TL;DR
This paper introduces a novel method for quantitatively analyzing brain growth in early childhood using deformation-based morphometry (DBM) on longitudinal MRI data. By creating population-averaged atlases and employing advanced registration techniques, the authors provide a robust framework for understanding neurodevelopmental trajectories in infants, particularly those born preterm.
What Problem Does This Paper Address?
Understanding brain development in the first two years of life is crucial, as this period sees rapid growth and significant changes. However, studying this early period presents challenges: traditional cross-sectional studies are confounded by inter-subject variability, and existing neuroinformatic tools struggle with the rapid anatomical changes and intensity variations in immature brains. This paper aims to address this by developing a robust method for analyzing longitudinal brain growth in infants, focusing on preterm infants who are particularly vulnerable to developmental disruptions.
Methods
The core of this paper lies in its advanced image registration and analysis techniques. The authors employ Deformation-Based Morphometry (DBM), which uses the Jacobian determinant of deformation fields to quantify regional volumetric changes.
Here's a breakdown of their approach:
- Longitudinal Registration: For each subject, baseline (1-year) and follow-up (2-year) MRI scans are registered using a two-step process:
- Global Affine Registration: A 12-parameter affine transformation accounts for overall shape changes.
- Non-Rigid Free-Form Deformation (FFD): A B-spline-based FFD model captures local, detailed deformations.
- Jacobian Determinant Calculation: The Jacobian determinant of the combined transformation provides a voxel-wise measure of local volume change, indicating growth or shrinkage.
- Atlas Construction: To study population-level growth, the authors create "average space" atlases. This involves:
- Cross-sectional Registration: Subjects within each timepoint are registered to a reference subject.
- Transformation Averaging: The average of these inter-subject transformations is computed to create a common "average space." This process is designed to be robust to the choice of reference subject.
- Atlas Generation: Anatomical images, growth maps (Jacobian determinants), and tissue probability maps are then averaged in this common space to create population atlases for each timepoint.
- Tissue Segmentation: For validation, a tissue segmentation algorithm (EM-based) is used to classify grey matter, white matter, and CSF. This allows for an alternative estimation of tissue volumes and growth factors.
- Growth Estimation: Growth factors are estimated both from individual registrations (using Jacobian determinants) and from tissue segmentation (ratio of volumes). These are then compared.
Datasets
The study utilizes longitudinal MRI data from 25 preterm born children. These children were scanned at two timepoints: approximately 1 year of age and 2 years of age. The gestational age at birth averaged 27.7 weeks. The paper does not introduce new benchmark datasets.
Key Insights and Ideas
- DBM for Infant Brain Growth: The paper demonstrates the efficacy of DBM, particularly using Jacobian determinants, for quantifying subtle but significant brain growth in very young children, a task challenging for traditional methods like VBM.
- Population Atlases for Development: Creating average space atlases for anatomy, growth, and tissue classes at different timepoints provides a powerful tool for visualizing and analyzing population-level neurodevelopmental trajectories.
- Robustness to Reference Choice: The method for averaging transformations to create population atlases is shown to be robust, meaning the resulting atlases are not overly biased by the selection of a single reference subject.
- Validation of DBM: The strong agreement between growth estimates derived from DBM and those from tissue segmentation provides crucial validation for the DBM approach in this age group.
- Preterm Infant Development: The study highlights the potential of this methodology to characterize brain development in vulnerable populations like preterm infants, where early disruptions can have long-term consequences.
Key Results
- Consistent Growth Estimates: The study shows good agreement between growth factor estimates derived from registration-based DBM and those obtained from tissue segmentation for both grey matter (GM) and white matter (WM).
- Quantitative Growth: For the exemplar group of 25 ex-preterm infants, grey matter showed an increase of approximately 10-12%, while white matter growth was around 25-30% between 1 and 2 years of age.
- Low Negative Jacobian Determinants: The DBM approach, with appropriate regularization, generated very few voxels with negative Jacobian determinants (indicating volume loss), suggesting the method is stable for capturing growth.
- Tissue Volume Changes: Mean grey matter volume increased from 665.7 cm³ to 746.7 cm³, and white matter volume increased from 198.7 cm³ to 254.8 cm³ in the cohort between ages 1 and 2.
Limitations and Future Work
- White Matter Variability: The authors note higher variability in white matter growth estimates compared to grey matter, potentially due to its smaller volume, changes in myelination, and partial volume effects.
- Term-Born Controls: Due to ethical considerations regarding sedation, the study primarily focused on preterm infants and did not include term-born controls. The authors suggest their techniques are transferable to such data.
- Specific Region Analysis: While the current study provides global and tissue-class level growth estimates, future work aims to extend the method to analyze growth in more specific brain structures by combining DBM with advanced structural segmentation techniques.
- Exploring Deformation Field Properties: The authors suggest that further analysis of the longitudinal deformation fields (e.g., using curl or divergence) could reveal more detailed patterns of development.
- Alternative Atlas Construction: Investigating different approaches to atlas creation, beyond the one presented, could be a fruitful area for future research.
Summary updated