Hand classification of fMRI ICA noise components

Abstract

We present a practical “how-to” guide to help determine whether single-subject fMRI independent components (ICs) characterise structured noise or not. Manual identification of signal and noise after ICA decomposition is required for efficient data denoising: to train supervised algorithms, to check the results of unsupervised ones or to manually clean the data. In this paper we describe the main spatial and temporal features of ICs and provide general guidelines on how to evaluate these. Examples of signal and noise components are provided from a wide range of datasets (3T data, including examples from the UK Biobank and the Human Connectome Project, and 7T data), together with practical guidelines for their identification. Finally, we discuss how the data quality, data type and preprocessing can influence the characteristics of the ICs and present examples of particularly challenging datasets.

Preserved network functional connectivity underlies cognitive reserve in multiple sclerosis

“…For example, the network of GM regions whose WM tract disruption was significantly associated with visual/spatial memory ( Figure 5) was made up of region-pairs including right-hemispheric occipitotemporal as well as a right-hemispheric network of hippocampal connections. We opted to follow recently published guidelines for reliable hand-classification of noise components (Griffanti et al, 2017), but with larger training datasets, automated software might be able to make more reproducible classifications. Impairment of cognitive processing speed as measured by SDMT was significantly related to a much more diffuse network of structurally disrupted pairs of GM regions (Figure 4), which is consistent with our understanding of SDMT as a cognitive assessment that is highly sensitive to structural damage in PwMS (Benedict et al, 2017 (Birn et al, 2013), future work should consider longer acquisition times.…”

“…Preprocessing steps included removal of volumes 1-2, slice-time correction, motion correction, intensity normalization, high-pass temporal filtering (2,000 s), brain extraction, fieldmap unwarping, and 4 mm spatial smoothing. Independent component analysis, hand-classification, and removal of noise components were completed using FSL MELODIC according to recently published guidelines (Griffanti et al, 2017). Independent component analysis, hand-classification, and removal of noise components were completed using FSL MELODIC according to recently published guidelines (Griffanti et al, 2017).…”

Network‐targeted approach and postoperative resting‐state functional magnetic resonance imaging are associated with seizure outcome

“…8,23 Each brain network oscillates in its oxygen concentration independently from the other brain networks, which makes major brain networks distinguishable from each other. Independent component analysis (ICA) is a mathematical process that analyzes the raw rs-fMRI signal and separates it into detected oscillating subsignals.…”

Spatial parcellations, spectral filtering, and connectivity measures in fMRI: Optimizing for discrimination

“…prisingly well under certain configurations of atlas choices and dependency measures. This result suggests an additional ability of highdimensionality ICA to isolate information from noise-typically found in this range of higher frequencies(Griffanti et al, 2017). However, the use of partial correlation maintains performance at high levels even at high cutoffs.…”