Open Collections

A robust strategy for cleaning motion artifacts in resting state fMRI

University of British Columbia

There is an increasing recognition that different preprocessing approaches for functional magnetic resonance imaging (fMRI) data may have a profound impact on downstream analyses. A critical element of standard preprocessing of fMRI data is motion correction, as head motion during fMRI scanning can induce changes in blood oxygenation level dependent (BOLD) signals that may confound estimations of brain activity, particularly connectivity estimates between brain regions. In this thesis, we propose an approach which explicitly decouples the changes due to brain activation and changes due to motion. Independent Vector Analysis (IVA) is used to determine the optimal combination of basis volumes to match reference volumes. Such an approach, which we call Motion correction with IVA (McIVA), is amenable to wide-scale parallelization. The mutual information between the first volume and all subsequent volumes declined more slowly after preprocessing with McIVA compared to the raw data and other motion correction schemes. Remarkably, since the final motion-correction volume in McIVA is based on a combination of images, we show that McIVA's error can be actually less than interpolation error. What is more, McIVA resulted in the lowest connectivity across a range of spatial distances between regions of interest (ROI) pairs. When a volume is severely corrupted, the IVA fails to converge, providing a principled way to determine if a given time point is too corrupted for recovery. Finally, we assessed the effects of McIVA on two popular denoising methods, aCompCor and ICA-AROMA on resting-state fMRI derived. McIVA resulted in reducing inflated connectivity estimates, while still retaining an adequate degree of freedom. Though the proposed method is based on resting-state fMRI data, it could be applied to task-related fMRI data as well. We conclude that the proposed approach is superior for removing motion-related artifacts and reducing biases in functional connectivity estimates induced by head movement.

Text

2020-06-30

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/70677

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/