TL;DR
This paper introduces the first population-specific brain charts for China, revealing distinct neurodevelopmental trajectories compared to Western cohorts. These new charts significantly improve the accuracy of predicting healthy brain development and identifying neurological disorders in Chinese individuals, highlighting the critical need for diverse, representative brain atlases in precision medicine.
What Problem Does This Paper Address?
Current lifespan brain charts, crucial for understanding neurodevelopment and identifying brain disorders, are predominantly based on European and North American populations. This "Western-centric" bias leads to a severe underrepresentation of Asian populations, including China, which constitutes 18% of the world's population. This lack of representation raises critical questions about the universality of existing brain charts and their clinical utility for non-Western populations. Previous studies have shown ethnic disparities in brain structure and developmental trajectories, suggesting that a one-size-fits-all approach is inadequate. This paper aims to address this gap by creating the first comprehensive, population-specific brain growth charts for Chinese individuals.
Methods
The study employed a rigorous, multi-stage approach:
- Data Aggregation and Quality Control: Large-scale structural MRI datasets from across China (Chinese Lifespan Brain Mapping Consortium, C-LBM) and Western countries were aggregated. A unified, four-stage quality control (QC) pipeline, combining automated algorithms and expert visual inspection, was applied to all structural MRI images to ensure data consistency and reliability.
- Structural MRI Data Processing: Age-specific processing strategies were used, with a harmonized pipeline for individuals aged two years and older, utilizing FreeSurfer v6.0.0. For younger participants (0-2 years), specialized pipelines (iBEAT v2.0 and dHCP) optimized for early-age neuroimaging were employed.
- Phenotype Quantification: A comprehensive set of 296 structural brain phenotypes were quantified, including global brain measures (e.g., total intracranial volume, cortical GMV), subcortical regional volumes (e.g., caudate, hippocampus), and cortical regional measures (e.g., thickness, GMV, surface area, folding index) across 68 cortical regions.
- Normative Growth Curve Modeling: Generalized Additive Models for Location, Scale and Shape (GAMLSS) were used to model nonlinear growth patterns across the lifespan. The generalized gamma (GG) distribution was found to be most suitable for brain structural data. Fractional polynomial functions were employed for age-related smoothing.
- Cross-Population Comparison: Permutation tests were used to rigorously compare peak age differences in maturation timelines between the Chinese and Western models.
- Predictive Accuracy and Clinical Relevance: The predictive accuracy of both Chinese and Western models was evaluated on independent Chinese datasets. The models' ability to detect pathological deviations in clinical cohorts (Alzheimer's disease, MCI, schizophrenia, MDD) was also assessed.
- Sensitivity Analyses: Multiple sensitivity analyses (bootstrap resampling, age-matched sampling, leave-one-site-out, split-half replication, imaging quality control, and ethnic subsample validation) were conducted to ensure the robustness and reproducibility of the findings.
Datasets
- Chinese Lifespan Brain Mapping (C-LBM) Consortium (Phase I): 43,037 healthy participants (aged 0-100 years) from 384 sites across 29 provinces in China. 98.5% self-identified as Han Chinese.
- Western Lifespan Neuroimaging Dataset: 56,339 healthy participants (aged 0-100 years) from 174 scanner sites across eight predominantly Western countries (United States, United Kingdom, Canada, Ireland, Netherlands, Belgium, Switzerland, Australia). 87.8% self-identified as white.
- Chinese Case-Control Datasets: Used for evaluating predictive accuracy and clinical relevance. This included healthy controls (N=2,591) and patients with Alzheimer's disease (N=399), mild cognitive impairment (N=283), schizophrenia (N=691), and major depressive disorder (N=1,492).
Key Insights and Ideas
- Neurodevelopmental Divergence: The study definitively shows that brain development is not universal. Chinese and Western populations exhibit distinct timelines for maturation, challenging the assumption that Western-derived charts can be applied globally.
- The Importance of "Normative" Context: What is considered "normal" brain development is population-specific. Using Western charts for Chinese individuals can lead to misestimations of both typical development and pathological deviations.
- Cerebellar Acceleration: A surprising finding is the accelerated growth of the cerebellum in the Chinese population compared to Western cohorts.
- Hierarchical Cortical Development: The study confirms a hierarchical pattern in cortical development, with sensorimotor regions maturing earlier than paralimbic regions, but highlights population-specific nuances in this timing.
- Need for Diverse Data: The findings strongly advocate for the creation of diverse, ethnically representative neuroimaging datasets to advance equitable precision neuroscience and improve clinical validity worldwide.
Key Results
- Population-Specific Charts: The study successfully constructed the first population-specific brain growth charts for China, covering 296 structural brain phenotypes across the lifespan.
- Distinct Maturation Timelines: Chinese individuals showed prolonged cortical and subcortical maturation, accelerated cerebellar growth, and earlier development of sensorimotor regions compared to Western cohorts.
- Improved Predictive Accuracy: The Chinese-specific models demonstrated significantly higher out-of-sample predictive accuracy for healthy Chinese individuals compared to Western models (Cohen's d = 0.66, mean directional error = 0.84). At least a 10% improvement was observed for 41 phenotypes, and at least 5% for 94 phenotypes.
- Reduced Misestimation in Clinical Cohorts: Chinese-specific models significantly reduced the misestimation of deviation scores in Chinese patients with AD, MCI, SCZ, and MDD, identifying more extreme deviations and providing a more accurate characterization of brain structural variations.
- Robustness: Extensive sensitivity analyses confirmed the robustness and reproducibility of the findings.
Limitations and Future Work
- Fetal Stage Exclusion: The study did not include the fetal stage due to limited availability of comparable neuroimaging data, especially for Chinese populations. Future work aims to incorporate fetal data.
- Focus on Chinese and Western Populations: The study primarily compared Chinese and Western populations. Future research should expand to include more diverse racial and ethnic groups to develop globally representative normative charts.
- Sample Acquisition Challenges: While efforts were made to collect data across Chinese provinces, sample acquisition in certain regions with limited scanners remains a challenge. Future work aims to include more geographically comprehensive datasets.
- Lack of Complementary Information: Most datasets used in the study lacked detailed information on cognitive performance, socioeconomic status, environmental exposure, and genetics. Future prospective, multimodal cohort studies are needed to investigate how these factors shape lifespan growth trajectories.
- Cross-Sectional Data: The study is based on cross-sectional data, which may underestimate age-related changes. Longitudinal data integration is essential for more accurate capture of brain growth trajectories.
Summary updated