[Paper Review] Benchmarking confound regression strategies for the control of motion artifact in studies of functional connectivity

Rastko Ćirić, Daniel H. Wolf+12 more|arXiv (Cornell University)|Aug 11, 2016

Functional Brain Connectivity Studies60 references18 citations

TL;DR

This study benchmarks 12 confound regression methods for reducing motion artifact in fMRI functional connectivity analyses using 193 young adults. It finds that global signal regression minimizes motion-connection relationships but exacerbates distance-dependent artifact, while censoring methods reduce both artifacts but consume more degrees of freedom, highlighting trade-offs in method selection based on research goals.

ABSTRACT

Since initial reports regarding the impact of motion artifact on measures of functional connectivity, there has been a proliferation of confound regression methods to limit its impact. However, recent techniques have not been systematically evaluated using consistent outcome measures. Here, we provide a systematic evaluation of 12 commonly used confound regression methods in 193 young adults. Specifically, we compare methods according to three benchmarks, including the residual relationship between motion and connectivity, distance-dependent effects of motion on connectivity, and additional degrees of freedom lost in confound regression. Our results delineate two clear trade-offs among methods. First, methods that include global signal regression minimize the relationship between connectivity and motion, but unmask distance-dependent artifact. In contrast, censoring methods mitigate both motion artifact and distance-dependence, but use additional degrees of freedom. Taken together, these results emphasize the heterogeneous efficacy of proposed methods, and suggest that different confound regression strategies may be appropriate in the context of specific scientific goals.

Motivation & Objective

• To systematically evaluate the efficacy of 12 confound regression strategies in controlling motion artifact in resting-state fMRI functional connectivity studies.
• To compare methods using three benchmarks: residual motion-connection relationships, distance-dependent motion effects, and degrees of freedom lost.
• To clarify the heterogeneous performance of existing methods and guide method selection based on scientific objectives.
• To emphasize the importance of transparent reporting of motion effects to distinguish true connectivity from motion-induced bias.

Proposed method

• The study evaluated 12 confound regression methods, including global signal regression, censoring (scrubbing, spike regression), CompCor, ICA-AROMA, and local white matter signal regression.
• Data from 193 young adults were analyzed using standardized preprocessing pipelines and functional connectivity estimation via Pearson correlation of regional time series.
• Three evaluation benchmarks were applied: (1) residual correlation between motion parameters and functional connectivity, (2) distance-dependent effects of motion on connectivity, and (3) degrees of freedom lost due to confound regression.
• Statistical modeling assessed how each method altered the relationship between motion and connectivity, particularly focusing on short- vs. long-range connections.
• Methods were compared using consistent outcome measures across a single, large, well-characterized dataset to ensure fair benchmarking.
• The analysis included both group-level inference and individual-level sensitivity to motion, with emphasis on replicability and biological relevance.

Experimental results

Research questions

• RQ1 Which confound regression method most effectively reduces the residual relationship between motion and functional connectivity?
• RQ2 How do different methods affect distance-dependent motion artifacts, particularly in short- versus long-range connections?
• RQ3 To what extent do confound regression methods reduce statistical degrees of freedom, and how does this impact inference?
• RQ4 Can any method simultaneously minimize motion-connection relationships and distance-dependent artifact without excessive loss of power?
• RQ5 How does method choice influence the interpretability and reliability of functional connectivity findings in the presence of motion?

Key findings

• Global signal regression (GSR) most effectively minimized the residual relationship between motion and functional connectivity, reducing motion-related spurious correlations.
• However, GSR unmasked or exacerbated distance-dependent motion artifact, increasing connectivity in short-range connections and decreasing it in long-range connections.
• Censoring methods such as scrubbing and spike regression mitigated both motion-connection relationships and distance-dependent artifact more effectively than GSR.
• Censoring methods incurred a significant loss of degrees of freedom, reducing statistical power and increasing variability in connectivity estimates.
• Methods using higher-order confound regressors (e.g., derivatives, quadratic terms) showed improved model fit compared to random regressors, suggesting specificity over noise.
• No single method outperformed all others across all benchmarks, indicating that method choice must align with specific research goals and data characteristics.