Tract shape modelling provides evidence of topological change in corpus callosum genu during normal ageing

Mark E. Bastin, Jakub P. Piatkowski, Amos J. Storkey, Laura J. E. Brown, Alasdair M. J. MacLullich, Jonathan D. ClaydenNeuroImage 2008 · DOI · OpenReview

TL;DR

This paper introduces a novel tractography method, Probabilistic Model-based Neighbourhood Tractography (PNT), that automatically and reliably segments white matter tracts in older adults. Using this method, the researchers found evidence of topological changes in the corpus callosum genu, suggesting a specific vulnerability of this brain region to normal aging.

What Problem Does This Paper Address?

Understanding how the brain changes with age is crucial for diagnosing and treating age-related neurological conditions. Diffusion Magnetic Resonance Imaging (dMRI) and tractography offer powerful tools to study white matter connectivity. However, a major challenge in tractography is the variability in results due to the subjective placement of "seed points" used to initiate the tract reconstruction. This makes it difficult to consistently identify the same white matter tracts across different individuals, especially in aging populations where brain atrophy can further complicate segmentation. Existing methods often require manual intervention or complex ROI definitions, which are time-consuming and can introduce bias. This paper aims to address this by developing an automated method for reliable tract segmentation in older adults.

Methods

The core innovation is the Probabilistic Model-based Neighbourhood Tractography (PNT). This method builds upon an earlier "neighbourhood tractography" (NT) approach. Here's how it works:

  • Tract Shape Modelling: PNT represents white matter tracts as B-spline curves defined by "knot points." The shape and length of these curves are then compared.
  • Probabilistic Framework: PNT uses probability distributions to model the variability in tract shape and length observed in a training dataset of younger individuals. This allows it to quantify how well a candidate tract matches a predefined "reference tract" (an idealized representation of the fasciculus).
  • Automatic Seed Point Placement: Instead of relying on manual seed point placement, PNT automatically identifies the seed point that generates the tract most closely matching the reference tract based on the probabilistic model. This is achieved by running tractography from multiple seed points within a neighborhood and selecting the one that yields the best "goodness-of-match" score (log-ratio, R).
  • Comparison Method: The researchers also compared PNT to a simpler "registration method" where seed points from a standard brain atlas are directly transferred to the subject's native space.

Datasets

  • Training Data: Tract data from fifteen younger healthy volunteers (mean age 31.5 ± 8.1 years) were used to train the tract shape models.
  • Evaluation Data: dMRI data from forty older volunteers (mean age 75.9 ± 4.9 years) with no history of stroke or neurological disorder were used to evaluate the PNT method.

Key Insights and Ideas

  • Automated Reliability: PNT significantly improves the reliability of tract segmentation in older adults compared to simple seed point registration, reducing the need for manual intervention.
  • Topological Change in Genu: The study provides evidence that the corpus callosum genu undergoes topological changes with age, as indicated by a negative correlation between the goodness-of-match score (R) and age. This suggests that the shape of this tract deviates more from its idealized form as individuals get older.
  • Frontal Lobe Vulnerability: The findings, particularly the specific changes observed in the genu (a frontal white matter structure), support the "frontal aging hypothesis," which suggests that frontal brain regions are disproportionately affected by aging.
  • The Log-Ratio (R) as a Metric: The log-ratio (R) proves to be a valuable, objective measure of how well a segmented tract matches a reference tract, correlating with anatomical plausibility.

Key Results

  • Acceptable Tracts: PNT successfully generated anatomically plausible tracts in 70-90% of subjects, a significant improvement over the registration method (2.5-60%).
  • Reduced Variability: PNT significantly reduced the inter-subject variability (coefficients of variation) of diffusion parameters (<D> and FA) compared to the registration method, indicating more consistent tract measurements.
  • Age Correlation in Genu: A significant negative correlation was found between the log-ratio (R) and age for the corpus callosum genu (p=0.04). A trend towards a significant negative correlation was also observed between tract-averaged fractional anisotropy (FA) and age (p<0.01).
  • Localization of Changes: These age-related changes were largely specific to the corpus callosum genu, with no significant correlations found for other investigated tracts.

Limitations and Future Work

  • Dependency on Tractography Quality: PNT's performance is inherently dependent on the quality of the underlying tractography data. Areas with complex white matter architecture or partial volume averaging (e.g., CCG, uncinate fasciculi) showed lower success rates.
  • Reference Tract Definition: The current method relies on a reference tract derived from a single individual. The authors are exploring ways to make reference tracts independent of specific datasets, for example, by using published human white matter atlases.
  • Unsupervised Learning: Future work will investigate "unsupervised learning" techniques to train the model using candidate tracts directly, rather than relying on a separate training dataset.
  • Tentative Topological Findings: The authors acknowledge that the findings regarding topological changes should be treated as tentative and require further replication.
  • Underlying Process: The precise underlying biological process driving the observed topological and microstructural changes in aging white matter remains unclear.

Summary updated

BibTeX
@article{bastin2008tract,
  author = {Mark E. Bastin and Jakub P. Piatkowski and Amos J. Storkey and Laura J. E. Brown and Alasdair M. J. MacLullich and Jonathan D. Clayden},
  title = {{Tract shape modelling provides evidence of topological change in corpus callosum genu during normal ageing}},
  journal = {NeuroImage},
  year = {2008},
  doi = {10.1016/j.neuroimage.2008.06.047},
}
Figure