Ultrafine brain intrinsic connectivity networks template via very-high-order independent component analysis of large-scale resting-state functional magnetic resonance imaging data

Shiva Mirzaeian, Kyle M. Jensen, Ram Ballem, Pablo Andrés-Camazón, Jiayu Chen, Vince D. Calhoun, Armin IrajiFrontiers in Neuroscience 2025 · DOI

Abstract

Spatial group independent component analysis (sgr-ICA) is a cornerstone of resting-state fMRI (rsfMRI) research for identifying intrinsic connectivity networks (ICNs). While lower-order decompositions reveal large-scale networks, higher-order models provide finer granularity but have historically been limited by sample size constraints. This study applies sgr-ICA with 500 components to a massive dataset of over 100,000 subjects to generate "NeuroMark-500," a robust, fine-grained ICN template. The template reveals schizophrenia-related dysconnectivity patterns—particularly in cerebellar and paralimbic regions—that are undetectable at coarser spatial scales.

Introduction: The Need for High-Resolution Brain Maps

Functional Magnetic Resonance Imaging (fMRI) measures blood flow to map brain communication. Researchers typically use Independent Component Analysis (ICA) to group brain regions into functional networks. Traditional "low-order" models (20–50 networks) provide a broad overview, while "high-order" models (up to 200 networks) offer more detail. However, pushing toward "very-high-order" ICA (500 components) has been challenging due to the need for immense datasets to ensure stability. This research aims to capture "ultrafine" granularity to detect subtle disease-related patterns that coarser maps might overlook.

Methodology and Data Processing

The researchers leveraged a staggering aggregate of over 100,517 subjects from more than 20 resting-state fMRI datasets. After rigorous quality control, 57,709 subjects were selected for the final analysis.

The analysis pipeline involves preprocessing 100,512 subjects, quality control, and a multi-stage sgr-ICA process to identify 131 stable ICNs.

The core methodology employed spatial group-level ICA (sgr-ICA) with a model order of 500. To manage computational demands, a multi-stage dimensionality reduction approach using Principal Component Analysis (PCA) was implemented. Stability was ensured through ICASSO indices and split-half reproducibility tests across 50 iterations. From the initial 500 components, 131 highly reliable ICNs were identified to form the NeuroMark-500 template.

Results: Fine-Grained Insights into Brain Function

The very-high-order ICA revealed a "branching" phenomenon, where large networks from lower-order models split into functionally distinct sub-units. For instance, the cerebellum, often treated as a single entity, was found to contain 27 distinct networks, highlighting its complex role in motor and cognitive integration. Similarly, the 500-order model successfully isolated the hippocampus from its neighbors, which is critical for accurately characterizing its role in psychiatric disorders.

Visualization of the 131 ICNs across seven brain domains and their corresponding spatial stability metrics.

Clinical Application in Schizophrenia

The NeuroMark-500 template was evaluated using the BSNIP dataset (502 individuals with schizophrenia and 640 typical controls). The high-resolution template detected significant dysconnectivity patterns missed by lower-order models:

  • Hypoconnectivity: Reduced connectivity between the cerebellum and subcortical domains (basal ganglia and thalamus).
  • Hyperconnectivity: Increased connectivity between the cerebellum and visual, sensorimotor, and higher cognitive domains.

These findings demonstrate that "ultrafine" networks are not only stable (with stability indices exceeding 0.80) but also offer superior sensitivity for identifying psychiatric biomarkers. The NeuroMark-500 template detected significant connectivity differences that were entirely missed by the previous state-of-the-art template, NeuroMark 2.2.

Discussion and Future Directions

While NeuroMark-500 provides unprecedented resolution, this study focused on static functional network connectivity. Future research should apply this 500-component resolution to dynamic connectivity to explore how these ultrafine networks interact over time. Additionally, further validation is required to determine the clinical utility of these markers for predicting individual patient outcomes. Despite the high computational cost, this template serves as a valuable reference for the neuroimaging community, potentially enhancing the clinical utility of rsfMRI in psychiatric research.

Summary updated

Abstract

Spatial group independent component analysis (sgr-ICA) is widely used in resting-state fMRI to identify intrinsic connectivity networks (ICNs). While lower-order decompositions reveal large-scale networks, higher-order models provide finer granularity but have been limited by small sample sizes. In this study, we applied sgr-ICA with 500 components to more than 100,000 subjects with rsfMRI to generate a robust fine-grained ICN template. Using this template, we examined whole brain functional network connectivity (FNC) in 502 individuals with schizophrenia and 640 typical controls and compared the findings with a lower order multiscale template. The 500-component template yielded a large set of reliable ICNs, particularly in the cerebellar and paralimbic regions, and revealed schizophrenia-related dysconnectivity patterns that were not detected at larger spatial scales. Specifically, we observed hypoconnectivity between the cerebellar and subcortical domains (basal ganglia and thalamus) and hyperconnectivity between the cerebellar domain and the visual, sensorimotor and higher cognitive domains. These results demonstrate that very high-order ICA can capture distinct fine-grained ICNs, improving the detection of disease-related connectivity differences and enriching current multiscale ICN templates. The derived ICNs can serve as a valuable reference for future studies and potentially enhance the clinical utility of rsfMRI in psychiatric research.
BibTeX
@inproceedings{mirzaeian2025ultrafine,
  author = {Shiva Mirzaeian and Kyle M. Jensen and Ram Ballem and Pablo Andrés-Camazón and Jiayu Chen and Vince D. Calhoun and Armin Iraji},
  title = {{Ultrafine brain intrinsic connectivity networks template via very-high-order independent component analysis of large-scale resting-state functional magnetic resonance imaging data}},
  booktitle = {Frontiers in Neuroscience},
  year = {2025},
  doi = {10.3389/fnins.2025.1672129},
}
Figure