Effects of diffusion MRI spatial resolution on human brain short-range association fiber reconstruction and structural connectivity estimation

Jialan Zheng, Ziyu Li, Wen Zhong, Ziang Wang, Zihan Li, Hongjia Yang, Mingxuan Liu, Xiaozhi Cao, Congyu Liao, S. David, Susie Y. Huang, Qiyuan TianbioRxiv 2025 · DOI

TL;DR

This study systematically investigates how the spatial resolution of diffusion MRI (dMRI) impacts the reconstruction of short-range association fibers (SAFs) and the estimation of brain connectivity. The findings reveal that lower resolution significantly degrades SAF reconstruction and connectivity estimates, with more advanced modeling techniques showing greater robustness. The paper provides practical recommendations for optimizing dMRI acquisition and analysis for SAF research.

What Problem Does This Paper Address?

Short-range association fibers (SAFs), also known as U-fibers, are crucial for communication between adjacent cortical areas. However, their small size (around 1.5mm thick layer of superficial white matter) and complex, gyration-following paths make them difficult to accurately map using conventional, lower-resolution dMRI. This limitation leads to an underestimation of their contribution to brain connectivity. As high-resolution dMRI techniques emerge, there's a need to systematically evaluate their impact on SAF reconstruction and connectivity estimation, and to understand how different tractography methods perform across various resolutions.

Methods

The researchers employed a comprehensive approach using both simulated and empirical dMRI data. They evaluated 20 different tractography pipelines, varying:

  • Fiber Modeling: Single-fiber (Diffusion Tensor Imaging - DTI) vs. multi-fiber (Constrained Spherical Deconvolution - CSD, including single-tissue SSST-CSD and multi-tissue MSMT-CSD).
  • Tracking Strategy: Deterministic vs. probabilistic tracking.
  • Post-processing: With or without Spherical-deconvolution Informed Filtering of Tractograms (SIFT) and Anatomical Constrain Tracking (ACT).

They quantitatively assessed the results using global (GSCS) and regional (RSCS) short-range connectivity strength metrics. The study also explored the effect of up-sampling lower-resolution data to a higher nominal resolution.

Datasets

  • Empirical Data: dMRI data from 20 healthy young adults acquired at three native resolutions: 0.96 mm, 1.5 mm, and 2 mm isotropic. This was achieved using the advanced gSlider sequence for sub-millimeter resolution and standard EPI sequences for lower resolutions.
  • Simulation Data: Data derived from the Human Connectome Project (HCP) WU-Minn consortium, down-sampled to simulate various resolutions (1.25 mm, 1.5 mm, 2 mm, 2.5 mm isotropic).

Key Insights and Ideas

  • Resolution is King for SAFs: The study provides strong quantitative evidence that higher spatial resolution is critical for accurately mapping SAFs and estimating their contribution to brain connectivity.
  • Advanced Models Matter: CSD-based multi-tissue models (MSMT-CSD) are significantly more robust to resolution changes than simpler DTI-based single-fiber models, highlighting the importance of accounting for crossing fibers.
  • SIFT and ACT Enhance Robustness: Incorporating SIFT and ACT into tractography pipelines significantly improves the reliability of SAF reconstruction and connectivity estimates across different resolutions.
  • Up-sampling is a Partial Fix: While up-sampling can partially mitigate the negative effects of lower resolution, it generally cannot fully recover the information lost compared to natively acquired high-resolution data.
  • Regional Variability: The impact of resolution is not uniform across the brain, with certain cortical regions (e.g., pars opercularis, precuneus, superior temporal) being more susceptible to resolution-induced errors.

Key Results

  • Decreased Connectivity at Lower Resolution: Lower spatial resolution consistently led to a significant decrease in Global Short-range Connectivity Strength (GSCS). For example, DTI models showed GSCS reductions of up to 13.1% when moving from 0.96 mm to 2 mm resolution.
  • Model Sensitivity: DTI single-fiber models were most sensitive to resolution changes, exhibiting larger GSCS decreases compared to CSD-based multi-fiber models.
  • Tracking Method Influence: Probabilistic tracking generally showed greater robustness to resolution decrease than deterministic tracking. Methods incorporating SIFT and ACT were more robust than those without.
  • Up-sampling Benefits: Data up-sampling generally increased GSCS, partially offsetting the reduction from lower native resolution, though it rarely restored GSCS to native high-resolution levels.
  • Visual Improvements: Higher resolution dMRI visually improved the representation of superficial white matter architecture, with clearer delineation of SAFs and sharper GM-WM boundaries.

Limitations and Future Work

  • Lack of Ground Truth: The study acknowledges the absence of a definitive "ground truth" for SAF reconstruction, necessitating validation studies using methods like tracer injections in post-mortem brains.
  • Healthy Adults Focus: The current study focused on healthy adults. Future research should investigate the sensitivity of SAF and connectivity biomarkers in patient populations (e.g., neurodegenerative diseases) and in children.
  • Tractography Algorithms: While commonly used algorithms were evaluated, future studies could incorporate emerging surface-based methods for a more comprehensive understanding of resolution effects.
  • Clinical Translation: The findings suggest that high-resolution dMRI is crucial for accurate SAF mapping, which could advance the study of brain connectivity in both neuroscience and clinical applications. The authors provide practical guidelines for optimizing acquisition and processing strategies.

Summary updated

Abstract

Abstract Short-range association fibers (SAFs) are critical for cortical communications but are often underestimated in conventional resolution diffusion magnetic resonance imaging (dMRI) since they locate within a ∼1.5mm thin layer of superficial white matter. With the emergence of high-resolution diffusion imaging techniques, this study timely evaluated the effects of image spatial resolution on SAF reconstruction using simulation data and multi-resolution (2, 1.5, and 0.96 mm iso.) empirical data acquired on the same 20 healthy subjects using the gSlider sequence and 20 widely used tractography approaches. Resolution effects were qualitatively assessed through model fitting and tractography results and quantitatively evaluated using global and regional short-range connectivity strength (SCS). It is found that lower resolution systematically reduces SCS across all methods in a spatially varying manner. Moreover, tractography methods exhibit significant differences in resolution sensitivity, with diffusion tensor imaging (DTI) based single-tissue single-fiber tractography showing greater vulnerability than constrained spherical deconvolution (CSD)-based multi-tissue multi-fiber tractography. Probabilistic tracking with anatomical constraints (ACT) and filtering (SIFT) improves robustness. Finally, up-sampling data to nominally higher resolution partially mitigates resolution-induced degradation and improves SAF reconstruction accuracy, particularly for DTI tractography. Based on these findings, higher resolution and multi-shell imaging is recommended if possible. For a given dataset, data up-sampling and DTI-based probabilistic tracking with ACT is recommended for single-shell low b-value data. CSD-based probabilistic tracking with ACT and SIFT is recommended for single-shell higher b-value data and multi-shell data. In summary, this study systematically and quantitatively evaluated resolution effects on SAF reconstruction and structural connectivity estimation and provided practical guidelines for more accurate mapping of SAFs. These advances hold promises to improve the characterization of healthy and diseased human brains in a wide range of neuroscientific and clinical applications.
BibTeX
@misc{zheng2025effects,
  author = {Jialan Zheng and Ziyu Li and Wen Zhong and Ziang Wang and Zihan Li and Hongjia Yang and Mingxuan Liu and Xiaozhi Cao and Congyu Liao and S. David and Susie Y. Huang and Qiyuan Tian},
  title = {{Effects of diffusion MRI spatial resolution on human brain short-range association fiber reconstruction and structural connectivity estimation}},
  year = {2025},
  doi = {10.1101/2025.06.04.657810},
  howpublished = {bioRxiv},
}
Figure