Voxel-wise functional connectomics using arterial spin labeling fMRI: the role of denoising.

Brain Connect. 2015 May 28;

Authors: Liang X, Connelly A, Calamante F

Abstract
The objective of this study was to investigate voxel-wise functional connectomics using arterial spin labeling (ASL) fMRI. Since ASL signal has intrinsically low signal-to-noise ratio (SNR), the role of denoising is evaluated; in particular, a novel denoising method, dual-tree complex wavelet transform (DT-CWT) combined with the non-local means (NLM) algorithm is implemented and evaluated. Simulations were conducted to evaluate the performance of the proposed method in denoising images, and in detecting functional networks from noisy data (including the accuracy and sensitivity of detection). In addition, denoising was applied to in-vivo ASL datasets, followed by network analysis using graph theoretical approaches. Efficiencies cost was used to evaluate the performance of denoising in detecting functional networks from in-vivo ASL fMRI data. Simulations showed that denoising is effective in detecting voxel-wise functional networks from low SNR data and/or from data with small total number of time points. The capability of denoised voxel-wise functional connectivity analysis was also demonstrated with in-vivo data. We concluded that denoising is important for voxel-wise functional connectivity using ASL fMRI, and that the proposed DT-CWT-NLM method should be a useful ASL pre-processing step. Key words: Arterial spin labeling, voxel-wise functional connectomics, denoising, non-local means, dual-tree complex wavelet, network efficiency.

PMID: 26020288 [PubMed - as supplied by publisher]