نهمین کنگره نقشه برداری مغز

نمایش مقاله

Title:

Network-Specific Susceptibility of Resting-State Functional Connectivity to Imaging-Site Heterogeneity and Head Motion in ABIDE

Subject:

Autism and Brain Mapping

Presenter Author:
Name:	Leila Ghobadi

Author:
Name 1:	Leila Ghobadi - Corresponding - Presenter
Affiliation 1:	Department of Statistics, Faculty of Science, Shiraz University, Shiraz, Iran

Name 2:	Alireza Nematollahi
Affiliation 2:	Department of Statistics, Faculty of Science, Shiraz University, Shiraz, Iran

Name 3:
Affiliation 3:

Name 4:
Affiliation 4:

Name 5:
Affiliation 5:

Name 6:
Affiliation 6:

Name 7:
Affiliation 7:

Name 8:
Affiliation 8:

Name 9:
Affiliation 9:

Name 10:
Affiliation 10:

Abstract

Original Research

introduction :

Multisite resting-state functional magnetic resonance imaging (rs-fMRI) enables large-scale investigation of functional brain organization in autism spectrum disorder (ASD), but functional connectivity (FC) estimates remain susceptible to imaging-site heterogeneity and residual head motion. We examined whether susceptibility to these nuisance sources varies across functional network interactions and whether their network-level susceptibility profiles are concordant.

Methods:

We analyzed a quality-controlled ABIDE I cohort of 818 participants (373 with ASD and 445 typically developing [TD]) from 19 imaging sites after quality control including mean framewise displacement (FD) < 0.20 mm. Resting-state regional time series were obtained using the C-PAC preprocessing pipeline and CC200 parcellation. Pearson correlations were Fisher r-to-z transformed and aggregated into 28 predefined within- and between-network interactions using a high-confidence CC200-to-Yeo-7 mapping. For each interaction, FC was modeled using multivariable linear regression with imaging site as a fixed categorical effect and mean FD, age, sex, and diagnosis as covariates. Susceptibility was quantified using partial R² from full versus reduced models, with Benjamini–Hochberg false-discovery-rate (FDR) correction applied separately to site and motion effects. Concordance between site- and motion-susceptibility profiles was assessed using Spearman correlation with site-stratified participant-level bootstrap confidence intervals, focusing on relative cross-network patterns rather than absolute effect magnitudes. Robustness was evaluated using TD-only models, an alternative CC200-to-Yeo-7 mapping rule, and replacement of mean FD with percentage of high-motion frames (func_perc_fd).

Results:

Site susceptibility varied substantially across network interactions (partial R² = 0.038–0.235), with all 28 interactions significant after FDR correction. The largest site-related effects involved Default Mode–Default Mode, Limbic–Default Mode, Limbic–Limbic, and Dorsal Attention–Default Mode connectivity. Mean FD differed across sites (η² = 0.155) and was higher in ASD than in TD participants (Cohen's d = 0.30). Motion susceptibility was heterogeneous (partial R² from <0.001 to 0.056), with 13 of 28 interactions significant after FDR correction. The largest effects involved Ventral Attention–Default Mode, Limbic–Default Mode, and Dorsal Attention–Default Mode connectivity. Site- and motion-susceptibility profiles showed modest positive concordance (Spearman ρ = 0.276, bootstrap 95% CI: 0.022–0.498). The site susceptibility profile was largely preserved in TD-only models (ρ = 0.849). Replacing mean FD with func_perc_fd largely preserved the motion susceptibility profile (ρ = 0.829), with 12 of 28 interactions significant after FDR correction. Results were robust to the alternative network-mapping rule (ρ = 0.978 for site; ρ = 0.957 for motion).

Conclusion:

Imaging-site and residual head-motion susceptibility were not uniformly distributed across functional network interactions and showed only modest concordance. Given the association between site and motion variability, site effects should be interpreted as composite site-related heterogeneity rather than the influence of a single acquisition factor. These findings suggest that global nuisance summaries may not fully capture network-specific vulnerability of FC estimates in multisite neuroimaging. Results are specific to cross-sectional ABIDE I data and the cortical Yeo-7 representation used here.

Keywords:

Autism Spectrum Disorder; Magnetic Resonance Imaging; Head Movements; Brain Mapping; Connectome