Removal of high frequency contamination from motion estimates in single-band fMRI saves data without biasing functional connectivity

Abstract

<jats:title>Abstract</jats:title><jats:p>Denoising fMRI data requires assessment of frame-to-frame head motion and removal of the biases motion introduces. This is usually done through analysis of the parameters calculated during retrospective head motion correction (i.e., ‘motion’ parameters). However, it is increasingly recognized that respiration introduces factitious head motion via perturbations of the main (B0) field. This effect appears as higher-frequency fluctuations in the motion parameters (&gt; 0.1 Hz, here referred to as ‘HF-motion’), primarily in the phase-encoding direction. This periodicity can sometimes be obscured in standard single-band fMRI (TR 2.0 – 2.5 s.) due to aliasing. Here we examined (1) how prevalent HF-motion effects are in seven single-band datasets with TR from 2.0 - 2.5 s and (2) how HF-motion affects functional connectivity. We demonstrate that HF-motion is relatively trait-like and more common in older adults, those with higher body mass index, and those with lower cardiorespiratory fitness. We propose a low-pass filtering approach to remove the contamination of high frequency effects from motion summary measures, such as framewise displacement (FD). We demonstrate that in most datasets this filtering approach saves a substantial amount of data from FD-based frame censoring, while at the same time reducing motion biases in functional connectivity measures. These findings suggest that filtering motion parameters is an effective way to improve the fidelity of head motion estimates, even in single band datasets. Particularly large data savings may accrue in datasets acquired in older and less fit participants.</jats:p><jats:sec><jats:title>Highlights</jats:title><jats:list list-type="simple"><jats:list-item><jats:label>-</jats:label><jats:p>Single-band fMRI motion traces show factitious high-frequency content (<jats:italic>HF-motion</jats:italic>)</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>The magnitude of HF-motion relates to age and other demographic factors</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>HF-motion elevates framewise displacement (FD) and causes data loss</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>Substantial fMRI data can be recovered from censoring by filtering motion traces</jats:p></jats:list-item><jats:list-item><jats:label>-</jats:label><jats:p>Filtering motion traces reduces motion artifacts in functional connectivity</jats:p></jats:list-item></jats:list></jats:sec>

Modular and state-relevant connectivity in high-frequency resting-state BOLD fMRI data: An independent component analysis

“…As such, even though there appeared to be little effect of motion in isolation, careful consideration should be applied to interpretations made regarding results within these domains. Future work would benefit from combinations of ICA, nuisance spectra (Fair et al, 2020; Gratton et al, 2019) and data-driven optimization not only to isolate signal of interest, but also design objective filters to minimize artifact, as artifact and signal of interest may not be compartmentalized to the evenly-spaced frequency bins selected for this analysis. Were such approaches to be utilized, they would need to be robust to acquisition parameters such as the Nyquist frequency limits set by the TR, and signal-and-artifact frequency overlap.…”

“…As with any fMRI study, the effects of motion may significantly contribute to the measured BOLD signals, particularly long vs. short range FNC values. Furthermore, recent work has identified that short TR multi-band sequences have respiration artifacts that have alias with motion estimates, and without information specific to respiration (e.g., waist bellows), notch filtering is recommended to mitigate respiration (Fair et al, 2020; Gratton et al, 2019). For the current study, however, motion measures likely mixed with respiration were fitted as a single regressor, both to avoid the loss of frequency-specific information associated with notch filtering, and because respiration information was not collected in the dataset.…”

Age-related change in task-evoked amygdala—prefrontal circuitry: a multiverse approach with an accelerated longitudinal cohort aged 4-22 years

“…Finally, our findings on age-related change in amygdala and mPFC function may be biased or confounded by age-related differences in head motion 70 , anatomical image quality and alignment 71,72 , signal dropout, and physiological artifacts [73][74][75] An accelerated longitudinal design was used such that participants' starting ages at scan 1 comprised the entire range of sample ages (4-22 years old), and coverage was approximately balanced across the entire age range (see Figure 1B). The design was structured into 3 study 'waves' corresponding with recruitment efforts and visit protocols.…”

A longitudinal resource for studying connectome development and its psychiatric associations during childhood

“…For the single band data (TR=2500ms), this filter is also not applicable (Nyquist frequency = 0.2Hz). We then attempted the technique suggested by Gratton et al 111 and applied a low-pass filter with a cutoff frequency at 0.1Hz. Motion parameters (FD) were recalculated after applying each of the filters.…”