Resting-state functional MRI (rs-fMRI) is widely used to study intrinsic brain organization and its alterations in psychiatric disorders such as schizophrenia. Traditional rs-fMRI workflows often rely on volumetric normalization and fixed atlas parcellations derived from neurotypical populations. While these approaches facilitate group-level comparisons, they may obscure individual variation in cortical organization and intrinsic dynamics.
This study compared analytic choices across preprocessing and parcellation strategies to determine how they influence both static functional connectivity and dynamic measures based on quasi-periodic patterns (QPPs) in schizophrenia. Two independent cohorts were analyzed to assess reproducibility and sensitivity of findings.
Analyses were performed on two independent schizophrenia datasets: MRI site 1 (n=159) and MRI site 2 (n=255). Four preprocessing and parcellation strategies were evaluated, including at least one individualized surface-based parcellation pipeline (referred to here as IndiPar) and multiple atlas-based pipelines. The pipelines were compared on measures of static functional connectivity and dynamic QPP-derived metrics.
Dynamic QPP metrics included assessment of default mode–dorsal attention network opposition, QPP component rank, explained variance attributable to QPPs, and QPP event rate. The study also examined associations between these rs-fMRI measures and clinical symptom severity measured with PANSS.
Across both cohorts, the individualized surface-based parcellation (IndiPar) approach consistently yielded stronger and more pronounced dynamic signatures than atlas-based approaches. IndiPar better preserved subject-specific cortical organization compared with fixed atlas parcellations derived from neurotypical templates. This preservation of individual variability translated into greater sensitivity for detecting intrinsic dynamics and patient-control differences.
The authors report that IndiPar detected more prominent QPP dynamics, stronger default mode/dorsal attention opposition, and a larger fraction of explained variance attributable to QPPs compared with atlas-based pipelines. IndiPar also produced larger and more reproducible differences between patients and controls in measures derived from both static functional connectivity and QPP event-rate analyses.
Parcellation choice influenced the magnitude and reproducibility of patient-control differences in static functional connectivity. IndiPar produced larger effect sizes and more reproducible connectivity differences across the two datasets relative to atlas-based pipelines. The study highlights that maintaining subject-specific cortical topography during parcellation can improve detection of connectivity alterations associated with schizophrenia.
IndiPar consistently enhanced characterization of QPP-related dynamics. Specific observations included:
In the larger cohort (MRI site 2, n=255), IndiPar detected a significantly increased QPP event rate in patients relative to controls and identified a greater number of associations between QPP measures and symptom severity.
The study examined relationships between rs-fMRI measures and PANSS symptom severity. Although IndiPar revealed more symptom associations in the larger cohort, the authors report limited stability of these associations across the two independent datasets. This limited cross-dataset replication underscores challenges in deriving robust brain–symptom relationships from heterogeneous clinical samples and suggests that symptom associations may be sensitive to cohort-specific factors or analytic choices.
Findings indicate that preprocessing and parcellation choices substantially influence rs-fMRI results in schizophrenia research. An individualized surface-based parcellation appears to better preserve subject-specific variability and improves detection of both static connectivity alterations and intrinsic dynamic patterns such as QPPs. These improvements may enhance sensitivity for detecting patient-control differences and, in some cases, reveal additional symptom associations.
However, the limited replication of symptom associations between cohorts highlights ongoing challenges: heterogeneous psychiatric populations and dataset-specific factors can reduce the stability of brain–symptom links. The source reports these cohort replication issues but does not provide further detail on potential moderators or remedies beyond the parcellation comparison.
The authors have made data and code available at the project repository: https://github.com/Dalimear/Schizophrenia-qpp-cPCA-dynamics. Funding included ERDF-Project BRAINSCAPE (CZ.02.01.01/00/23_020/0008560). The authors declared no competing interests.
This study demonstrates that individualized surface-based parcellation enhances characterization of resting-state brain dynamics in schizophrenia relative to atlas-based pipelines. IndiPar improved detection of QPP dynamics, default mode–dorsal attention opposition, explained variance, and patient-control differences in both static and dynamic measures. Nonetheless, limited cross-dataset stability of symptom associations emphasizes the challenge of establishing reproducible brain–symptom relationships in heterogeneous psychiatric cohorts.