Skip to main content
NCBI home page
Human Brain Mapping logoLink to Human Brain Mapping
. 2024 Jul 24;45(11):e26773. doi: 10.1002/hbm.26773

4D dynamic spatial brain networks at rest linked to cognition show atypical variability and coupling in schizophrenia

Krishna Pusuluri 1,✉, Zening Fu 1, Robyn Miller 1, Godfrey Pearlson 2, Peter Kochunov 3, Theo G M Van Erp 4,5, Armin Iraji 1,6, Vince D Calhoun 1
PMCID: PMC11267451  PMID: 39045900

Abstract

Despite increasing interest in the dynamics of functional brain networks, most studies focus on the changing relationships over time between spatially static networks or regions. Here we propose an approach to study dynamic spatial brain networks in human resting state functional magnetic resonance imaging (rsfMRI) data and evaluate the temporal changes in the volumes of these 4D networks. Our results show significant volumetric coupling (i.e., synchronized shrinkage and growth) between networks during the scan, that we refer to as dynamic spatial network connectivity (dSNC). We find that several features of such dynamic spatial brain networks are associated with cognition, with higher dynamic variability in these networks and higher volumetric coupling between network pairs positively associated with cognitive performance. We show that these networks are modulated differently in individuals with schizophrenia versus typical controls, resulting in network growth or shrinkage, as well as altered focus of activity within a network. Schizophrenia also shows lower spatial dynamical variability in several networks, and lower volumetric coupling between pairs of networks, thus upholding the role of dynamic spatial brain networks in cognitive impairment seen in schizophrenia. Our data show evidence for the importance of studying the typically overlooked voxel‐wise changes within and between brain networks.

Keywords: brain networks, dynamic spatial network connectivity (dSNC), inter‐network dynamic volumetric coupling, resting state fMRI (rsfMRI), schizophrenia, spatial network dynamics

Spatially dynamic brain networks show significant volumetric coupling with synchronized growth and shrinkage, referred to as dynamic spatial network connectivity (dSNC). Dynamic variability in such networks and coupling between network pairs are positively associated with cognitive performance, while showing negative association with schizophrenia, highlighting their possible role in cognitive impairment.

graphic file with name HBM-45-e26773-g002.jpg

1. INTRODUCTION

Resting‐state functional magnetic resonance imaging (rsfMRI) investigates spontaneous neural activity indirectly via blood‐oxygen‐level‐dependent (BOLD) signal (Matsui et al., 2016; Schwalm et al., 2017). Human rsfMRI data have been extensively studied to identify brain networks with coherent patterns of spatio‐temporal activity that are disrupted in numerous brain disorders including schizophrenia (SZ) (Adhikari et al., 2019; Aine et al., 2017; Calhoun et al., 2014; Damaraju et al., 2014; Iraji et al., 2019; Pini et al., 2020). Recent studies examined temporal fluctuations by employing sliding‐window approaches to measure dynamic functional network connectivity (dFNC) and found that SZ is strongly associated with the temporal dynamics of whole‐brain network connectivity (Damaraju et al., 2014; Miller et al., 2016; Sakoğlu et al., 2010). A common assumption for most studies is that these networks are fixed spatially over the course of a typical scan, failing to fully capture the highly dynamic nature of the brain networks whose functional connectivity can vary over time, while the networks themselves can also evolve temporally at the voxel‐wise scale (e.g., shrink or grow) (Iraji et al., 2020). A model of the spatial chronnectome was proposed in (Iraji et al., 2019) to identify spatially dynamic network features and voxel‐level variations in network coupling that were affected in SZ. Cognitive performance, symptom severity, and drug scores of subjects and their links to resting state dynamic functional network connectivity, multimodal biomarkers, and multi‐spatial‐scale dynamic functional connectivity have been previously investigated (Hare et al., 2019; Iraji, Faghiri, Zening, Rachakonda, et al., 2022; Sendi et al., 2021; Sui et al., 2018). Data‐driven analytical approaches such as independent component analysis (ICA) typically assume linear separability of spatial and temporal dynamics of brain networks. Here we introduce an approach to study spatially dynamic 4D brain networks (3D voxel‐wise changes over time windows) by utilizing a volumetric measure. We investigate whether significant volumetric coupling (i.e., synchronized shrinkage and growth) exists between networks over time. Extending the idea of dFNCs, we refer to such spatial coupling as dynamic spatial network connectivity (dSNC). In addition, we study how the spatial variability and other dynamic features of these networks, and the dSNC between network pairs are modulated in SZ patients compared to typical controls. Linear regression analysis is employed to reveal the associations of dynamic spatial brain network features with cognitive performance, drug dosage and symptom scores. This work highlights the importance of studying the (voxel‐wise) spatial dynamics of brain networks (Iraji et al., 2019) that could characterize brain health and disorder, and help in the development of relevant biomarkers.

2. METHODS

2.1. Data collection and preprocessing

We analyzed 3‐Tesla rsfMRI data from 508 subjects collected in three studies—The Functional Imaging Biomedical Informatics Research Network (FBIRN) (Damaraju et al., 2014), the Center for Biomedical Research Excellence (COBRE) (Aine et al., 2017), and a Maryland Psychiatric Research Center (MPRC) (Adhikari et al., 2019) study. In addition to the previously described inclusion criteria (Iraji et al., 2023; Iraji, Faghiri, Zening, Kochunov, et al., 2022), we only selected the 508 subjects who also have genomic data (for use in other parts of this project), resulting in 315 controls (CN) and 193 SZ patients. rsfMRI data preprocessing was performed using the statistical parametric mapping toolbox (SPM12, http://www.fil.ion.ucl.ac.uk/spm/). This involved the exclusion of the first five volumes for magnetization equilibrium, rigid body motion correction to account for head motion, slice‐timing correction, alignment of subject data to Montreal Neurological Institute template, resampling to 3‐mm3 isotropic voxels, and spatial smoothing using a 6‐mm full width at half maximum Gaussian kernel. In addition, we detrended, despiked, motion corrected, and filtered the voxel‐level time courses to reduce noise and nuisance signals. Further details of the datasets and preprocessing can be found in (Iraji et al., 2023; Iraji, Faghiri, Zening, Kochunov, et al., 2022).

2.2. Analysis pipeline

The analysis pipeline is depicted in Figure 1. In the first step, we performed group‐level spatially constrained independent component analysis (sICA) of rsfMRI data with a model order of 20 components, as described in (Iraji et al., 2019), using the group ICA of fMRI toolbox (GIFT) software package (Iraji et al., 2021).

FIGURE 1.

FIGURE 1

Open in a new tab

Analysis pipeline showing (1) group level spatial ICA followed by (2) subject level MOO‐ICAR over sliding windows using prior components/spatial networks (ICNs) from step (1) as reference to maintain correspondence of ICNs across subjects and time windows. (3) Volume of each network at each window is computed as the number of voxels with z‐scored activity above a volume threshold, V th. Volumetric coupling (VC) matrices are computed at the subject level (measured as correlations between volumes of ICNs across time windows). Further analysis is done using (4) one‐sample, two‐sample t‐tests and regression analysis. Detailed description of the pipeline is presented in Section 2.2.

(https://trendscenter.org/software/gift/). Of these 20 components, 14 were identified as neuro‐related brain networks, as shown in Figure 2. sICA identified and separated these brain intrinsic connectivity networks (ICNs) and their associated time courses. In the second step, a sliding‐window approach combined with spatially constrained ICA was performed for each subject across time windows, using multi‐objective optimization ICA with reference (MOO‐ICAR) (Lin et al., 2010; Yuhui & Fan, 2013). This ensured the correspondence of ICNs across subjects and time windows by using the components obtained from group level ICA analysis in the previous step as reference. Our recent study showed that MOO‐ICAR is capable of effectively estimating large‐scale networks from short time segments (Iraji, Zening, et al., 2022). This step captured, for each subject, the brain networks (ICNs) that were spatially varying over time (windows), as well as their time courses within each window. This allowed for further analysis of such spatially dynamic brain networks, inter‐network couplings, as well as different modulation under disease conditions. A sliding window length of 60 s (30 × TR, where the repetition time TR = 2 s) was employed based on previous research (Iraji et al., 2019) and falling within the recommended ranges from multiple studies (Iraji et al., 2021). Further details on the first two steps of the analysis pipeline can be found in (Iraji et al., 2023).

FIGURE 2.

FIGURE 2

Open in a new tab

The mean activity maps across typical control subjects for each of 14 relevant brain networks are shown along three planar cross sections—sagittal, coronal, and transverse—taken near the voxel with the highest activity. The maps are z‐scored and only show regions with z ≥ 2, with anatomical images overlaid in the background. aDMN, default mode—anterior; ATN, attention—dorsal; CER, cerebellar; FRNT, frontal; FPN‐L, frontoparietal left; FPN‐R, frontoparietal right; MTR‐P, somatomotor primary; MTR‐S, somatomotor secondary; pDMN, default mode posterior; VIS‐S, visual secondary; SN, salience; SUB, subcortical; TEMP, temporal; VIS‐P, visual primary.

In the third step, we performed a volumetric analysis for each subject by measuring the volume of each dynamic spatial brain network (computed as the number of z‐scored voxels with activity above a volume threshold z‐score, V th, picked from the values [0.5, 1, 1.5, 2, 2.5, 3, 3.5]) as a function of time (windows). While a volume threshold z‐score around 1.645 corresponds to a 5% p‐value for a one‐sample t‐test to identify voxels with higher than average activity, we employed several volume thresholds above and below this value (arbitrarily chosen at steps of size 0.5) to analyze the most active regions of a network as opposed to the wider regions, as shown in Figure 3. With a higher threshold, we can identify the regions of a network with the most active voxels, while using lower thresholds reveals the wider network. As elaborated further in Section 3, this allows us to study the dynamic nature of these networks and answer complex questions including whether activity is spatially more focused or diluted in SZ, whether there is higher or reduced dynamism in these networks in SZ, and how altered spatial focus of activity is associated with various cognitive, symptom and drug scores. We also computed the volumetric coupling (correlations converted into Fisher's z‐scores) between such networks over time, resulting in a 14 × 14 volumetric coupling (VC) matrix for each subject, corresponding to the 14 relevant brain networks, as shown in Step 3 of Figure 1. A large positive coupling/correlation value in the matrix implies that the two corresponding networks grow and shrink together, while a negative value implies that one network grows while the other shrinks.

FIGURE 3.

FIGURE 3

Open in a new tab

Primary visual network (VIS‐P) shown using a discontinuous color map for regions with z ≥ 0.5, with gradually increasing volume thresholds in steps of size 0.5, revealing the most active (central) regions of the network at higher thresholds while lower thresholds show the wider network.

In the fourth step, we used the subject‐level VC matrices, as well as the mean and standard deviations (SD) of the volumes of networks, to perform hypothesis testing. We identified dynamic spatial brain networks with significant VC in each group using one‐sample t‐tests, as well as studied group differences in mean/SD network volumes and VC of network pairs between the control group and individuals with schizophrenia using two‐sample t‐tests. A 5% false discovery rate (FDR) (Benjamini & Yekutieli, 2005) was applied to correct for multiple comparisons at each threshold or across all thresholds (results shown in bold). Performing such analyses at multiple volume thresholds (V th) allowed us to examine the networks at different levels of granularity, and identify changes in regions of either more focused or wide spread activity. We also performed robust linear regression analysis of cognitive scores of all subjects, and symptom and drug scores of patients against the mean/SD network volumes and VC of network pairs as predictors, while using age, gender and scanning site as covariates. The positive and negative syndrome scale (PANSS) scores (Hare et al., 2019) were available for the FBIRN and COBRE datasets, while Brief Psychiatric Rating Scale (BPRS) scores were available for MPRC dataset, which were then converted into PANSS total scores using the matching obtained in (Iraji, Faghiri, Zening, Rachakonda, et al., 2022; Leucht et al., 2013). Chlorpromazine (CPZ) equivalent drug scores (Hare et al., 2019; Woods, 2003) were available for all three datasets. Across both patients and controls, we also employed two cognitive scores (Sui et al., 2018) (processing speed, working memory) available for all three datasets and two cognitive scores (visual learning, verbal learning) available only for FBIRN and COBRE, after separately normalizing the scores for each dataset. On the whole, 178 patients had PANSS total scores, 111 patients had CPZ drug scores, 459 subjects had processing speed scores, 377 subjects had working memory scores, 254 subjects had verbal learning scores, and 251 subjects had visual learning scores available that were employed in the analyses.

3. RESULTS AND DISCUSSION

3.1. Dynamic spatial brain networks

Fourteen relevant brain networks are identified, as shown in Figure 2. These are labeled VIS‐P (visual primary), SUB (subcortical), MTR‐P (somatomotor primary), CER (cerebellar), ATN (attention—dorsal), FRNT (frontal), MTR‐S (somatomotor secondary), FPN‐R (frontoparietal right), VIS‐S (visual secondary), pDMN (default mode posterior), FPN‐L (frontoparietal left), SN (salience), TEMP (temporal), and aDMN (default mode—anterior) networks. These networks are spatially dynamic at the voxel level and undergo changes in volumes over time, as demonstrated in Figure 4 for the networks CER and pDMN in CN group. The figure shows the mean activity maps and demonstrates the variability in the volumes of these networks across CN subjects over windows with the largest volumes (top 10%) or the smallest volumes (bottom 10%), measured at V th = 2. In addition, the difference maps between the largest and the smallest mean volumes depict the regions where volumetric changes occur. As can be inferred from Figures 2 and 4 and the color bars next to them, by only examining regions above a particular V th, one can investigate either the most active/focused region of the network (high V th) or a more widespread region (low V th).

FIGURE 4.

FIGURE 4

Open in a new tab

Spatial variability and regions with volumetric changes in cerebellar (CER—left) and posterior default mode (pDMN—right) networks revealed using mean activity maps across subjects in CN group, over windows with bottom 10% smallest volumes or top 10% largest volumes measured at V th = 2, and the differences between the two.

3.2. Altered spatial focus of activity and lower spatial network dynamics in SZ

Volume of each network was measured per subject per window at different volume thresholds (V th) to determine subject level mean and SD of network volume. We analyzed results at different volume thresholds to determine whether the most active regions of the network are affected differently from the wider network. Two sample t‐test results comparing CN and SZ groups for mean volumes of networks are shown in Table 1, while those for the SD volumes using non‐parametric rank‐sum tests are shown in Table 2, with a 5% false discovery rate (FDR) (Benjamini & Yekutieli, 2005) for multiple comparisons at each threshold or across all thresholds (in bold). Table 1 reveals that several networks show significant mean volume changes in SZ as they grow or shrink during the scan. For SUB, CER and FPN‐R networks, CN shows lower volumes (blue) at low thresholds and higher volumes (red) at higher thresholds compared to SZ, implying that spatially focused activity in these networks is lowered in SZ as activity becomes more widespread or spatially diluted. For the networks VIS‐S and TEMP, it is vice versa with activity becoming more spatially focused and less widespread/diluted in SZ. ATN and pDMN networks only shrink at moderate thresholds. Table 2 shows that in the case of SD of volumes, the networks VIS‐P, MTR‐P, CER, ATN, FRNT, MTR‐S, VIS‐S, pBMN and TEMP have lower variability or spatial dynamics in SZ at different thresholds, while no networks seem to have significantly higher variability in SZ as compared to CN. For each network, results at different thresholds reveal how spatially focused or widespread are the reduced spatial dynamics seen in SZ. For example, the network MTR‐S shows most significantly lower spatial dynamics in SZ at V th = 1, which gradually become less significant as we move towards more widespread (smaller V th) or more spatially focused (higher V th) regions of the network.

TABLE 1.

(CN − SZ) two‐sample t‐test results comparing mean network volumes across CN and SZ groups at different volume thresholds.

Volume threshold 0.5 1 1.5 2 2.5 3 3.5
VIS‐P 0.0094396 −0.67467 −0.2901 −0.16913 −0.24318 −0.18351 −0.017194
SUB −2.1411 −0.095268 1.6995 1.9817 1.8919 1.7924 1.5886
MTR‐P 0.76234 −0.015179 −1.0331 −1.2044 −1.289 −1.1102 −0.79436
CER −3.3364 −0.17031 2.9802 3.1319 2.6908 “2.128” 1.5517
ATN 1.5017 −1.3744 −2.7792 −2.216 −1.3068 −0.75839 −0.49952
FRNT −0.69518 −1.5552 −0.20074 0.25048 0.62401 0.87852 1.0343
MTR‐S 1.1157 0.043892 −1.3458 −1.3825 −1.0508 −0.48056 −0.097005
FPN‐R −2.407 −2.4334 0.52785 1.4762 2.0225 2.4475 “2.3738”
VIS‐S 1.7599 −1.2723 −3.051 −2.5426 −2.2593 −1.3419 −0.63712
pDMN 1.0274 −4.835 −5.6172 −3.1533 −1.1735 −0.21363 0.059581
FPN‐L −1.2931 −1.7746 0.13007 0.87923 1.3016 1.556 1.8707
SN −0.82516 −1.3297 −0.082985 0.19803 0.33028 0.64614 1.2361
TEMP 1.8193 −2.7545 −3.8361 −2.8574 −1.9538 −1.3492 −0.94463
aDMN −1.1466 −0.2445 0.79391 0.82288 0.70661 0.76537 0.7978
Open in a new tab

Note: k = −log10(p) × sign(t) values are shown, with significant ones depicted in red for CN > SZ, and blue for CN < SZ. Significant results at each threshold with 5% FDR for multiple comparisons (14 comparisons, one for each network) are shown, while those that remain significant with 5% FDR across all thresholds (14 × 7 comparisons) are shown in bold. Note that CER network at V th = 3, and FPN‐R network at V th = 3.5 (shown in double quotes) are significant with FDR across all thresholds, but not with FDR at the single threshold (while less likely in general, this is still possible depending upon the combination of p‐values for which FDR is calculated across a single or all thresholds).

TABLE 2.

(CN − SZ) two‐sample non‐parametric rank‐sum test results comparing the standard deviation of network volumes over windows per subject across CN and SZ groups at different volume thresholds.

Volume threshold 0.5 1 1.5 2 2.5 3 3.5
VIS‐P 0.90706 1.8232 1.6755 1.4843 0.24438 −0.39964 −0.80985
SUB 0.70326 1.0632 1.1278 0.60998 1.2882 1.2383 0.99868
MTR‐P 1.9253 2.1333 3.2954 2.8933 0.24956 −0.44333 −1.073
CER 1.5615 0.98867 1.4843 1.2669 “2.2754” “2.0787” 1.4702
ATN 2.2537 2.3241 2.7664 2.0437 −0.079203 −1.1611 −1.1815
FRNT 0.36165 1.1527 1.9253 0.19987 −0.25118 0.20323 0.56631
MTR‐S 2.0564 3.702 3.4156 2.9502 1.0961 −0.29175 −1.0386
FPN‐R 1.1521 0.76978 1.4329 1.4521 1.5733 1.207 1.0758
VIS‐S 0.71425 1.592 2.0414 1.4924 0.14984 −0.84875 −1.618
pDMN 0.24085 1.3512 2.675 0.15411 −0.79823 −0.49077 −0.25248
FPN‐L 1.2426 0.44649 1.3667 1.1967 0.76978 0.74373 1.2266
SN 0.28565 1.5809 1.0004 0.67029 0.6536 0.46802 0.52484
TEMP 2.554 2.2529 3.0793 1.9662 0.064951 −0.98922 −1.2315
aDMN 0.7902 0.59129 1.4012 0.98978 0.93498 0.49941 0.36678
Open in a new tab

Note: k = −log10(p) × sign(z) values are shown, with significant CN > SZ ones depicted in red (no networks found with significant CN < SZ). Significant results at each threshold with 5% FDR for multiple comparisons (14 comparisons, one for each network) are shown, while those that remain significant with 5% FDR across all thresholds (14 × 7 comparisons) are shown in bold. Note that CER network at V th = 2.5, 3 (shown in double quotes) is significant with FDR across all thresholds, but not with FDR at single thresholds.

3.3. Dynamic spatial networks show significant volumetric coupling (dSNC)

Volumetric coupling between spatial dynamic network pairs (dSNC) was measured per subject as the correlation between their volumes over time windows, measured at different volume thresholds (V th). Results from one sample t‐tests checking for non‐zero correlations, as shown in Figure 5 and Table 3, demonstrate that several dynamic spatial network pairs across groups have significant VC. Positive VC between two networks indicates that both networks grow and shrink together, while negative VC indicates one network grows while the other shrinks. The connectogram in Figure 5 at V th = 2 shows strong positive coupling (yellow/orange lines) between the network pairs (VIS‐P/VIS‐S), (MTR‐P/MTR‐S) and (FPN‐R/FPN‐L). The network aDMN has a strong negative coupling (blue line) with ATN, and a strong positive coupling (red line) with FRNT. Similarly, the network SN has strong negative coupling (blue line) with pDMN, and strong positive coupling (red lines) with both the networks SUB and MTR‐S. A false discovery rate (FDR) of 5% to correct for multiple comparisons (Benjamini & Yekutieli, 2005) at this threshold shows that out of 91 possible network pair combinations for the 14 relevant brain networks, 62 network pairs in controls have significant VC, of which 34 are positively, and 28 are negatively coupled. A similar analysis for patients with SZ shows that 45 brain network pairs have significant VC, with 30 network pairs having positive and 15 networks having negative coupling. For the combined dataset with subjects from both groups, 65 network pairs have significant VC, with 35 positively and 30 negatively coupled. The total number of network pairs with significant positive/negative VC at different volume thresholds are depicted in Table 3, which shows that several network pairs are volumetrically coupled across groups with 5% FDR at threshold‐level, and most of the network pairs show significant coupling even with 5% FDR applied across all thresholds (in bold).

FIGURE 5.

FIGURE 5

Open in a new tab

A connectogram depicting networks with strong volumetric coupling, based on the results of one sample t‐tests for controls at V th = 2. Connections are drawn using the values of k = −log10(p) × sign(t) for each network pair, where t is the computed t‐statistic, and p is the resulting p‐value. Only strong connections with abs(k) > 10 are shown, while the 5% FDR threshold for multiple comparisons at abs(k) = 1.5391 (corresponding to the critical p‐value of .0289) shows that 62 network pairs (out of 91 possible combinations) have significant volumetric coupling, of which 34 are positively coupled (i.e., when one network grows, so does the other) and 28 negatively coupled (i.e., when one network grows, the other shrinks). The connectogram shows strong positive coupling (yellow/orange lines) between the network pairs (VIS‐P/VIS‐S), (MTR‐P/MTR‐S), (FPN‐R/FPN‐L), while strong negative coupling (blue lines) exists between the networks (aDMN/ATN), (pDMN/SN).

TABLE 3.

Number of network pairs with significant non‐zero positive (red) or negative (blue) volumetric coupling (dSNC) as found from one‐sample t‐test results at different volume thresholds (out of 91 possible network pair combinations, chosen from 14 networks).

Volume threshold All subjects +ve All subjects −ve CN +ve CN −ve SZ +ve SZ −ve
0.5 32, 32 35, 34 30, 30 30, 30 21, 21 20, 20
1 91, 91 0, 0 91, 91 0, 0 83, 79 0, 0
1.5 57, 57 6, 6 49, 49 6, 6 49, 47 5, 5
2 35, 35 30, 30 34, 34 28, 28 30, 30 15, 15
2.5 28, 28 33, 33 30, 30 29, 29 22, 22 16, 17
3 30, 30 20, 20 29, 30 19, 20 21, 21 7, 8
3.5 40, 41 3, 3 35, 37 2, 2 22, 26 0, 0
Open in a new tab

Note: Results with 5% FDR correction applied for multiple (91) comparisons at each given threshold are shown, while significant results with 5% FDR correction applied across all thresholds (91 × 7 total comparisons) are shown in bold. The CN group consistently shows more network pairs with significant positive/negative volumetric coupling than the SZ group.

3.4. Reduced volumetric coupling (dSNC) in SZ

One‐sample t‐test results in Table 3 show that fewer number of dynamic spatial brain networks show significant VC in SZ compared to CN group, implying that VC, whether positive or negative, is lower in SZ as compared to CN. Two‐sample t‐tests of subject level VC measures were performed at different V th to identify volumetrically coupled network pairs with different modulations under SZ disease conditions as compared to CN. Multiple comparisons were accounted for using 5% FDR correction at each threshold or across all thresholds (in bold). As shown in Table 4, we identified several network pairs with significant group differences in (CN − SZ) two‐sample t‐tests. A significantly negative (CN − SZ) t‐statistic could imply that either SZ patients gained positive VC that did not exist in CN, or that SZ patients lost negative VC that existed in CN. Similarly, a significant positive (CN − SZ) t‐statistic could imply either that SZ patients lost positive VC existing in CN, or that SZ patients gained negative VC that did not exist in CN. This is resolved by looking at the corresponding one‐sample t‐test results in Table 4, which show that in all the cases SZ patients lost either positive/negative VC existing in CN, but no networks gain positive/negative VC in SZ compared to CN. The network pairs (VIS‐S/FPN‐R) at V th = {0.5, 2, 2.5} and (SUB/VIS‐P) at V th = {0.5, 2} have significant negative VC in CN, which is lost in SZ. The network pairs (TEMP/MTR‐S), (CER/SUB) at V th = 0.5 and (SN/VIS‐S) at V th = 1 have significant positive VC in CN, which is lost in SZ. The network pair (aDMN–FRNT) at V th = 1 shows highly significant positive VC in CN, that is somewhat lost in SZ as shown by two‐sample t‐test results, but still remains significant in one‐sample t‐test results. Even with FDR applied across all thresholds, the positive coupling of (SN/VIS‐S) at V th = 1 and the negative coupling of (VIS‐S/FPN‐R) at V th = 2 remain significantly higher in CN compared to SZ.

TABLE 4.

Summary table showing dynamic spatial brain network pairs whose volumetric coupling (dSNC) is significantly affected under SZ.

Volume threshold Network1 Network2 Two sample t‐test (CN‐SZ) One sample t‐test (all subjects) One sample t‐test (CN) One sample t‐test (SZ)
0.5 SUB VIS‐P −3.2803 −5.9809 −8.4771 −0.13055
VIS‐S FPN‐R −3.538 −2.6627 −5.9664 0.41982
TEMP MTR‐S 3.3503 1.6406 4.0571 −0.76595
CER SUB 2.9146 5.9675 7.7499 0.22573
1 SN VIS‐S 4.2897 12.1594 15.1279 0.76533
aDMN FRNT 3.1873 25.3679 21.6856 5.1741
2 SUB VIS‐P −3.3981 −4.8994 −7.8222 0.019538
VIS‐S FPN‐R −4.0738 −2.1391 −5.3376 0.79012
2.5 VIS‐S FPN‐R −3.3537 −2.4936 −5.5555 0.41586
Open in a new tab

Note: Significant (CN − SZ) two‐sample t‐test results with 5% FDR correction for multiple (91) comparisons at each V th are shown, along with the corresponding one‐sample t‐test results; All values shown are k = −log10(p) × sign(t). Network pairs shown in blue lost negative VC in SZ compared to CN group, those shown in red lost positive VC in SZ, while the one shown in orange lost some positive VC in SZ compared to CN in two‐sample t‐test, but still remains significant in one sample t‐test in SZ; Results that remain significant with 5% FDR across all thresholds (91 × 7 comparisons) are shown in bold (SN/VIS‐S at V th = 1 and VIS‐S/FPN‐R at V th = 2). Results show SZ patients lost positive/negative VC compared to the control group (CN).

3.5. Association of 4D dynamic spatial brain networks with subject cognitive, symptom and drug scores

In order to investigate possible links between dynamic spatial brain networks and subject cognitive, symptom and drug scores, we performed robust linear regression analysis with 4 cognitive scores (visual learning, verbal learning, processing speed and working memory) for all subjects, chlorpromazine (CPZ) drug score for patients, and the positive and negative syndrome scale (PANSS) total symptom score for patients, each separately regressed against the subject‐wise means and SD of the volumes of 14 different networks across windows, as well as the subject‐wise volumetric coupling between network pairs (91 combinations) as predictors. This was repeated at 7 different volume thresholds, for a total of 4998 separate regressions (6 scores × 7 thresholds × (14 mean network volumes + 14 SD network volumes + 91 VC of network pairs)). In each case, age, gender and imaging site were used as covariates. Table 5 shows the features of dynamic spatial brain networks with significant positive (red) or negative (blue) association to subject scores, with 5% FDR correction applied for multiple comparisons at each given threshold (14 total comparisons for mean/SD and 91 for VC), or 5% FDR correction across all thresholds (14 × 7 comparisons for mean/SD and 91 × 7 for VC; shown in bold). Results reveal strong association of dynamic spatial brain network features with cognitive, drug and symptom scores.

TABLE 5.

Association of dynamic spatial brain networks with subject cognitive, symptom and drug scores via robust linear regression analysis: 4 cognitive scores (visual learning, verbal learning, processing speed and working memory) for all subjects, chlorpromazine (CPZ) equivalent drug score for patients, and the positive and negative syndrome scale (PANSS) total symptom score for patients are each separately regressed against the subject‐wise means and standard deviations (SD) of the volumes of 14 different networks across windows, as well as the volumetric coupling between network pairs (91 combinations) as predictors.

Score V th Mean network volume Std. network volume Network pairs VC
Visual learning 0.5 SUB (−3.7497) CER (−2.7101)
FPN‐R (−2.8957) FPN‐L (−2.1757)
TEMP (3.3449)
1 SUB (−2.2909) VIS‐S (−2.0565) FPN‐L (2.219) TEMP (4.4672) VIS‐S_SN (3.6095)
pDMN (−3.7004) SN (−4.0015)
TEMP (−2.5697)
1.5 TEMP (−5.2192) VIS‐P (3.0656) TEMP (2.2133)
aDMN (2.1947)
2 SUB (2.582) CER (1.8409) VIS‐P (2.535)
FPN‐R (2.0186) FPN‐L (2.0236)
TEMP (−4.5004)
2.5 SUB (3.0009) CER (2.155)
FPN‐R (2.6259) TEMP (−3.3722)
3 SUB (3.1906) CER (2.5547) SUB (3.1232) CER (1.9753)
FPN‐R (3.0619) TEMP (−2.2514) FPN‐R (2.0044) FPN‐L (2.2775)
TEMP (−1.9108)
3.5 SUB (3.1741) CER (2.5964) SUB (2.9359) CER (2.4758)
FPN‐R (2.8418) SN (2.0359) FPN‐R (2.211) FPN‐L (1.7988)
SN (2.0853)
Verbal learning CER (−2.5356) FPN‐R (−2.6899)
0.5 FPN‐L (−2.6967)
1 VIS‐S_SN (3.309)
2 CER (2.1507) FPN‐R (2.4745)
FPN‐L (2.599)
2.5 CER (2.0852) MTR‐S (−2.8836)
FPN‐R (2.2112) FPN‐L (2.4156)
3 MTR‐S (−3.0627)
3.5 MTR‐S (−2.5588) SN (2.4424) SN (2.6854)
Processing speed 0.5 TEMP (2.5217) MTR‐S (2.5104)
1 pDMN (−3.1079) MTR‐S (2.783) SN (2.0394) VIS‐S_SN (3.3209)
TEMP (2.8588)
1.5 pDMN (−2.7853) TEMP (−3.1043) MTR‐S (2.6377) VIS‐S (2.6716)
2 TEMP (−3.3893) MTR‐S (3.5447) VIS‐S (2.7952)
2.5 TEMP (−3.5632) MTR‐S (2.5309)
3 TEMP (−3.014) SN (2.4798)
3.5 SN (2.6027) SN (3.1808)
Working memory 0.5 SUB_FPN‐R (3.3628)
1 MTR‐P_ATN (3.0683)
FRNT_FPN‐L (3.6719)
2.5 SUB_FPN‐R (3.4186)
FPN‐R_VIS‐S (−3.0587)
3 VIS‐P_TEMP (3.0692)
SUB_FPN‐R (3.1227)
CPZ drug 1.5 MTR‐S (−2.4957) SN (2.7973)
PANSS total 1.5 CER (2.7301)
2 CER (2.6539)
2.5 CER (2.5912)
Open in a new tab

Note: This is repeated at seven different volume thresholds. Age, gender and imaging site are used as covariates in each case. Values shown are k = −log10(p) × sign(t), where t is the ratio of the regression coefficient to the standard error of its estimate. Predictors with positive coefficients are shown in red, while those with negative coefficients are shown in blue. Significant results with 5% FDR correction applied for multiple comparisons at each given threshold or across all thresholds (in bold) are shown. Results show strong association of dynamic spatial brain network features with cognitive, drug and symptom scores.

As seen with two‐sample t‐test results comparing CN versus SZ mean network volumes in Section 3.2 and Table 1, regression results in Table 5 for various scores against the mean network volumes in all subjects also reveal how altered spatial focus of activity affects these scores. For example, visual learning shows negative association with SUB, CER and FPN‐R at low thresholds (V th = 0.5), and positive association with those networks at high thresholds (V th = 2, 2.5, 3, 3.5), implying that more spatially focused activity in these networks has a positive effect on the score, while less focused or more wide spread (spatially diluted) activity has a negative effect. Similarly, positive associations are seen for more spatially focused activity in CER, FPN‐R and FPN‐L with verbal learning, while more wide spread activity is associated with lower scores. For both visual learning and processing speed, more wide spread activity in TEMP is associated with improved scores, whereas more spatially focused activity is linked to lower scores.

Regression results for SD in Table 5 show that all the scores predominantly show positive association (red) with SD network volumes (with the exception of visual learning with TEMP at V th = 3), implying that higher dynamic variability in these networks is linked to improved scores. Comparing these results against the results from two‐sample rank‐sum tests for CN versus SZ SD network volumes in Section 3.2 and Table 2 which showed SZ only reduced dynamic variability in these networks, highlights the role of spatial dynamic variability in cognitive performance and SZ.

As described for two‐sample t‐test results for VC in SZ versus CN in Section 3.4 and Table 4, regression results against VC can only be resolved properly by also analyzing the corresponding one‐sample VC results looking for non‐zero coupling between these network pairs. A significant positive regression coefficient for a score against the VC of a network pair could imply that the network pair has positive VC, and a higher magnitude of such coupling leads to higher performance/score, or alternatively, that the network has negative VC and higher magnitude of such coupling leads to lower performance/score. Similarly, a significant negative regression coefficient implies that the network has negative VC and higher magnitude of the coupling leads to improved scores, or alternatively, that the network has positive VC and higher magnitude leads to lower scores. In each case, we examined the corresponding one‐sample t‐test results (values not shown) for non‐zero VC and found that positive regression coefficients against VC in Table 5 are due to higher magnitudes of positive VC in the corresponding networks associated with improved scores, and negative regression coefficients are due to higher magnitudes of negative VC in the corresponding networks associated with improved scores, thus highlighting in both cases that improved scores are associated with stronger VC (positive/negative), whereas SZ is associated with the loss of such VC (positive/negative) between network pairs as seen in Section 3.4 and Table 4.

4. CONCLUSIONS

We identified several spatially dynamic brain networks that expand or shrink over time. SZ is linked to such network dynamics, as some networks showed higher volumes in SZ (ATN, pDMN), some showed predominantly less spatially focused activity and higher wide spread activity (SUB, CER, FPN‐R), while a few others showed predominantly higher spatially focused activity and lower wide spread activity (VIS‐S, TEMP). In addition, several networks (VIS‐P, MTR‐P, CER, ATN, FRNT, MTR‐S, VIS‐S, pDMN, TEMP) showed consistently lower spatial dynamical variability in SZ compared to CN, while no networks showed higher variability in SZ. We also demonstrated that dynamic spatial brain networks showed significant volumetric coupling with synchronized shrinkage/expansion, that we refer to as dynamic spatial network connectivity (dSNC). Such network coupling is altered in SZ, and patients showed significantly lower positive (TEMP/MTR‐S, CER/SUB, SUB/VIS‐S, aDMN/FRNT) or negative (SUB/VIS‐P, VIS‐S/FPN‐R) volumetric coupling between network pairs compared to controls. On the other hand, no network pairs were found to have significantly higher positive/negative volumetric coupling in SZ.

Our results showing significant group differences in brain network behaviors between CN and SZ are consistent with and extend findings from several previous studies (Baker et al., 2014; Calhoun et al., 2009; Damaraju et al., 2014; Erdeniz et al., 2017; Garrity et al., 2007; Gavrilescu et al., 2010; Iraji et al., 2019; Jafri et al., 2008; Skudlarski et al., 2010; Vercammen et al., 2010; Zeng et al., 2018). Reduced functional connectivity in SZ has been consistently reported in several of these studies and our results showed that such reduced connectivity in SZ extends beyond the temporal domain of static networks to spatially dynamic networks as well. The dysfunction of the visual system (Deng et al., 2019; Lang et al., 2016), the reduced suppression of DMN activity and its aberrant functional connectivity (Garrity et al., 2007; Karbasforoushan & Woodward, 2012; Karlsgodt et al., 2010), and the altered connectivity in frontal and temporal regions (Fornito et al., 2012; Van Den Heuvel & Fornito, 2014) have been reported previously in SZ. Disruptions in DMN, SN, FPN and ATN networks, and their connectivity could be related to psychopathology and cognitive disturbances (Hare et al., 2019; Iraji et al., 2023; Menon et al., 2023). The language related regions MTR‐S (speech production) and TEMP (semantic integration) have been shown to be disrupted and their functional/spatial connectivity altered in SZ (Cao et al., 2019; Damaraju et al., 2014; Wei et al., 2022; Woodward et al., 2012). Motor impairments from early stages, temporal cortex dysfunction, impaired psycho‐social functioning, genetic risk for MTR and TEMP networks have also been associated with SZ (Bombin et al., 2005; Chan et al., 2010; Iraji et al., 2023; Javitt & Sweet, 2015). The altered dSNC that we found in this study between TEMP/MTR‐S and CER/SUB in SZ is consistent with the cerebellar‐thalamic‐cortical dysconnectivity model of psychosis (Cao et al., 2019; Damaraju et al., 2014; Wei et al., 2022; Woodward et al., 2012).

Impairments in speed of processing and attention/vigilance were seen previously in SZ compared to healthy subjects (van Erp et al., 2015). SN and DMN networks, and their functional connectivity were associated with symptom and cognitive scores in SZ (Hare et al., 2019; Sui et al., 2018). Our robust linear regression analyses showed that several features of dynamic spatial brain networks, such as the subject‐wise means or standard deviations of network volumes and the volumetric coupling between network pairs were associated with cognitive performance, CPZ equivalent drug dosage and symptom scores, as well as highlighted the role of altered spatial focus of activity in these networks and its effects on these scores. While cognitive, symptom and drug scores showed strong positive association with higher dynamic variability in network volumes and higher VC (positive/negative) between network pairs, SZ showed lower dynamic variability in the volumes of these networks as well as lower VC between network pairs compared to controls, suggesting a possible mechanism for cognitive defects of SZ wherein reduced network‐level spatial dynamics and reduced inter‐network VC could lead to diminished performance. Based on the dysconnectivity hypothesis (Friston et al., 2016), disruptions in brain wide functional/spatial networks and their connectivity could be expected to be associated with altered behaviors and psychopathology. This is supported by our findings of significant disruptions in dynamic spatial brain network features and dSNC seen in SZ.

While we analyzed combined data from three separate datasets for greater statistical power in this study, future work should replicate these findings in independent and larger datasets and investigate the interplay of functional and spatial dynamic networks to further establish their links to SZ associated disruptions of behavior and the underlying mechanisms. In addition to the possible effects of CPZ equivalents explored in current work, future studies should also investigate anti‐cholinergic burden as another likely cause of medication‐induced cognitive variance (Joshi et al., 2021). Visualizing these dynamic changes in spatial networks and their coupling could help identify the mechanisms by which these networks expand/shrink and how they are affected in SZ. In addition, one could also employ other metrics for dynamic spatial brain network coupling beyond simple volumetric measures, as well as adapt and combine with other techniques to identify spatiotemporal structures, such as dynamic mode decomposition where modes have intrinsic temporal behaviors (Kunert‐Graf et al., 2019). Future research could investigate the impact of sliding window lengths on dynamic spatial brain networks, their coupling, and the observed significant effects of SZ. Our work highlights the importance of capturing spatiotemporal structure within and among networks and should be explored further to identify possible indicators of healthy and disordered brain function.

CONFLICT OF INTEREST STATEMENT

The authors of this study declare no competing interests.

ACKNOWLEDGMENTS

This work is funded in part by the NSF grant 2112455 and the NIH grant R01MH123610 to Dr. Vince D. Calhoun, and the NIH grant 5R01MH119251 to Dr. Armin Iraji. We would like to thank the members of TReNDS Center for valuable discussions.

Pusuluri, K. , Fu, Z. , Miller, R. , Pearlson, G. , Kochunov, P. , Van Erp, T. G. M. , Iraji, A. , & Calhoun, V. D. (2024). 4D dynamic spatial brain networks at rest linked to cognition show atypical variability and coupling in schizophrenia. Human Brain Mapping, 45(11), e26773. 10.1002/hbm.26773

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

REFERENCES

  1. Adhikari, B. M. , Elliot Hong, L. , Sampath, H. , Chiappelli, J. , Jahanshad, N. , Thompson, P. M. , Rowland, L. M. , Calhoun, V. D. , Xiaoming, D. , Chen, S. , & Kochunov, P. (2019). Functional network connectivity impairments and core cognitive deficits in schizophrenia. Human Brain Mapping, 40(16), 4593–4605. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Aine, C. J. , Bockholt, H. J. , Bustillo, J. R. , Cañive, J. M. , Caprihan, A. , Gasparovic, C. , Hanlon, F. M. , Houck, J. M. , Jung, R. E. , Lauriello, J. , Liu, J. , Mayer, A. R. , Perrone‐Bizzozero, N. I. , Posse, S. , Stephen, J. M. , Turner, J. A. , Clark, V. P. , & Calhoun, V. D. (2017). Multimodal neuroimaging in schizophrenia: description and dissemination. Neuroinformatics, 15(4), 343–364. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baker, J. T. , Holmes, A. J. , Masters, G. A. , Thomas Yeo, B. T. , Krienen, F. , Buckner, R. L. , & Öngür, D. (2014). Disruption of cortical association networks in schizophrenia and psychotic bipolar disorder. JAMA Psychiatry, 71(2), 109–118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Benjamini, Y. , & Yekutieli, D. (2005). False discovery rate‐adjusted multiple confidence intervals for selected parameters. Journal of the American Statistical Association, 100(469), 71–81. [Google Scholar]
  5. Bombin, I. , Arango, C. , & Buchanan, R. W. (2005). Significance and meaning of neurological signs in schizophrenia: Two decades later. Schizophrenia Bulletin, 31(4), 962–977. [DOI] [PubMed] [Google Scholar]
  6. Calhoun, V. D. , Eichele, T. , & Pearlson, G. (2009). Functional brain networks in schizophrenia: A review. Frontiers in Human Neuroscience, 3, 833. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Calhoun, V. D. , Miller, R. , Pearlson, G. , & Adalı, T. (2014). The chronnectome: Time‐varying connectivity networks as the next frontier in fmri data discovery. Neuron, 84(2), 262–274. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cao, H. , Ingvar, M. , Hultman, C. M. , & Cannon, T. (2019). Evidence for cerebello‐thalamo‐cortical hyperconnectivity as a heritable trait for schizophrenia. Translational Psychiatry, 9(1), 192. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Chan, R. C. K. , Xu, T. , Heinrichs, R. W. , Yu, Y. , & Wang, Y. (2010). Neurological soft signs in schizophrenia: A meta‐analysis. Schizophrenia Bulletin, 36(6), 1089–1104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Damaraju, E. , Allen, E. A. , Belger, A. , Ford, J. M. , McEwen, S. , Mathalon, D. H. , Mueller, B. A. , Pearlson, G. D. , Potkin, S. G. , Preda, A. , Turner, J. A. , Vaidya, J. G. , van Erp, T. G. , & Calhoun, V. D. (2014). Dynamic functional connectivity analysis reveals transient states of dysconnectivity in schizophrenia. NeuroImage: Clinical, 5, 298–308. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Deng, Y. , Liu, K. , Cheng, D. , Zhang, J. , Chen, H. , Chen, B. , Li, Y. , Wang, W. , Kong, Y. , & Wen, G. (2019). Ventral and dorsal visual pathways exhibit abnormalities of static and dynamic connectivities, respectively, in patients with schizophrenia. Schizophrenia Research, 206, 103–110. [DOI] [PubMed] [Google Scholar]
  12. Erdeniz, B. , Serin, E. , Ibadi, Y. , & Taş, C. (2017). Decreased functional connectivity in schizophrenia: The relationship between social functioning, social cognition and graph theoretical network measures. Psychiatry Research: Neuroimaging, 270, 22–31. [DOI] [PubMed] [Google Scholar]
  13. Fornito, A. , Zalesky, A. , Pantelis, C. , & Bullmore, E. T. (2012). Schizophrenia, neuroimaging and connectomics. NeuroImage, 62(4), 2296–2314. [DOI] [PubMed] [Google Scholar]
  14. Friston, K. , Brown, H. R. , Siemerkus, J. , & Stephan, K. E. (2016). The dysconnection hypothesis (2016). Schizophrenia Research, 176(2–3), 83–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Garrity, A. G. , Pearlson, G. D. , McKiernan, K. , Lloyd, D. , Kiehl, K. A. , & Calhoun, V. D. (2007). Aberrant “default mode” functional connectivity in schizophrenia. American Journal of Psychiatry, 164(3), 450–457. [DOI] [PubMed] [Google Scholar]
  16. Gavrilescu, M. , Rossell, S. , Stuart, G. W. , Shea, T. L. , Innes‐Brown, H. , Henshall, K. , McKay, C. , Sergejew, A. A. , Copolov, D. , & Egan, G. F. (2010). Reduced connectivity of the auditory cortex in patients with auditory hallucinations: A resting state functional magnetic resonance imaging study. Psychological Medicine, 40(7), 1149–1158. [DOI] [PubMed] [Google Scholar]
  17. Hare, S. M. , Ford, J. M. , Mathalon, D. H. , Damaraju, E. , Bustillo, J. , Belger, A. , Lee, H. J. , Mueller, B. A. , Lim, K. O. , Brown, G. G. , Preda, A. , van Erp, T. G. M. , Potkin, S. G. , Calhoun, V. D. , & Turner, J. A. (2019). Salience—Default mode functional network connectivity linked to positive and negative symptoms of schizophrenia. Schizophrenia Bulletin, 45(4), 892–901. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Iraji, A. , Chen, J. , Lewis, N. , Faghiri, A. , Zening, F. , Agcaoglu, O. , Kochunov, P. , Adhikari, B. M. , Mathalon, D. H. , Pearlson, G. D. , Macciardi, F. , Preda, A. , van Erp, T. G. M. , Bustillo, J. R. , Díaz‐Caneja, C. M. , Andrés‐Camazón, P. , Dhamala, M. , Adali, T. , & Calhoun, V. D. (2023). Spatial dynamic subspaces encode sex‐specific schizophrenia disruptions in transient network overlap and its links to genetic risk. Biological Psychiatry. 10.1016/j.biopsych.2023.12.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Iraji, A. , Deramus, T. P. , Lewis, N. , Yaesoubi, M. , Stephen, J. M. , Erhardt, E. , Belger, A. , Ford, J. M. , McEwen, S. , Mathalon, D. H. , Mueller, B. A. , Pearlson, G. D. , Potkin, S. G. , Preda, A. , Turner, J. A. , Vaidya, J. G. , van Erp, T. G. M. , & Calhoun, V. D. (2019). The spatial chronnectome reveals a dynamic interplay between functional segregation and integration. Human Brain Mapping, 40(10), 3058–3077. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Iraji, A. , Faghiri, A. , Lewis, N. , Zening, F. , Rachakonda, S. , & Calhoun, V. D. (2021). Tools of the trade: Estimating time‐varying connectivity patterns from fmri data. Social Cognitive and Affective Neuroscience, 16(8), 849–874. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Iraji, A. , Faghiri, A. , Zening, F. , Kochunov, P. , Adhikari, B. M. , Belger, A. , Ford, J. M. , McEwen, S. , Mathalon, D. H. , Pearlson, G. D. , Potkin, S. G. , Preda, A. , Turner, J. A. , Van Erp, T. G. M. , Chang, C. , & Calhoun, V. D. (2022). Moving beyond the ‘cap’ of the iceberg: Intrinsic connectivity networks in fmri are continuously engaging and overlapping. NeuroImage, 251, 119013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Iraji, A. , Faghiri, A. , Zening, F. , Rachakonda, S. , Kochunov, P. , Belger, A. , Ford, J. M. , McEwen, S. , Mathalon, D. H. , Mueller, B. A. , Pearlson, G. D. , Potkin, S. G. , Preda, A. , Turner, J. A. , van Erp, T. G. M. , & Calhoun, V. D. (2022). Multi‐spatial‐scale dynamic interactions between functional sources reveal sex‐specific changes in schizophrenia. Network Neuroscience, 6(2), 357–381. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Iraji, A. , Miller, R. , Adali, T. , & Calhoun, V. D. (2020). Space: A missing piece of the dynamic puzzle. Trends in Cognitive Sciences, 24(2), 135–149. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Iraji, A. , Zening, F. , Faghiri, A. , Duda, M. , Chen, J. , Rachakonda, S. , DeRamus, T. P. , Kochunov, P. , Adhikari, B. M. , Belger, A. , Ford, J. M. , Mathalon, D. H. , Pearlson, G. D. , Potkin, S. G. , Preda, A. , Turner, J. A. , van Erp, T. G. M. , Bustillo, J. R. , Yang, K. , … Calhoun, V. D. (2022). Canonical and replicable multi‐scale intrinsic connectivity networks in 100k+ resting‐state FMRI datasets. Human Brain Mapping, 44(17), 5729–5748. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Jafri, M. J. , Pearlson, G. D. , Stevens, M. , & Calhoun, V. D. (2008). A method for functional network connectivity among spatially independent resting‐state components in schizophrenia. NeuroImage, 39(4), 1666–1681. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Javitt, D. C. , & Sweet, R. A. (2015). Auditory dysfunction in schizophrenia: Integrating clinical and basic features. Nature Reviews Neuroscience, 16(9), 535–550. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Joshi, Y. B. , Thomas, M. L. , Braff, D. L. , Green, M. F. , Gur, R. C. , Gur, R. E. , Nuechterlein, K. H. , Stone, W. S. , Greenwood, T. A. , Lazzeroni, L. C. , MacDonald, L. R. , Molina, J. L. , Nungaray, J. A. , Radant, A. D. , Silverman, J. M. , Sprock, J. , Sugar, C. A. , Tsuang, D. W. , Tsuang, M. T. , … Light, G. A. (2021). Anticholinergic medication burden–associated cognitive impairment in schizophrenia. American Journal of Psychiatry, 178(9), 838–847. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Karbasforoushan, H. , & Woodward, N. D. (2012). Resting‐state networks in schizophrenia. Current Topics in Medicinal Chemistry, 12(21), 2404–2414. [DOI] [PubMed] [Google Scholar]
  29. Karlsgodt, K. H. , Sun, D. , & Cannon, T. D. (2010). Structural and functional brain abnormalities in schizophrenia. Current Directions in Psychological Science, 19(4), 226–231. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Kunert‐Graf, J. M. , Eschenburg, K. M. , Galas, D. J. , Nathan Kutz, J. , Rane, S. D. , & Brunton, B. W. (2019). Extracting reproducible time‐resolved resting state networks using dynamic mode decomposition. Frontiers in Computational Neuroscience, 13, 75. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Lang, X. , Wang, L. , Zhuo, C.‐J. , Jia, F. , Wang, L.‐N. , & Wang, C.‐L. (2016). Reduction of interhemispheric functional connectivity in sensorimotor and visual information processing pathways in schizophrenia. Chinese Medical Journal, 129(20), 2422–2426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Leucht, S. , Rothe, P. , Davis, J. M. , & Engel, R. R. (2013). Equipercentile linking of the BPRS and the PANSS. European Neuropsychopharmacology, 23(8), 956–959. [DOI] [PubMed] [Google Scholar]
  33. Lin, Q.‐H. , Liu, J. , Zheng, Y.‐R. , Liang, H. , & Calhoun, V. D. (2010). Semiblind spatial ica of fmri using spatial constraints. Human Brain Mapping, 31(7), 1076–1088. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Matsui, T. , Murakami, T. , & Ohki, K. (2016). Transient neuronal coactivations embedded in globally propagating waves underlie resting‐state functional connectivity. Proceedings of the National Academy of Sciences, 113(23), 6556–6561. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Menon, V. , Palaniyappan, L. , & Supekar, K. (2023). Integrative brain network and salience models of psychopathology and cognitive dysfunction in schizophrenia. Biological Psychiatry, 94(2), 108–120. [DOI] [PubMed] [Google Scholar]
  36. Miller, R. L. , Yaesoubi, M. , Turner, J. A. , Mathalon, D. , Preda, A. , Pearlson, G. , Adali, T. , & Calhoun, V. D. (2016). Higher dimensional meta‐state analysis reveals reduced resting fmri connectivity dynamism in schizophrenia patients. PLoS One, 11(3), e0149849. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Pini, L. , Jacquemot, C. , Cagnin, A. , Meneghello, F. , Semenza, C. , Mantini, D. , & Vallesi, A. (2020). Aberrant brain network connectivity in presymptomatic and manifest huntington's disease: A systematic review. Human Brain Mapping, 41(1), 256–269. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Sakoğlu, Ü. , Pearlson, G. D. , Kiehl, K. A. , Wang, Y. M. , Michael, A. M. , & Calhoun, V. D. (2010). A method for evaluating dynamic functional network connectivity and task‐modulation: Application to schizophrenia. Magnetic Resonance Materials in Physics, Biology and Medicine, 23, 351–366. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Schwalm, M. , Schmid, F. , Wachsmuth, L. , Backhaus, H. , Kronfeld, A. , Jury, F. A. , Prouvot, P.‐H. , Fois, C. , Albers, F. , van Alst, T. , et al. (2017). Cortex‐wide bold fmri activity reflects locally‐recorded slow oscillation‐associated calcium waves. eLife, 6, e27602. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Sendi, M. S. E. , Zendehrouh, E. , Ellis, C. A. , Liang, Z. , Zening, F. , Mathalon, D. H. , Ford, J. M. , Preda, A. , van Erp, T. G. M. , Miller, R. L. , Pearlson, G. D. , Turner, J. A. , & Calhoun, V. D. (2021). Aberrant dynamic functional connectivity of default mode network in schizophrenia and links to symptom severity. Frontiers in Neural Circuits, 15, 649417. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Skudlarski, P. , Jagannathan, K. , Anderson, K. , Stevens, M. C. , Calhoun, V. D. , Skudlarska, B. A. , & Pearlson, G. (2010). Brain connectivity is not only lower but different in schizophrenia: A combined anatomical and functional approach. Biological Psychiatry, 68(1), 61–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Sui, J. , Qi, S. , van Erp, T. G. M. , Bustillo, J. , Jiang, R. , Lin, D. , Turner, J. A. , Damaraju, E. , Mayer, A. R. , Cui, Y. , Fu, Z. , du, Y. , Chen, J. , Potkin, S. G. , Preda, A. , Mathalon, D. H. , Ford, J. M. , Voyvodic, J. , Mueller, B. A. , … Calhoun, V. D. (2018). Multimodal neuromarkers in schizophrenia via cognition‐guided mri fusion. Nature Communications, 9(1), 3028. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Van Den Heuvel, M. P. , & Fornito, A. (2014). Brain networks in schizophrenia. Neuropsychology Review, 24, 32–48. [DOI] [PubMed] [Google Scholar]
  44. van Erp, T. G. M. , Preda, A. , Turner, J. A. , Callahan, S. , Calhoun, V. D. , Bustillo, J. R. , Lim, K. O. , Mueller, B. , Brown, G. G. , Vaidya, J. G. , et al. (2015). Neuropsychological profile in adult schizophrenia measured with the cminds. Psychiatry Research, 230(3), 826–834. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Vercammen, A. , Knegtering, H. , den Boer, J. A. , Liemburg, E. J. , & Aleman, A. (2010). Auditory hallucinations in schizophrenia are associated with reduced functional connectivity of the temporo‐parietal area. Biological Psychiatry, 67(10), 912–918. [DOI] [PubMed] [Google Scholar]
  46. Wei, Y. , Xue, K. , Yang, M. , Wang, H. , Chen, J. , Han, S. , Wang, X. , Li, H. , Zhang, Y. , Song, X. , & Cheng, J. (2022). Aberrant cerebello‐thalamo‐cortical functional and effective connectivity in first‐episode schizophrenia with auditory verbal hallucinations. Schizophrenia Bulletin, 48(6), 1336–1343. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Woods, S. W. (2003). Chlorpromazine equivalent doses for the newer atypical antipsychotics. Journal of Clinical Psychiatry, 64(6), 663–667. [DOI] [PubMed] [Google Scholar]
  48. Woodward, N. D. , Karbasforoushan, H. , & Heckers, S. (2012). Thalamocortical dysconnectivity in schizophrenia. American Journal of Psychiatry, 169(10), 1092–1099. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Yuhui, D. , & Fan, Y. (2013). Group information guided ica for fmri data analysis. NeuroImage, 69, 157–197. [DOI] [PubMed] [Google Scholar]
  50. Zeng, L.‐L. , Wang, H. , Panpan, H. , Yang, B. , Weidan, P. , Shen, H. , Chen, X. , Liu, Z. , Yin, H. , Tan, Q. , Wang, K. , & Hu, D. (2018). Multi‐site diagnostic classification of schizophrenia using discriminant deep learning with functional connectivity MRI. eBioMedicine, 30, 74–85. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Articles from Human Brain Mapping are provided here courtesy of Wiley

RESOURCES