine published by Wiley Periodicals, Inc.
BibTeX
@inproceedings{kaden2015quantitative,
  author = {Enrico Kaden and Frithjof Kruggel and Daniel C. Alexander},
  title = {{Quantitative mapping of the per‐axon diffusion coefficients in brain white matter}},
  booktitle = {Magnetic},
  year = {2015},
  doi = {10.1002/mrm.25734},
}
Figure