[Paper] Cross-Population White Matter Atlas Creation for Concurrent Mapping of Brain Connections in Neonates and Adults with Diffusion MRI Tractography

Source: arXiv - 2512.20370v1

Overview

A new diffusion‑MRI white‑matter (WM) atlas—NABA (Neonatal/Adult Brain Atlas)—brings together brain‑tract data from newborns and adults into a single, common space. By doing so, researchers can directly compare how WM pathways mature from birth through adulthood, opening doors to earlier detection of neurodevelopmental disorders and more precise brain‑imaging pipelines.

Key Contributions

• Unified cross‑population WM atlas built from thousands of tractography streamlines spanning neonates (including pre‑term) and adults.
• Data‑driven fiber clustering pipeline that tolerates the huge anatomical variability between newborn and mature brains.
• Standardized WM parcellation template enabling one‑click mapping of the same tract across age groups.
• Comprehensive validation: feasibility of joint mapping, developmental trajectories, sex differences, and pre‑term effects all demonstrated on the same atlas.
• Open‑source resources (code & atlas files) released for the neuro‑imaging community, ready to plug into existing dMRI processing stacks (e.g., DIPY, MRtrix3, FSL).

Methodology

1. Data collection – Diffusion MRI scans from two large cohorts: (a) neonates (0–2 months, including pre‑term) and (b) healthy adults.
2. Pre‑processing – Standard denoising, motion correction, and bias‑field removal; tensors and higher‑order models (e.g., constrained spherical deconvolution) were fitted to obtain fiber orientation distributions.
3. Whole‑brain tractography – Probabilistic streamlines were generated for each subject (≈1 M fibers).
4. Fiber clustering – A two‑stage, data‑driven algorithm:
   • Within‑subject clustering groups similar streamlines using a spectral‑embedding distance metric.
   • Across‑subject consensus clustering aligns clusters from neonates and adults into a common set of 72 canonical tracts.
5. Atlas construction – The consensus clusters are warped into a shared anatomical template (MNI‑like for adults, neonatal‑adjusted for infants) and stored as probabilistic tract masks.
6. Validation – Tract‑wise fractional anisotropy (FA) trajectories were extracted and compared across age, sex, and birth‑status groups.

Why it matters for developers: The pipeline is modular, relies on open‑source libraries, and can be scripted end‑to‑end (Docker‑compatible). Plug‑and‑play with existing dMRI workflows is possible without re‑training deep‑learning models.

Results & Findings

These patterns line up with known neurodevelopmental milestones (e.g., early motor pathway maturation) and provide quantitative baselines for future clinical studies.

Practical Implications

• Accelerated biomarker discovery – Clinicians can now compare a newborn’s tract FA directly against adult reference values, flagging atypical development earlier.
• Standardized preprocessing – Development teams building neuro‑imaging platforms can embed NABA as a default WM parcellation, reducing the need for population‑specific atlases.
• Cross‑age machine‑learning models – Training a single model to predict outcomes (e.g., neurodevelopmental scores) across ages becomes feasible because the input feature space (tract‑wise FA) is consistent.
• Educational tools – Interactive visualizations of the same tract at birth vs. adulthood can be integrated into VR/AR training modules for neurosurgeons or radiologists.
• Regulatory & data‑sharing – A unified atlas simplifies multi‑site studies, as data from neonatal NICUs and adult hospitals can be harmonized under a common coordinate system.

Limitations & Future Work

• Resolution mismatch – Neonatal scans have lower spatial resolution; some fine tracts may be under‑represented.
• Population bias – The adult cohort is predominantly young‑to‑middle‑aged; extending to older adults will be needed for lifespan studies.
• FA‑only focus – While FA is a convenient microstructural metric, incorporating additional measures (e.g., neurite orientation dispersion, myelin water fraction) could enrich the atlas.
• Longitudinal validation – Future work should track the same infants over months/years to confirm that cross‑sectional trajectories hold longitudinally.

Bottom line: NABA bridges a critical gap between neonatal and adult diffusion MRI, giving developers a ready‑to‑use, cross‑population white‑matter map that can power the next generation of brain‑development analytics.

Authors

• Wei Zhang
• Yijie Li
• Ruixi Zheng
• Nir A. Sochen
• Yuqian Chen
• Leo R. Zekelman
• Ofer Pasternak
• Jarrett Rushmore
• Yogesh Rathi
• Nikos Makris
• Lauren J. O’Donnell
• Fan Zhang

Paper Information

• arXiv ID: 2512.20370v1
• Categories: cs.NE
• Published: December 23, 2025
• PDF: Download PDF