Supervised Dictionary Learning for Inferring Concurrent Brain Networks

Abstract

Task-based fMRI (tfMRI) has been widely used to explore functional brain networks via predefined stimulus paradigm in the fMRI scan. Traditionally, the general linear model (GLM) has been a dominant approach to detect task-evoked networks. However, GLM focuses on task-evoked or event-evoked brain responses and possibly ignores the intrinsic brain functions. In comparison, dictionary learning and sparse coding methods have attracted much attention recently, and these methods have shown the promise of automatically and systematically decomposing fMRI signals into meaningful task-evoked and intrinsic concurrent networks. Nevertheless, two notable limitations of current data-driven dictionary learning method are that the prior knowledge of task paradigm is not sufficiently utilized and that the establishment of correspondences among dictionary atoms in different brains have been challenging. In this paper, we propose a novel supervised dictionary learning and sparse coding method for inferring functional networks from tfMRI data, which takes both of the advantages of model-driven method and data-driven method. The basic idea is to fix the task stimulus curves as predefined model-driven dictionary atoms and only optimize the other portion of data-driven dictionary atoms. Application of this novel methodology on the publicly available human connectome project (HCP) tfMRI datasets has achieved promising results.

Connectome-scale functional intrinsic connectivity networks in macaques

“…Dictionary learning and sparse representation techniques have been successfully applied for brain fMRI time series analysis and functional brain network identification (e.g., Oikonomou et al, 2012; Abolghasemi et al, 2013; Zhao S et al, 2015; Zhang S et al, 2015; Jiang et al, 2014; Jiang et al, 2015a; Jiang et al, 2015b; Lv et al, 2015a; Lv et al, 2015b; Lv et al, 2015c; Makkie et al, 2015). In this work, the computational framework of identification of connectome-scale ICNs in each individual subject via dictionary learning and sparse representation is illustrated in Fig.…”

Constructing fine-granularity functional brain network atlases via deep convolutional autoencoder

“…Thus, faithful reconstruction and quantitative modeling of those concurrent neural networks from noisy fMRI data has been of a major neuroscientific research topic for years (Bullmore and Sporns, 2009; Dosenbach et al, 2006; Duncan, 2010; Fedorenko et al, 2013; Fox et al, 2005; Huettel, Scott A., Allen W. Song, 2004; Pessoa et al, 2012). Popular brain network reconstruction techniques based on fMRI data include general linear model (GLM) (Friston et al, 1994; Worsley, 1997) for task-based fMRI (tfMRI), independent component analysis (ICA) (Beckmann et al, 2005; Calhoun et al, 2004) for resting state fMRI (rsfMRI), and dictionary learning/sparse representation (Ge et al, 2016; Jiang et al, 2015; Li et al, 2016; Lv et al, 2015a, 2015b, 2015c, 2015d; Xintao Hu et al, 2015; Zhang et al, 2016; Zhao et al, 2016, 2015) for both tfMRI and rsfMRI, all of which can effectively reconstruct concurrent network maps from whole brain fMRI data. For instance, by using the dictionary learning and sparse coding algorithms (Ge et al, 2016; Jiang et al, 2015; Li et al, 2016; Lv et al, 2015a, 2015b, 2015c, 2015d; Mairal et al, 2010; Xintao Hu et al, 2015; Zhang et al, 2016; Zhao et al, 2015), several hundred of concurrent functional brain networks, characterized by both spatial maps and associated temporal time series, can be effectively decomposed from either tfMRI or rsfMRI data of an individual brain.…”

“…Popular brain network reconstruction techniques based on fMRI data include general linear model (GLM) (Friston et al, 1994; Worsley, 1997) for task-based fMRI (tfMRI), independent component analysis (ICA) (Beckmann et al, 2005; Calhoun et al, 2004) for resting state fMRI (rsfMRI), and dictionary learning/sparse representation (Ge et al, 2016; Jiang et al, 2015; Li et al, 2016; Lv et al, 2015a, 2015b, 2015c, 2015d; Xintao Hu et al, 2015; Zhang et al, 2016; Zhao et al, 2016, 2015) for both tfMRI and rsfMRI, all of which can effectively reconstruct concurrent network maps from whole brain fMRI data. For instance, by using the dictionary learning and sparse coding algorithms (Ge et al, 2016; Jiang et al, 2015; Li et al, 2016; Lv et al, 2015a, 2015b, 2015c, 2015d; Mairal et al, 2010; Xintao Hu et al, 2015; Zhang et al, 2016; Zhao et al, 2015), several hundred of concurrent functional brain networks, characterized by both spatial maps and associated temporal time series, can be effectively decomposed from either tfMRI or rsfMRI data of an individual brain. Pooling and integrating the spatial maps of those functional networks from many brains such as those of Human Connectome Project (HCP) subjects can significantly advance our understanding of the regularity and variability of brain functions across individuals and populations (Lv et al, 2015a, 2015b; Zhao et al, 2016)(Zhao et al, 2016)(Zhao et al, 2016)(Zhao et al, 2016)(Zhao et al, 2016)(Zhao et al, 2016).…”

Higher-Order Block Term Decomposition for Spatially Folded fMRI Data

“…Tensor decomposition methods have been introduced in the field of fMRI data analysis, for effectively performing blind source separation in multi-subject (or multi-trial) experiments. The dimensions of the tensor in most of these methods are voxels × time × subject/trial, and hence in single-subject, singletrial cases a multiway tensor can not be formed and matrix decomposition methods (ICA [9], k-SVD [44], etc.) are used instead.…”