Evaluating denoising strategies in resting-state functional magnetic resonance in traumatic brain injury (EpiBioS4Rx)

Author(s)

Weiler, Marina ; Casseb, Raphael F. ; de Campos, Brunno M. ; Crone, Julia S. ; Lutkenhoff, Evan S. ; Vespa, Paul M. ; Monti, Martin M. ; EpiBioS4Rx Study Grp

• Document type: Journal article
• Source: Human Brain Mapping; v. 43, n. 15, p. 10-pg., 2022-06-20.

Abstract

Resting-state functional MRI is increasingly used in the clinical setting and is now included in some diagnostic guidelines for severe brain injury patients. However, to ensure high-quality data, one should mitigate fMRI-related noise typical of this population. Therefore, we aimed to evaluate the ability of different preprocessing strategies to mitigate noise-related signal (i.e., in-scanner movement and physiological noise) in functional connectivity (FC) of traumatic brain injury (TBI) patients. We applied nine commonly used denoising strategies, combined into 17 pipelines, to 88 TBI patients from the Epilepsy Bioinformatics Study for Anti-epileptogenic Therapy clinical trial. Pipelines were evaluated by three quality control (QC) metrics across three exclusion regimes based on the participant's head movement profile. While no pipeline eliminated noise effects on FC, some pipelines exhibited relatively high effectiveness depending on the exclusion regime. Once high-motion participants were excluded, the choice of denoising pipeline becomes secondary - although this strategy leads to substantial data loss. Pipelines combining spike regression with physiological regressors were the best performers, whereas pipelines that used automated data-driven methods performed comparatively worse. In this study, we report the first large-scale evaluation of denoising pipelines aimed at reducing noise-related FC in a clinical population known to be highly susceptible to in-scanner motion and significant anatomical abnormalities. If resting-state functional magnetic resonance is to be a successful clinical technique, it is crucial that procedures mitigating the effect of noise be systematically evaluated in the most challenging populations, such as TBI datasets. (AU)

• FAPESP's process: 13/07559-3 - BRAINN - The Brazilian Institute of Neuroscience and Neurotechnology
• Grantee: Fernando Cendes
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC
• FAPESP's process: 20/00019-7 - Advanced analysis of MR imaging in epilepsy to predict surgery outcome: a machine learning approach
• Grantee: Raphael Fernandes Casseb
• Support Opportunities: Scholarships in Brazil - Post-Doctoral