Extraction of Time-Varying Spatiotemporal Networks Using Parameter-Tuned Constrained IVA

Abstract

Dynamic functional connectivity (dFC) analysis is an effective way to capture the networks that are functionally associated and continuously changing over the scanning period. However, these methods mostly analyze the dynamic associations across the activation patterns of the spatial networks while assuming that the spatial networks are stationary. Hence, a model that allows for the variability in both domains and reduces the assumptions imposed on the data provides an effective way for extracting spatio-temporal networks. Independent vector analysis is a joint blind source separation technique that allows for estimation of spatial and temporal features while successfully preserving variability. However, its performance is affected for higher number of datasets. Hence, we develop an effective two-stage method to extract time-varying spatial and temporal features using IVA, mitigating the problems with higher number of datasets while preserving the variability across subjects and time. The first stage is used to extract reference signals using group independent component analysis (GICA) that are used in a parameter-tuned constrained IVA (pt-cIVA) framework to estimate time-varying representations of these signals by preserving the variability through tuning the constraint parameter. This approach effectively captures variability across time from a large-scale resting-state fMRI data acquired from healthy controls and patients with schizophrenia and identifies more functionally relevant connections that are significantly different among healthy controls and patients with schizophrenia, compared with the widely used GICA method alone.

Spatial Dynamic Functional Connectivity Analysis Identifies Distinctive Biomarkers in Schizophrenia

“…For each d n , we obtain 10 solutions using parameter-tuned cIVA with γ n = 3, n = 1, …, N , using the IVA-L-SOS algorithm for different random initializations and select the most consistent run using the method described in Long et al (2018b). IVA-L-SOS is a type of IVA algorithm that assumes the sources are multivariate Laplacian distributed and exploits second order statistics (SOS) (Bhinge et al, 2019). This algorithm provides a better match to the properties of fMRI sources, since fMRI sources are in general expected to have a super-Gaussian distribution, like Laplacian (Calhoun and Adali, 2012), and are correlated across windows.…”

“…Independent vector analysis (IVA) provides a general and flexible framework to spatio-temporal dFNC analysis and estimates window-specific time courses and spatial maps. However its performance degrades with increase in the size of the data, for a given number of samples (Bhinge et al, 2019). Hence, in this work, we use a data-driven method to jointly extract spatio-temporal patterns using the subsequent extraction of exemplar and dynamic components using constrained IVA (SED-cIVA) method (Bhinge et al, 2019), from a large-scale fMRI data acquired from 91 healthy controls and 88 patients with schizophrenia.…”

“…However its performance degrades with increase in the size of the data, for a given number of samples (Bhinge et al, 2019). Hence, in this work, we use a data-driven method to jointly extract spatio-temporal patterns using the subsequent extraction of exemplar and dynamic components using constrained IVA (SED-cIVA) method (Bhinge et al, 2019), from a large-scale fMRI data acquired from 91 healthy controls and 88 patients with schizophrenia. This two-stage method preserves variability in both domains while addressing the issue with large-scale data, by using a parameter tuning technique.…”

Brain network analyses in clinical neuroscience

“…Condition-specific brain states are characterised by reduced and less defined functional connectivity and more frequent recurrence [74,[95][96][97][98][99][100][101]. A higher temporal variability of dFC time-courses, was reported in schizophrenic patients in several attention, perceptual and emotion regulation RSNs [102][103][104][105][106], and related to a disruption in perceptual functions [105,107]. In contrast, lower temporal variability was reported for the default mode and fronto-parietal networks [105,107].…”

Dynamic Modeling of Common Brain Neural Activity in Motor Imagery Tasks

“…Several strategies for raw data aggregation can be implemented for building group inferences, including serial/parallel combinations of subject-level feature sets to form a more extensive multi-subject array (Lio and Boulinguez, 2016). Instead, data-driven approaches have also been employed to infer collective feature structures, like joint diagonalization (Gong et al, 2018), temporally constrained sparse representation (Zhang et al, 2019), canonical correlation analysis (de Cheveigné et al, 2019), and versions derived from independent Component Analysis (Emge et al, 2018;Huster and Raud, 2018;Bhinge et al, 2019), among others.…”