Robustness of NeuroMark-Derived Functional Networks to fMRI Spatial Normalization Across the Human Lifespan

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Data Analysis

Background:

• NeuroMark is a hybrid independent component analysis (ICA) framework for extracting functional network features.
• It integrates spatial templates with spatially constrained ICA for data-driven decomposition.
• Current methods typically use direct spatial normalization to an adult template, raising questions about its optimality for non-adult populations.

Purpose of the Study:

• To evaluate the impact of two normalization strategies on NeuroMark's functional network features across different age groups.
• To determine if an age-specific normalization step improves feature stability compared to direct normalization to an adult template.

Main Methods:

• Compared direct normalization to an adult EPI template (EPInorm) with age-specific T1 template normalization followed by transformation to the adult EPI template (T1toEPInorm).
• Evaluated two strategies across three large datasets: infancy, development, and aging.
• Assessed spatial correspondence of intrinsic connectivity networks, similarity of individual-level time courses and spatial maps, and preservation of functional network connectivity (FNC) measures.

Main Results:

• Both EPInorm and T1toEPInorm yielded highly spatially correspondent intrinsic connectivity networks across all age cohorts (mean correspondence > 0.99).
• Individual-level functional network features showed high similarity, with time courses generally more consistent than spatial maps.
• Functional network connectivity (FNC) measures were robustly preserved across scans, regardless of the normalization strategy used (95% of FNC with r > 0.8670).

Conclusions:

• NeuroMark generates highly stable functional network features, irrespective of whether an age-specific normalization step is included.
• Direct normalization to the adult EPI template is a robust and efficient strategy for NeuroMark.
• NeuroMark with direct normalization facilitates harmonizable, large-scale, lifespan neuroimaging studies, ensuring broad dataset comparability without introducing template-related biases.