Dimensionality of ICA in resting-state fMRI investigated by feature optimized classification of independent components with SVM

Abstract

Different machine learning algorithms have recently been used for assisting automated classification of independent component analysis (ICA) results from resting-state fMRI data. The success of this approach relies on identification of artifact components and meaningful functional networks. A limiting factor of ICA is the uncertainty of the number of independent components (NIC). We aim to develop a framework based on support vector machines (SVM) and optimized feature-selection for automated classification of independent components (ICs) and use the framework to investigate the effects of input NIC on the ICA results. Seven different resting-state fMRI datasets were studied. 18 features were devised by mimicking the empirical criteria for manual evaluation. The five most significant (p < 0.01) features were identified by general linear modeling and used to generate a classification model for the framework. This feature-optimized classification of ICs with SVM (FOCIS) framework was used to classify both group and single subject ICA results. The classification results obtained using FOCIS and previously published FSL-FIX were compared against manually evaluated results. On average the false negative rate in identifying artifact contaminated ICs for FOCIS and FSL-FIX were 98.27 and 92.34%, respectively. The number of artifact and functional network components increased almost linearly with the input NIC. Through tracking, we demonstrate that incrementing NIC affects most ICs when NIC < 33, whereas only a few limited ICs are affected by direct splitting when NIC is incremented beyond NIC > 40. For a given IC, its changes with increasing NIC are individually specific irrespective whether the component is a potential resting-state functional network or an artifact component. Using FOCIS, we investigated experimentally the ICA dimensionality of resting-state fMRI datasets and found that the input NIC can critically affect the ICA results of resting-state fMRI data.

Methods for cleaning the BOLD fMRI signal

“…Therefore, several procedures have been proposed for assisting automated classification, which mainly differ in the algorithms used for supervised classification (and if necessary feature selection), the number and definitions of the spatial and temporal features used in the classification, as well as the type of fMRI data they are optimized to work with, either task-based, resting state, or both (Beall and Lowe, 2007; Bhaganagarapu et al, 2013; De Martino et al, 2007; Douglas et al, 2011; Formisano et al, 2002; Griffanti et al, 2014; Kochiyama et al, 2005; Liao et al, 2006; Perlbarg et al, 2007; Pruim et al, 2015a; 2015b; Rummel et al, 2013; Salimi-Khorshidi et al, 2014; Sochat et al, 2014; Soldati et al, 2009; Storti et al, 2013; Sui et al, 2009; Thomas et al, 2002; Tohka, et al, 2008; Wang and Li, 2015, Xu et al, 2014). …”

“…This strategy might be particularly recommended if the goal is to reveal resting state functional networks across subjects (Du et al, 2016; Wang and Li, 2015). Furthermore, it is important to consider that in case of group studies comparing patients and healthy controls (or two populations), the classifier must be trained in a subset of the healthy controls, rather than on an equal number of patients and controls (Griffanti et al, 2016).…”

“…In practice, it has been shown that the number of noise ICs is lower or approximately equal than that of neuronal-related BOLD ICs for low order ICA, but the former are about twice larger in case of high order ICA. Similarly, ICs describing a single functional network can split into several ICs at increasing orders (Abou Elseoud et al, 2011; Wang and Li, 2015). In theory, the number of ICA components could be estimated based on the space spanned by the non-thermal components of the data.…”

Studying Sub-Dendrograms of Resting-State Functional Networks with Voxel-Wise Hierarchical Clustering

“…At the large-scale and global levels, our understanding of brain connectivity topology relies mainly on the analysis of anatomical and functional connections measured by non-invasive brain imaging (Hilgetag et al, 2000 ; Zhou et al, 2006 ; Cohen et al, 2008 ; Ferrarini et al, 2009 ; Park and Friston, 2013 ; Zhen et al, 2013 ; Russo et al, 2014 ). Both clustering (Golland et al, 2008 ; Wang and Li, 2013 ) and ICA of brain imaging data (Wang and Li, 2015 ) are particularly important data-driven approaches to study brain network organization. Accumulating results from clustering and ICA studies demonstrate that the cerebral cortex can be divided into the extrinsic and intrinsic systems at the global scale (Golland et al, 2008 ).…”

Heterogeneous fractionation profiles of meta-analytic coactivation networks

“…Although prior work has sought to develop analytic strategies for automatically identifying an optimal model order of interest (Beckmann and Smith, 2004; Himberg et al, 2004; Li et al, 2007; Ray et al, 2013), these methods are somewhat arbitrary and usually depend upon a number of factors (e.g., field strength, number of time points, number of subjects, and data quality). A recent study established the importance of this dimensionality parameter (Wang and Li, 2015), demonstrating that the number of components can critically affect ICA results. Only a few studies have directly compared ICA-based resting state networks across different model orders (Smith et al, 2009; Kiviniemi et al, 2009; Abou-Elseoud et al, 2010; Pamilo et al, 2012), suggesting a hierarchical network structure (i.e., the 20 networks observed at a low-dimensionality ICA can be decomposed into distinct sub-networks at a model order of 70).…”