Surface- and volume-based functional connectivity in the first episode psychosis patients

Abstract

INTRODUCTION: Functional connectivity analysis has an important tool in exploring the brain dynamics in both healthy and patient populations (Caputi et al., 2021). The majority of functional connectivity (FC) studies utilize volume-based preprocessing approaches (Tomecek et al., 2024). However, surface based measures can capture better the topography of brain networks, offering additional insights. The disruptions in brain network connectivity in the first episode psychosis is believed to be an important factor in an emergence of symptoms. The aim of this study is to compare the volume- and surfacebased FC in FEP and healthy controls to better understand the patterns of network connectivity in the psychosis onset. METHODS: The dataset included T1 structural and resting-state scans 124 first episode psychosis patients and 75 age and gender matched healthy controls,45 females, being part of the Early-Stage Schizophrenia Outcome Study. The data was collected using a 3T Siemens Trio MRI scanner using EPI sequence, for resting-state 10 min, 300 volumes, TR=2s, TE=30ms, flip angle = 70° and structural T1-weighted structural volume was acquired using MPRAGE sequence, TI/TR/ TE = 900/2300/4.63 ms, flip angle 10°. The structural data were parcellated using recon-all procedure implemented in the Freesurfer (ver. 7.2.0). The connectivity analyses were performed using the CONN toolbox release 22a. The volumebased preprocessing pipeline included realignment and unwarping, slice timing correction, outlier detection, structural and functional direct segmentation and normalization to MNI template and smoothing (FWHM=8mm). In the surface-based pipeline realignment, slice timing correction, outlier detection were also performed, followed by indirect co-registration of functional to structural data, resampling of functional data within the cortical surface and smoothing of surface level functional data using 40 diffusion steps (equivalent to volumetric 8mm kernel). Next, the data resulting from both pipelines were filtered using band-pass filter (0.004 – 0.08 Hz) and denoised using aCompCor procedure (5 parameters for white matter and CSF) and motion artifacts were removed using 6 realignment parameters and their first order derivatives. Finally ROI-to-ROI connectivity matrices were calculated using Deskian-Killiany atlas (both volumetric and surface based with 68 ROIs) for each subject, and later used for second level between-subjects comparisons. Cluster-level inferences were based on parametric statistics from Gaussian Random Field theory. Results were thresholded using p-FDR < 0.05 cluster-size threshold. RESULTS: Using surface and volume-based strategies we observed different connectivity patterns while comparing FEP patients against the healthy controls. In the volume approach negative correlation between the insula (both right and left) and the left posterior cingulate was observed. In the surface-based analysis this connection did not reach statistical significance. Both approaches yielded positive change of correlations in the occipital regions. However, in the volume-based FC an overall pattern of hyper-connectivity was obtained, whereas surface-based FC was more balanced, displaying also hypo-connectivity. CONCLUSIONS: The choice between the surface- and volume-based connectivity analyses can be critical in functional imaging. In the studies of cortical activity and connectivity surface-based methods can offer significant advantages in reducing signal noise and increasing sensitivity. Approaches based on volumetric processing may introduce errors in anatomical positioning, further influencing the results of group level analyses. Adopting different strategies in functional connectivity may improve accuracy of studies and enhance our understanding of brain function in both healthy and clinical populations. ACKNOWLEDGEMENTS: The publication was supported by ERDF-Project Brain dynamics, No. CZ.02.01.01/00/22_008/0004643 and Czech Health Research Council Project No. NU21-08-00432.