Fixel-based analysis of the preterm brain: Disentangling bundle-specific white matter microstructural and macrostructural changes in relation to clinical risk factors

Diliana Pecheva, Jacques‐Donald Tournier, Maximilian Pietsch, Daan Christiaens, Dafnis Batallé, Daniel C. Alexander, Joseph V. Hajnal, A. David Edwards, Hui Zhang, Serena J. CounsellNeuroImage 2019 · DOI

TL;DR

This study introduces Fixel-Based Analysis (FBA) to examine white matter development in preterm infants, revealing that clinical risk factors like prematurity and ventilation duration impact specific fiber bundles rather than the brain uniformly. This advanced technique offers a more nuanced understanding of how these factors contribute to both the microscopic structure and macroscopic organization of white matter tracts.

What Problem Does This Paper Address?

Preterm birth, affecting a significant portion of newborns, is linked to a wide range of developmental impairments, including cognitive, motor, and language delays. These outcomes are often associated with various perinatal risk factors such as prematurity itself, the need for mechanical ventilation, and nutritional support. Traditional diffusion MRI (dMRI) techniques, particularly diffusion tensor imaging (DTI), have been instrumental in identifying white matter abnormalities in preterm infants. However, DTI models simplify complex white matter architecture by assuming a single dominant fiber direction within each voxel. This simplification can obscure the specific impact of risk factors on individual fiber bundles, especially in regions with crossing fibers, which are common in the developing brain. This paper addresses this limitation by employing Fixel-Based Analysis (FBA), a more sophisticated dMRI technique that can disentangle and quantify changes within individual fiber populations, providing a finer-grained understanding of how perinatal insults affect white matter development.

Methods

The core innovation in this study is the application of Fixel-Based Analysis (FBA). FBA builds upon advanced diffusion modeling techniques like multi-shell multi-tissue constrained spherical deconvolution (CSD) to estimate the fiber orientation density function (ODF) for each voxel. From these ODFs, FBA extracts "fixels," which represent individual fiber populations within a voxel. The study then quantifies two key fixel-based measures:

  • Apparent Fibre Density (FD): This measures the intra-axonal volume of a specific fiber population, reflecting its microscopic integrity. Reduced FD can indicate fewer or smaller axons.
  • Fibre Cross-Section (FC): This quantifies the macroscopic size or cross-sectional area of a fiber bundle. It's derived from the Jacobian determinant of the registration warp, reflecting local expansion or contraction of the tract.

These measures were then statistically analyzed to assess their relationship with various perinatal risk factors. The study also utilized standard dMRI acquisition protocols and image processing pipelines, including motion correction, denoising, and bias field correction, before applying FBA.

Datasets

The study utilized diffusion MRI (dMRI) data from 50 preterm infants (28 male) born between 24.0 and 32.9 weeks gestational age (GA) and scanned at term equivalent age (TEA), specifically between 38.6 and 47.1 weeks postmenstrual age (PMA). The data was acquired on a 3 Tesla Philips MRI scanner. The study did not introduce a new benchmark dataset but rather applied its novel analysis method to an existing cohort.

Key Insights and Ideas

  • Fiber-Specific Impact: The most significant insight is that perinatal risk factors don't uniformly affect white matter. Instead, they have specific impacts on individual fiber bundles, even within complex regions of crossing fibers. This highlights the limitations of voxel-wise analyses that average across multiple fiber populations.
  • Microstructure vs. Macrostructure: The study effectively disentangles the effects on microscopic (FD) and macroscopic (FC) white matter organization. For instance, while some factors affected both, others primarily impacted FC, suggesting that changes in tract size might be a more prominent consequence of certain perinatal insults than changes in axon density.
  • Maturation Matters: The study confirms that white matter development continues significantly from preterm birth to term equivalent age, with PMA (age at scan) being positively correlated with all FBA measures (FD, FC, FDC) across various tracts. This underscores the critical window of development during this period.
  • Clinical Relevance: The findings directly link specific clinical risk factors (gestational age, ventilation duration, TPN, birthweight) to distinct white matter alterations, providing a more precise understanding of the biological mechanisms underlying neurodevelopmental risks in preterm infants.

Key Results

  • Gestational Age (GA): Higher GA at birth was positively correlated with FD, FC, and FDC across numerous white matter tracts, including commissural fibers (corpus callosum, anterior commissure) and association fibers.
  • Mechanical Ventilation: Longer duration of mechanical ventilation was negatively correlated with all three FBA measures (FD, FC, FDC). Notably, FBA could pinpoint that FC in the corticospinal tract was affected by ventilation duration, while association fibers in the same region were not.
  • Total Parenteral Nutrition (TPN): TPN duration showed a negative correlation with FC in specific tracts like the corticospinal tract and cerebellum. However, this correlation was not significant after removing a single outlier.
  • Birthweight: Birthweight z-scores were positively correlated with FC and FDC, particularly in tracts like the corpus callosum and corticospinal tract, but not with FD.
  • Sex Differences: While males showed higher FC in several tracts (cingulum, superior longitudinal fasciculus), these differences were not observed when controlling for brain volume. No significant differences in FD were found between sexes.

Limitations and Future Work

The authors acknowledge the relatively small study group size (n=50) and suggest that larger studies incorporating neurodevelopmental outcome measures in later childhood are needed. They also note that while FBA provides more detailed insights, it is still limited by the quality of dMRI data and the accuracy of the registration and modeling techniques.

Future work could involve extending FBA to multivariate analyses to better account for the interdependencies between various risk factors and brain volume. Investigating the interplay between FBA findings and cortical development could also provide a more comprehensive understanding of the anomalous grey and white matter development observed in preterm infants. The authors also suggest that FBA's ability to identify specific tract vulnerabilities could aid in anticipating future developmental impairments.

Summary updated

Abstract

Diffusion MRI (dMRI) studies using the tensor model have identified abnormal white matter development associated with perinatal risk factors in preterm infants studied at term equivalent age (TEA). However, this model is an oversimplification of the underlying neuroanatomy. Fixel-based analysis (FBA) is a novel quantitative framework, which identifies microstructural and macrostructural changes in individual fibre populations within voxels containing crossing fibres. The aim of this study was to apply FBA to investigate the relationship between fixel-based measures of apparent fibre density (FD), fibre bundle cross-section (FC), and fibre density and cross-section (FDC) and perinatal risk factors in preterm infants at TEA. We studied 50 infants (28 male) born at 24.0-32.9 (median 30.4) weeks gestational age (GA) and imaged at 38.6-47.1 (median 42.1) weeks postmenstrual age (PMA). dMRI data were acquired in non-collinear directions with b-value 2500 s/mm<sup>2</sup> on a 3 Tesla system sited on the neonatal intensive care unit. FBA was performed to assess the relationship between FD, FC, FDC and PMA at scan, GA at birth, days on mechanical ventilation, days on total parenteral nutrition (TPN), birthweight z-score, and sex. FBA reveals fibre population-specific alterations in FD, FC and FDC associated with clinical risk factors. FD was positively correlated with GA at birth and was negatively correlated with number of days requiring ventilation. FC was positively correlated with GA at birth, birthweight z-scores and was higher in males. FC was negatively correlated with number of days on ventilation and days on TPN. FDC was positively correlated with GA at birth and birthweight z-scores, negatively correlated with days on ventilation and days on TPN and higher in males. We demonstrate that these relationships are fibre-specific even within regions of crossing fibres. These results show that aberrant white matter development involves both microstructural changes and macrostructural alterations.
BibTeX
@inproceedings{pecheva2019fixel,
  author = {Diliana Pecheva and Jacques‐Donald Tournier and Maximilian Pietsch and Daan Christiaens and Dafnis Batallé and Daniel C. Alexander and Joseph V. Hajnal and A. David Edwards and Hui Zhang and Serena J. Counsell},
  title = {{Fixel-based analysis of the preterm brain: Disentangling bundle-specific white matter microstructural and macrostructural changes in relation to clinical risk factors}},
  booktitle = {NeuroImage},
  year = {2019},
  doi = {10.1016/j.nicl.2019.101820},
}
Figure