Influence of Group‑Level Parcellations on Resting-State fMRI Connectome–Cognition Correlations In Brain Tumor Patients
Abstract
Purpose
Investigate the association between neurocognitive function (NCF) and functional connectomics derived from commonly used group-level brain parcellations on patients with brain tumors.
Methods
Presurgical resting-state (rs) functional MRI (fMRI), T1-weighted and FLAIR MR images from a dataset of 87 patients (mean age: 50 ± 15 years; 43 Male, 44 Female) with brain tumors were analyzed. Automated Anatomical Labelling Atlas 3 (AAL3), Glasser Atlas and Schaefer 200-region atlas were registered to patient space with SPM12. For each patient and the respective atlases, Fisher’s Z-transform was applied to the Pearson correlation of the average timeseries between parcels to obtain the functional connectivity matrix for calculating functional connectomics. Density thresholds of 0.1, 0.2 and 0.3 were applied to the functional connectivity matrix. Preoperative NCF measures standardized to demographically-adjusted normative data were converted to ordinal measures including normal limits and 3 levels of impairment severity for analysis. Spearman correlation was calculated between functional connectomics and NCF measures at each density threshold.
Results
For AAL3, statistically significant (p < 0.05) correlations were found for connectomic measures and NCF measures of language, diverse executive functions, and encoding, retrieval and consolidation episode memory processes. For the Glasser atlas, statistically significant correlations were found for connectomic measures and NCF measures of processing speed and verbal abstract reasoning aspects of executive function. For the Schaefer atlas, significant correlations were found for connectomic measures and only NCF measures of processing speed.
Conclusion
This work underscores how different group-level parcellations can influence the discovery of associations between functional connectomics and NCF measures. Despite lesion-induced anatomical distortions, group-level parcellations are commonly used in functional connectomics studies of patients with brain lesions. Knowledge pertaining to the influence of group-level parcellations on functional connectomics and their association with NCF is important to establish prediction models of NCF outcomes for the brain tumor patients.