Disrupted Dynamic Functional Network Connectivity in Fetal Alcohol Spectrum Disorders

Abstract

Background:

Prenatal alcohol exposure (PAE) can result in harmful and long-lasting neurodevelopmental changes. Children with PAE or a fetal alcohol spectrum disorder (FASD) have decreased white matter volume and resting-state spectral power compared to typically developing controls (TDC) and impaired resting-state static functional connectivity. The impact of PAE on resting-state dynamic functional network connectivity (dFNC) remains unknown.

Methods:

Using eyes-closed and eyes-open magnetoencephalography (MEG) resting-state data, global dFNC statistics and meta-states were examined in 89 children aged 6–16 years (51 TDC, 38 FASD). Source analyzed MEG data were used as input to group spatial independent component analysis to derive functional networks from which the dFNC was calculated.

Results:

During eyes-closed, relative to TDC, participants with FASD spent a significantly longer time in state 2, typified by anticorrelation (i.e. decreased connectivity) within and between default mode network (DMN) and visual network (VN), and state 4, typified by stronger internetwork correlation. The FASD group exhibited greater dynamic fluidity and dynamic range (i.e. entered more states, changed from one meta-state to another more often, traveled greater distances) than TDC. During eyes-open, TDC spent significantly more time in state 1, typified by positive intra and interdomain connectivity with modest correlation within frontal network (FN), while participants with FASD spent a larger fraction of time in state 2, typified by anticorrelation within and between DMN and VN and strong correlation within and between FN, attention network, and sensorimotor network.

Conclusions:

There are important resting-state dFNC differences between children with FASD and TDC. Participants with FASD exhibited greater dynamic fluidity and dynamic range and spent more time in states typified by anticorrelation within and between DMN and VN, and more time in a state typified by high internetwork connectivity. Taken together, these network aberrations indicate that prenatal alcohol exposure has a global effect on resting-state connectivity.

1. Introduction

Alcohol consumption during pregnancy is a leading, yet preventable, cause of neurodevelopmental disorders in the United States. Recent estimates of alcohol use in pregnancy are between 13.5% (Gosdin et al., 2022) to 30.3% (Ethen et al., 2009); however, in early pregnancy the rate may be as high as 24.4% in the general obstetrics population (Bakhireva et al., 2018). Prenatal alcohol exposure (PAE) can result in a range of developmental insults depending on amount of alcohol consumed, when in pregnancy the exposure occurred, and how often in pregnancy alcohol was consumed. Children with PAE often have long-lasting changes in physical, behavioral, and/or cognitive functioning. An umbrella term for disorders due to PAE is fetal alcohol spectrum disorders (FASD) and it includes children with fetal alcohol syndrome (FAS), partial fetal alcohol syndrome (pFAS), alcohol-related neurodevelopmental disorder (ARND), and neurobehavioral disorder associated with prenatal alcohol exposure (ND-PAE).

Several neurodevelopmental differences are reported in children with an FASD. Structural decreases have been reported in corpus callosum thickness (Riley et al., 1995), white matter integrity (Lebel et al., 2008, Sowell et al., 2008), and white matter tract connectivity derived from diffusion tensor imaging fractional anisotropy (Wozniak et al., 2006, Lebel et al., 2008, Sowell et al., 2008, Paolozza et al., 2017), including corticothalamic tracts (Stephen et al., 2021). Alterations in resting-state static functional network connectivity (sFNC) have also been reported in children with FASD using fMRI (Wozniak et al., 2011a, Santhanam et al., 2011, Wozniak et al., 2017). The default mode network (DMN), the cortical network highly connected during rest, had reduced connectivity in FASD when compared to controls (Santhanam et al., 2011). Interhemispheric connectivity has also been found to be impaired in children with PAE (Wozniak et al., 2011b), along with overall reduction in path length and global efficiency (Wozniak et al., 2013). Other evidence of altered local dynamics is the increased entropy in the right hemisphere of children with FAS (Candelaria-Cook et al., 2021) and significantly reduced alpha peak frequency in children and adolescents with PAE/FASD (Candelaria-Cook et al., 2022a). While differences in sFNC are important to highlight, group differences in dynamic functional network connectivity (dFNC) have not yet been studied.

Functional brain connectivity most commonly uses ‘static’ connectivity or sFNC to determine connectivity patterns in fMRI resting-state data (Iraji et al., 2021). The sFNC approach ascertains connectivity patterns based on a single connectivity or correlation value across the time analyzed which can often be as long as 5–10 minutes. While this approach has yielded reliable resting state networks (Li et al., 2018, Franco et al., 2013, Dinis Fernandes et al., 2020, Candelaria-Cook and Stephen, 2020), it is more likely that functional connectivity changes over short segments of time due to dynamic (time-varying) brain oscillations and that network correlations are not static (Sakoğlu et al., 2010, Iraji et al., 2021). The ‘dynamic’ connectivity or dFNC approach utilizes smaller time windows (~2–10 seconds) to examine temporal scale changes, evaluates how the interactions of functional sources change over time and tracks how brain networks vary over time producing different brain states. dFNC can reveal key differences in connectivity patterns, number of states entered, or the rate of change between states when clinical groups are compared. The approach has been successful in evaluating group differences including fMRI and MEG clinical data in patients with schizophrenia (Sanfratello et al., 2019a, Sanfratello et al., 2019b, Sakoğlu et al., 2010) and fMRI rest in children and adolescents (Agcaoglu et al., 2020). Due to the higher temporal resolution of MEG data, dFNC state metrics have been shown to have higher reliability than static connectivity metrics (Dimitriadis et al., 2018). Using dFNC to examine brain dynamics may reveal important differences due to PAE.

The current study was designed to examine resting-state dFNC in children with FASD and TDC. This is one of the first neurophysiological studies examining dFNC in FASD using MEG, as previous research used the sFNC approach and the fMRI modality. Given that previous research had shown children with FASD had decreased DMN and interhemispheric connectivity, we anticipated children with FASD would show group differences in dFNC here. Using eyes-closed and eyes-open MEG resting-state data, global dFNC statistics and meta-states were examined in 89 children 6 to 16 years of age (51 TDC, 38 FASD). Source analyzed MEG data were used as input to group spatial independent component analysis (ICA) to derive functional networks from which the dFNC was determined.

2. Materials and Methods

2.1. Participants

The present study included 89 children and adolescents, 38 participants diagnosed with an FASD and 51 age and sex matched typically developing controls, ranging in age from 6 to 16 years of age. Participants with an FASD were recruited from the Prenatal Exposures Clinic within the Center for Development and Disability at the University of New Mexico Health Sciences Center (UNM HSC). FASD classification was based on consensus and assessment from an interdisciplinary team with a clinical psychologist, neuropsychologist, and pediatrician. The prenatal alcohol exposure-related diagnosis including FAS, pFAS, or ARND was made following the diagnostic criteria laid out by Hoyme et al. (2016) and Stratton et al. (1996). The FASD participants were from the following FASD subgroups: 11 FAS, 8 pFAS, 15 ARND, and 4 PAE. Participants listed as PAE were children without a diagnosis but with confirmed prenatal alcohol exposure. Maternal alcohol consumption was confirmed either through direct confirmation by maternal interview, eyewitness reports of maternal drinking during pregnancy, or legal records confirming alcohol consumption during pregnancy (e.g. DWI arrest). Information on maternal alcohol consumption during pregnancy is collected as a part of the FASD clinical assessment; however, accurate estimates of quantity of alcohol consumption during pregnancy are often not available. TDC did not have known prenatal exposure to alcohol or other substances; nor did they have histories of developmental delays or neurological or psychological problems. For the present study, data were pooled from 2 separate studies with similar exclusion and inclusion criteria (Candelaria-Cook et al., 2021, Stephen et al., 2021, Candelaria-Cook et al., 2022a). For data to be included in the current analysis, participants with 1 minute of continuous movement-free, artifact-free data were included. Due to this requirement, 6 FASD subjects and 2 TDC subjects were scanned but not included in the study. The research protocol and procedures were approved by the UNM HSC Human Research Review Committee and were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments on ethical standards. Informed consent/assent was obtained from all participants included in this study and their parents.

2.2. MEG Data Acquisition

MEG data were collected at the Mind Research Network in a magnetically shielded room using a whole-cortex 306-channel MEG system (Elekta MEGIN). Prior to data acquisition, two electrooculogram electrodes were placed near the left and right eye, above the left eyebrow and lateral to the right outer canthi, and two electrocardiogram electrodes were placed below the left and right clavicles. Data from those electrodes were used to capture eyeblink, eye movement and heartbeat artifact. Three-dimensional digitization equipment (Polhemus FastTrack) was used to register the location of four head position indicator (HPI) coils located on the mastoid bone and upper forehead. The HPI coils allow for measurement of the head position within the MEG helmet during data collection. Using the digitization equipment, the head coordinate system was defined by the left and right preauricular and nasion fiducial points; additional general head shape points were collected to optimize the registration of the MEG and MRI coordinate systems. During the scan, children under 8 years old were in supine position during recordings, while adolescents sat upright in the scanner, both were monitored by audio and video links in the control room. Data were sampled at either 1000 Hz or 1200 Hz with an acquisition passband of 0.1–330 Hz. Continuous HPI monitoring (cHPI) was enabled to allow for head movement correction.

All participants had a one hour MEG visit which included sessions of eyes-closed and eyes-open resting-state tasks. Each phase of rest contained unique data triggers. For eyes-closed data, participants were instructed to rest quietly with their eyes closed. For eyes-open data, participants were instructed to rest quietly and look at a white fixation cross that was present on the center of the screen.

2.3. Structural MRI Data Acquisition

Sagittal T1-weighted anatomical MR images were obtained using a 3T Siemens Triotim MRI system with a 32-channel head coil or a 3T Siemens Prisma MRI system with a 32-channel head coil. The parameters of the multiecho 3D MPRAGE sequence on the Siemens Triotim were as follows: TR = 2530 ms, TE= 1.64, 3.5, 5.36, 7.22, 9.08 ms, TI= 1200 ms, flip angle = 7°, field of view (FOV) = 256 mm x 256 mm, matrix = 256 × 256, resolution= 1.0 × 1.0 × 1.0 mm, 192 slices, GRAPPA acceleration factor = 2. The parameters of the MPRAGE sequence on the Siemens Prisma were as follows: TR = 2500 ms, TE= 2.88 ms, TI= 1060 ms, flip angle = 8°, field of view (FOV) = 256 mm x 256 mm, matrix = 256 × 256, resolution=1.0 × 1.0 × 1.0 mm, 176 slices, parallel imaging = 2x.

2.4. MEG Data Processing

Raw MEG data were processed with Neuromag Max-Filter 2.1 or 2.2 software using the temporal extension of signal space separation (t-SSS) with movement compensation (Taulu and Kajola, 2005, Taulu and Hari, 2009). Signal space projection (SSP) (Uusitalo and Ilmoniemi, 1997) was used to remove heartbeat and eye-blink artifacts in MNE software (Gramfort et al., 2013b). To create a 1 minute continuous segment of movement-free, artifact-free data for each resting state, data were processed with 60 second epochs with rejection thresholds set to 6 picotesla for magnetometers and 4 picotescla for gradiometers. Data were visually scanned to ensure the 1 minute segment was artifact free, then the source time course was cropped and the first minute of continuous artifact free data was used for the remaining source analysis. To provide a metric of data quality for Table 1, the number of 2 second epochs rejected and retained with the same rejection thresholds for the duration of the recording is shown. This data is provided solely for the purposes of showing overall data quality between groups, as a 1 minute continuous segment of movement-free, artifact-free data for each resting state, which was equivalent between groups was used for remaining analyses.

Table 1:

Participant Characteristics

	TDC (Mean ±SEM)	PAE/FASD (Mean ±SEM)
Demographics
Sex (M/F)	25/26	20/18
Male average age (years)	9.43 ±0.54	9.45 ±0.54
Male age range	6.00–16.08	6.42–14.58
Female average age (years)	9.29 ±0.42	10.19 ±0.57
Female age range	6.17–15.75	6.83–16.08
Barrett SES*	48.28 ±1.57	30.38 ±2.56
Barrett SES range	18.50–65.50	11.50–66.00
MEG open epochs (% rejected)	2.97 ±0.63	4.57 ±1.11
MEG closed epochs (% rejected)	4.04 ±1.13	2.58 ±0.98

Asterisks represent significant differences between groups

^*(p<0.01).

Source analysis was performed with MNE software (Gramfort et al., 2014, Gramfort et al., 2013a) using dynamic statistical parametric mapping (dSPM) (Dale et al., 2000) as the inverse model. Automatic segmentation of the cortical surface was reconstructed from each participant’s T1-weighted MRI using FreeSurfer. A volume-based source space with a repeatedly subdivided octahedron as the spatial subsampling method, created 4,098 locations per hemisphere with a source space of 4.9 mm. For the dSPM approach, the regularization parameter corresponded with a signal-to-noise ratio of 3, source orientation had a 0.2 loose constraint, and the depth weighting was set to 0.8. A single layer boundary element model was used to create the forward solution. The dSPM inverse model identified where the estimated current at each cortical surface vertex differed significantly from the baseline noise, defined as empty room data. For processing efficiency, spatiotemporal source distribution maps were downsampled to 100 Hz sampling rate which provided 1–50 Hz frequency range. The source timecourse was cropped to the first 60 seconds of artifact-free, continuous data for each rest condition, morphed to MNI space, then converted to NIFTI format.

Preprocessed MEG source analyzed data were used to derive estimates of functional network connectivity (FNC) among networks. Using the GIFT toolbox (http://trendscenter.org/softwawre/gift), group spatial independent component analysis (ICA) was applied to MEG resting-state data, similar to the approach taken by (Sanfratello et al., 2019a, Sanfratello et al., 2019b) and (Houck et al., 2017). Subject-specific maps and timecourses were estimated using a back-reconstruction approach based on PCA compression and projection (Calhoun et al., 2001, Erhardt et al., 2011). MEG resting-state data were decomposed into 50 spatially independent components using the infomax algorithm (Bell and Sejnowski, 1995) and repeated 20 times to ensure stability in ICASSO (Himberg et al., 2004). The 50 ICA components were assessed for quality and non-artifactual components were retained for further analysis. The final number of components selected was determined based upon the following requirements: 1) networks were not lost by component reduction, and 2) a single area of activation was not broken up into numerous components, as detailed in (Sanfratello et al., 2019a). Furthermore, individual components were assessed qualitatively and components were removed when peak activation was situated in white matter, ventricles, or deep brain regions. Low quality components which were not focal or did not have visible peak activation were also removed. Components which were retained exhibited 1) peak activation in gray matter, 2) low spatial overlap with known artifact regions, and 3) locations within networks of interest. Artifactual components, those located in ventricles, peak activation voxel located in white matter or deep brain regions, and low quality components were removed from analysis. Location of component peak activation in MNI space was determined with the GIFT toolbox and component quality was examined using Freesurfer’s Freeview. Using the R label4MRI package (http://github.com/yunshiuan/label4MRI), component peak activation in MNI space was converted to closest brain region in the Automated Anatomical Labeling (AAL) atlas and Brodmann area (BA) regions. After evaluation, the components were manually grouped based on their anatomical and functional properties into resting-state networks defined as auditory network (AN), sensorimotor network (SN), visual network (VN), default-mode network (DMN), attention network (ATN), and frontal network (FN).

The resulting FNC components were further processed with the GIFT toolbox to derive estimates of dynamic functional network connectivity (dFNC). Using k-means clustering, derived from the distance methods of correlation, a meta-state analysis was performed on the defined resting-state networks. For 60 sec of 100 Hz data, which provided data in the 1–50 Hz frequency range with a TR=0.01 s, a 2 second window length was chosen, allowing for the investigation of 5,800 windows in total. We investigated a range of cluster states (2–7 states) until no additional cluster states were populated. We found that six cluster states were optimal for eyes-closed data, while two cluster states were optimal for eyes-open data. Group differences were first examined by comparing mean dwell time and fraction of time. Mean dwell time was defined as the average time a participant remains in each state before changing to another state. Here, time was represented as number of windows, therefore mean dwell time was the average number of windows a participant remained in each state before changing to another state. Fraction of time was defined as the percent of time a participant remains in each state over the 60 seconds or the percentage of occurrences. Group differences were further examined with global meta-state statistics first described in Miller et al. (2014, 2016). The global meta-state metrics used are representations of how much time a subject spent in each cluster state, such as number of states, change between states, state span, and total distance. For meta-state comparisons the number of states was defined as number of unique meta-states for each subject, changes between states was defined as the number of times each subject changed from one meta-state to another, state span was defined as the maximum L1 distance between the two most divergent meta-states for each subject, and total distance was defined as the summed L1 distance between successive meta-states for each subject.

Statistics were performed using SPSS, version 28 for Macintosh (IBM), on sample demographic data and summary k-means cluster information extracted from the GIFT toolbox. Two-sample t-tests were also calculated within the GIFT toolbox and validated with SPSS. Global Meta-state metrics used false discovery rate (FDR) correction to adjust for multiple comparisons. To examine the impact of potential confounding variables (age, sex, and SES), information on dwell time and fraction of time were analyzed with repeated measures analysis of variance (RM-ANOVAs) with statistical thresholds set at p<0.05. Greenhouse-Geisser corrections were made when violations of sphericity occurred. For eyes-closed resting data, the RM-ANOVAs for dwell time and fraction of time included the between-subject factors of Group (TDC, FASD) and Sex (Female, Male), the within-subject factors of Meta-state (States 1–6), and covariates of Age at scan and Barrett SES. For eyes-open resting data, the RM-ANOVAs for dwell time and fraction of time included the between-subject factors of Group (TDC, FASD) and Sex (Female, Male), the within-subject factors of Meta-state (States 1–2), and covariates of Age at scan and Barrett SES. Significant interactions were followed up with separate one-way ANOVAs with FDR correction to adjust for multiple comparisons.

3. Results

3.1. Demographics

Sample demographic information is presented in Table 1. There were no group differences in age or sex, p>0.05; therefore, the sample was well matched in terms of male and female average age and sex. In terms of socioeconomic status (SES), as measured by the Barrett SES measure, the FASD group had significantly lower SES when compared to TDC, F(1,73)=39.410, p<0.001.

3.2. MEG independent components

Of the 50 components modeled with group ICA, 43 components were retained in the eyes-closed condition and 44 were retained in the eyes-open condition. The resting-state networks grouped based on their anatomical and functional properties into AN, SN, VN, DMN, ATN, and FN for each resting-state are shown in Figure 1. The peak activation location for each component is listed in Tables 2 and 3. Peak activation coordinates were determined from within the GIFT toolbox and listed in MNI space.

Figure 1: MEG resting-state ICA component maps for each resting condition.

Individual components are grouped based on their anatomical and functional properties into auditory network (AN), sensorimotor network (SN), visual network (VN), default mode network (DMN), attention network (ATN), and frontal network (FN). Each color within each network represents a different component. There were 43 eyes-closed components: 3 AN, 5 SN, 4 VN, 11 DMN, 7 ATN, and 13 FN. There were 44 eyes-open components: 4 AN, 7 SN, 2 VN, 14 DMN, 4 ATN, and 13 FN.

Table 2:

MEG RSN Peak Activations, Eyes Closed

			Peak Coordinates
RSN	AAL Region	Brodmann Area	x	y	z
Auditory
IC 4	L superior temporal gyrus	L-BA22	−52.5	1.5	−12.5
IC 17	R superior temporal gyrus	R-BA22	53.5	2.5	−2.5
IC 36	L superior temporal gyrus	L_PrimSensory1	−61.5	−18.5	12.5
Sensorimotor
IC 1	R rolandic operculum	R-BA44	52.5	7.5	12.5
IC 18	R postcentral gyrus	R-BA6	53.5	−2.5	23.5
IC 27	R postcentral gyrus	L-Primotor4	−52.5	−13.5	38.5
IC 30	L supplementary motor area	L-BA8	−2.5	26.5	47.5
IC 39	L rolandic operculum	L-BA6	−56.5	2.5	13.5
Visual
IC 2	L middle temporal gyrus	L-BA39	−56.5	−47.5	22.5
IC 9	L inferior temporal gyrus	L-BA21	−56.5	−7.5	−26.5
IC 34	L inferior temporal gyrus	L-Fusiform37	−56.5	−57.5	−12.5
IC 40	L inferior temporal gyrus	L-BA21	−61.5	−32.5	−16.5
Default Mode
IC 5	R anterior cingulate	R-BA10	3.5	51.5	18.5
IC 7	L fusiform	L_Parahip36	−31.5	−22.5	−31.5
IC 8	L temporal pole, superior temporal	L-BA38	−47.5	12.5	−21.5
IC 14	L insula	L-Insula13	−27.5	16.5	−6.5
IC 20	R inferior frontal gyrus, pars orbitalis	R-BA47	47.5	17.5	−6.5
IC 24	L inferior frontal gyrus, pars orbitalis	L-BA47	−46.5	21.5	−6.5
IC 29	L middle occipital gyrus	L-BA19	−46.5	−72.5	3.5
IC 35	L temporal pole, superior temporal	L-BA6	−52.5	6.5	2.5
IC 38	R middle cingulate	R-BA8	3.5	37.5	33.5
IC 43	R parahippocampal	R-Parahipp36	23.5	−22.5	−16.5
IC 44	L superior frontal gyrus, medial	L-BA9	−12.5	37.5	33.5
Attention
6	R middle temporal gyrus	R-BA22	52.5	2.5	−16.5
12	R middle frontal gyrus	R-BA8	27.5	12.5	47.5
23	R superior frontal gyrus, dorsolateral	R-BA8	17.5	27.5	38.5
26	L inferior frontal gyrus, triangularis	L-BA8	−42.5	17.5	32.5
32	R middle temporal gyrus	R-BA22	58.5	−38.5	8.5
37	R middle temporal gyrus	R-BA22	62.5	−18.5	−6.5
45	L middle frontal gyrus	L-BA8	−26.5	22.5	38.5
Frontal
3	L superior frontal gyrus, dorsolateral	L-BA10	−41.5	6.5	−31.5
10	R middle frontal gyrus	R-BA10	27.5	46.5	13.5
11	L inferior frontal gyrus, triangularis	L-BA46	−46.5	26.5	18.5
16	L superior frontal gyrus, dorsolateral	L-BA6	−21.5	16.5	52.5
22	R inferior frontal gyrus, triangularis	R-BA45	48.5	22.5	7.5
25	R inferior frontal gyrus, triangularis	R-BA9	42.5	27.5	18.5
28	L middle frontal gyrus	L-BA9	−32.5	31.5	27.5
33	L inferior frontal gyrus, triangularis	L-BA45	−47.5	26.5	7.5
42	R middle frontal gyrus	R-BA9	37.5	21.5	33.5
47	R middle frontal gyrus	R-BA9	23.5	36.5	28.5
48	R inferior frontal gyrus, triangularis	R-BA46	43.5	36.5	2.5
49	L superior frontal gyrus, dorsolateral	L-BA10	−17.5	47.5	18.5
50	L inferior frontal gyrus, triangularis	L-BA46	−36.5	41.5	13.5

Location of component peak activation in MNI space determined with GIFT toolbox. Closest brain region corresponding to location of peak activation are listed within Automated Anatomical Labeling (AAL) and Brodmann area (BA) regions. Coordinate = max coordinate (mm) in MNI space. Other regions may be present in an IC. Abbreviations: RSN, resting-state network, IC, independent component; L, left; R, right; BA, Brodmann area; AAL, automated anatomical labeling

Table 3:

MEG RSN Peak Activations, Eyes Open

			Peak Coordinates
RSN	AAL Region	Brodmann Area	x	y	z
Auditory
IC 1	R superior temporal gyrus	R-BA22	53.5	2.5	−1.5
IC 10	L superior temporal gyrus	L-BA22	−52.5	1.5	−11.5
IC 13	R superior temporal gyrus	R-BA22	58.5	−42.5	13.5
IC 37	R superior temporal gyrus	R-BA22	62.5	−22.5	−2.5
Sensorimotor
IC 2	L rolandic operculum	L-BA6	−56.5	2.5	12.5
IC 4	L precentral gyrus	L-BA6	−51.5	7.5	28.5
IC 5	R rolandic operculum	L-BA44	57.5	1.5	12.5
IC 7	R postcentral gyrus	R-BA6	12.5	12.5	53.5
IC 18	L rolandic operculum	L-BA44	−51.5	7.5	2.5
IC 33	R postcentral gyrus	R-BA6	52.5	−3.5	33.5
IC 35	L postcentral gyrus	L-PrimSensory1	−56.5	−17.5	33.5
Visual
IC 27	L inferior temporal gyrus	L-BA21	−61.5	−27.5	33.5
IC 34	L inferior temporal gyrus	L-Fusiform37	−56.5	−53.5	−16.5
Default Mode
IC 6	L middle occipital gyrus	L-BA19	−46.5	−72.5	2.5
IC 8	L anterior cingulate	L-BA24	−2.5	22.5	22.5
IC 11	R inferior frontal gyrus, pars orbitalis	R-BA47	47.5	21.5	−2.5
IC 12	L insula	L-insula13	−31.5	16.5	7.5
IC 19	L temporal pole, superior temporal	L-BA36	−46.5	16.5	−16.5
IC 21	L inferior frontal gyrus, pars orbitalis	L-BA47	−46.5	21.5	−6.5
IC 22	L middle temporal gyrus	L-BA38	−46.5	6.5	−26.5
IC 23	L middle temporal gyrus	L-BA21	−61.5	−42.5	2.5
IC 26	R anterior cingulate	R-BA10	2.5	51.5	17.5
IC 28	R hippocampus	R-Hipp54	22.5	−13.5	−12.5
IC 30	R middle cingulate	R-BA9	8.5	41.5	32.5
IC 39	R superior frontal gyrus, medial	R-BA8	8.5	31.5	42.5
IC 42	R parahippocampal gyrus	R-Parahipp36	23.5	−42.5	−2.5
IC 49	L superior frontal gyrus, medial	L-BA9	−11.5	37.5	33.5
Attention
IC 15	L supplementary motor area	L-BA8	−7.5	26.5	47.5
IC 25	R middle temporal gyrus	R-BA22	52.5	2.5	−16.5
IC 31	R middle frontal gyrus	R-BA8	28.5	21.5	43.5
IC 40	L middle temporal gyrus	L-BA22	−61.5	−17.5	2.5
Frontal
IC 3	R inferior frontal gyrus, triangularis	R-BA9	47.5	21.5	23.5
IC 9	R middle frontal gyrus	R-BA9	37.5	36.5	18.5
IC 14	L middle frontal gyrus	L-BA9	−31.5	22.5	37.5
IC 17	L superior frontal gyrus, dorsolateral	L-BA10	−17.5	47.5	18.5
IC 20	L inferior frontal gyrus, triangularis	L-BA45	−47.5	26.5	7.5
IC 29	R inferior frontal gyrus, triangularis	R-BA46	42.5	37.5	2.5
IC 32	L middle frontal gyrus	L-BA6	−21.5	12.5	53.5
IC 38	L middle frontal gyrus	L-BA9	−31.5	32.5	27.5
IC 43	L inferior frontal gyrus, triangularis	L-BA46	−36.5	42.5	12.5
IC 44	R superior frontal gyrus, dorsolateral	R-BA10	23.5	47.5	12.5
IC 45	R middle frontal gyrus	R-BA9	27.5	31.5	32.5
IC 58	R inferior frontal gyrus, triangularis	R-BA45	52.5	21.5	8.5
IC 20	L inferior frontal gyrus, triangularis	L-BA46	−42.5	27.5	22.5

Location of component peak activation in MNI space determined with GIFT toolbox. Closest brain region corresponding to location of peak activation are listed within Automated Anatomical Labeling (AAL) and Brodmann area (BA) regions. Coordinate = max coordinate (mm) in MNI space. Other regions may be present in an IC. Abbreviations: RSN, resting-state network, IC, independent component; L, left; R, right; BA, Brodmann area; AAL, automated anatomical labeling.

3.3. Eyes-closed MEG dynamic functional connectivity

The temporal characteristics in eyes-closed resting-state data were stable with 6 k-means cluster states for both TDC and FASD participants. The number of participants which entered each state within 1 minute of eyes-closed data is as follows: 57 participants entered state 1, 40 participants entered state 2, 12 participants entered state 3, 15 participants entered state 4, 43 participants entered state 5, 60 participants entered state 6. The number of individuals per group per state is presented in Figures 2a and 2b, along with group average dFNC organized by network as correlation grids and overall state connectograms showing correlation between components. States 1, 5 and 6 display similar levels of positive intra and interdomain connectivity and are the 3 states TDC spent over 86% of their time. State 2 is characterized by anticorrelation within and between DMN and other networks, as well as VN. State 3 is characterized by strong correlation within and between SN and FN and other networks, but strong anticorrelation within and between VN and DMN and other networks. State 4 is characterized by strong correlation within and between all networks.

Figure 2a: Eyes-Closed Cluster States (1–3) for TDC and FASD.

MEG k-means cluster states derived from ICA components for TDC and FASD groups during the eyes-closed resting-state. At the top, states 1–3 are represented with an overall connectogram, color coded by network. The networks are organized as follows: AN, blue; SN, green; VN, red; DMN, orange; ATN, purple; and FN, yellow. Component to component correlations are shown, with cool colors representing strongly anticorrelated components and hot colors representing strongly correlated components. Relationships are shown between and within networks. The location of each component is displayed along the edge of the connectogram. Immediately below are dFNC results, organized by group, with network correlation grids (components x components). The number of participants who entered each state is indicated at the top of each correlation grid.

Figure 2b: Eyes-Closed Cluster States (4–6) for TDC and FASD.

MEG k-means cluster states derived from ICA components for TDC and FASD groups during the eyes-closed resting-state. At the top, states 4–6 are represented with an overall connectogram, color coded by network. The networks are organized as follows: AN, blue; SN, green; VN, red; DMN, orange; ATN, purple; and FN, yellow. Component to component correlations are shown, with cool colors representing strongly anticorrelated components and hot colors representing strongly correlated components. Relationships are shown between and within networks. The location of each component is displayed along the edge of the connectogram. Immediately below are dFNC results, organized by group, with network correlation grids (components x components). The number of participants who entered each state is indicated at the top of each correlation grid.

Participants with FASD spent significantly more time in state 2 than TDC, as measured by dwell time and percentage of occurrences, see Table 4. Participants with FASD also remained in state 4 for a significantly longer time compared to TDC, as measured by percentage of occurrences. This means participants with FASD spent significantly more time in both a state typified by anticorrelation within and between DMN, VN and other networks (AN, SM, ATN, FN), and a state typified by strong correlation between all networks.

Table 4:

Summary of Eyes-Closed k-means Cluster Statistics

	TDC Mean ±SEM	FASD Mean ±SEM	t-value	p-value
Mean Dwell time	(# windows)	(# windows)
State #1	178.32 ±71.07	130.89 ±39.95	0.536	0.594
State #2	66.03 ± 26.11	219.46 ±66.84	−2.340	0.022*
State #3	11.57 ±7.69	36.02 ±22.60	−1.131	0.261
State #4	5.41 ±2.50	271.07 ± 174.08	−1.747	0.845
State #5	633.91 ±241.11	125.89 ±62.57	1.806	0.075
State #6	331.69 ± 142.97	91.49 ±27.23	1.453	0.150
Total % of occurrences
State #1	28 % ±6	22 % ±6	0.615	0.540
State #2	12 % ±4	28 % ±7	−2.050	0.044*
State #3	2 % ±2	4 % ±3	−0.667	0.507
State #4	0.2 % ±0.1	9 % ±5	−2.200	0.031*
State #5	28 % ±6	14 % ±5	1.625	0.108
State #6	30 % ±6	22 % ±6	0.948	0.346
Global meta-state
Number of states	18.511	22.917	−2.298	0.024*
Change between states	180.340	221.111	−2.209	0.030*
State span	6.340	7.667	−3.218	0.002*
Total distance	186.809	228.111	−2.157	0.034*

All t-tests represent TDC-FASD.

^*Indicates significance at p<0.05.

There were also group differences in global meta-state statistics. Participants with FASD entered more states (number of unique windows for each subject), changed states more often (number of times each subject changes from one meta-state to another), had a higher state span (maximum L1 distance between states for each subject), and had a greater total distance (sum of L1 distances between successive meta-states for each subject), see Table 4 and Figure 3. Taken together, these results indicate participants with FASD exhibit increased dynamic fluidity and dynamic range and enter more states more often than TDC.

Figure 3: Eyes-Closed Global Meta-state Metrics for TDC and FASD.

MEG dFNC meta-state metrics showing the impact of prenatal alcohol exposure on dynamic fluidity (A-B) and dynamic range (C-D) in the eyes-closed resting-state data. Data are shown in truncated violin plots with solid lines at median and quartiles, dashed lines at mean and an estimation plot with the difference between group means with 95% confidence intervals. TDC data are shown in gray, FASD data are shown in teal. Asterisk (*) denotes mean difference p<0.05, FDR corrected. (A) Number of meta-states; Participants with FASD show higher dynamic fluidity by entering a higher number of distinct meta-states; (B) Number of timepoints at which subjects change between meta-states; Participants with FASD show higher dynamic fluidity by having a higher number of meta-state changepoints; (C) Distance between the two most divergent meta-states in L1 distance; Participants with FASD show higher dynamic range by having a larger radius hypercube of state space; (D) Total distance traveled as measured by summed L1 distance between successive meta-states; Participants with FASD show higher dynamic range by traveling more overall total distance through the state space.

As a follow-up to these results, we explored the impact of age at scan, SES, and sex on eyes-closed dFNC findings. When dwell time and fraction of time were examined, it was evident that participants dwelled in certain states and spent a different fraction of time in each state, but age at scan, SES, and sex had no impact on findings [Dwell time interaction of Meta-state by Age: F(2.525, 194.388)=4.775, p=0.005; Dwell time Meta-state main effect: F(2.525, 194.388)=4.467, p=0.007; all other effects and interactions for dwell time p>0.157; Fraction of time interaction of Meta-state by Age: F(3.681, 283.429)=4.004, p=0.004; Fraction of time Meta-state main effect: F(3.681, 283.429)=3.418, p=0.012; all other effects and interactions for fraction of time p>0.365]. Prior to multiple comparisons correction, there was a trend for age to confound the dwell time in meta-state 5 (p=0.010) and fraction of time in meta-state 6 (p=0.019), but the trends did not survive FDR correction. Age did impact fraction of time spent in meta-state 5 (p=0.005, uncorrected p-value), but because our main group differences occurred in meta-states 2 and 4, the age effects in meta-state 5 were not explored further. When global meta-state variables (number of states, change between states, state span, and total distance) were examined, SES impacted state span [State span: F(1, 83)=5.469, p=0.022; other meta-state variables and SES p>0.159], but age and sex had no impact on global meta-state variables.

3.4. Eyes-open MEG dynamic functional connectivity

The temporal characteristics in eyes-open resting-state data were stable with 2 k-means cluster states for both TDC and FASD participants. The number of participants which entered each state within 1 minute of eyes-closed data is as follows: 80 participants entered state 1, 51 participants entered state 2. The number of individuals per group per state is presented in Figure 4, along with group average dFNC organized by network as correlation grids and overall state connectogram showing correlation between components. State 1 is characterized by similar levels of positive intra and interdomain connectivity with modest correlation within FN and it is the state TDC and FASD participants spent a majority of their time. State 2 is characterized by strong correlation within and between FN, ATN, SN but anticorrelation within and between VN and DMN.

Figure 4: Eyes-Open Cluster States (1–2) for TDC and FASD.

MEG k-means cluster states derived from ICA components for TDC and FASD groups during the eyes-open resting-state. At the top, states 1–2 are represented with an overall connectogram, color coded by network. The networks are organized as follows: AN, blue; SN, green; VN, red; DMN, orange; ATN, purple; and FN, yellow. Component to component correlations are shown, with cool colors representing strongly anticorrelated components and hot colors representing strongly correlated components. Relationships are shown between and within networks. The location of each component is displayed along the edge of the connectogram. Immediately below are dFNC results, organized by group, with network correlation grids (components x components). The number of participants who entered each state is indicated at the top of each correlation grid.

TDC spent significantly more time in state 1 than participants with FASD, see Table 5, as measured by dwell time and percentage of occurrences. Meanwhile, participants with FASD remained in state 2 for a significantly longer time compared to TDC, as measured by percentage of occurrences. This means TDC spent significantly more time in a state typified by positive intra and interdomain connectivity with modest correlation within FN, while participants with FASD spent a larger fraction of time in a state typified by strong correlation within and between FN, ATN, SN but anticorrelation within and between VN and DMN. There were no group differences in global meta-state statistics during an eyes-open state, although there were trends for participants with FASD to enter more states and have a higher state span.

Table 5:

Summary of Eyes-Open k-means Cluster Statistics

	TDC Mean ±SEM	FASD Mean ±SEM	t-value	p-value
Mean Dwell time	(# windows)	(# windows)
State #1	3414.04 ±372.51	2234.11 ±422.84	2.263	0.026*
State #2	387.61 ±199.19	890.30 ±285.38	−1.545	0.126
Total % of occurrences
State #1	82 % ±5	63 % ±7	2.229	0.028*
State #2	18 % ±5	37 % ±7	−2.229	0.028*
Global meta-state
Number of states	2.46	2.69	−1.736	0.086
Change between states	52.64	53.42	−0.120	0.905
State span	2.92	3.39	−1.736	0.086
Total distance	105.28	106.83	−0.120	0.905

All t-tests represent TDC-FASD.

^*indicates significance at p<0.05.

As a follow-up to these results, we explored the impact of age at scan, SES, and sex on eyes-open dFNC findings. When dwell time and fraction of time were examined, it was evident that participants dwelled in certain states and spent a different fraction of time in each state, but age at scan, SES, and sex had no impact on findings [Dwell time Meta-state main effect: F(1.000, 81.000)=5.127, p=0.026; all other effects and interactions for dwell time p>0.094; Fraction of time Meta-state main effect: F(1.000, 81.000)=4.214, p=0.043; all other effects and interactions for fraction of time p>0.072]. When global meta-state variables (number of states, change between states, state span, and total distance) were examined, there was no impact of age at scan, SES, or sex on the findings (p>0.075).

4. Discussion

Here, we have examined resting-state dFNC in FASD and TDC using MEG. Our analyses revealed important resting-state dFNC differences exist in children with FASD. In an eyes-closed resting-state, relative to TDC, participants with FASD spent significantly longer time in state 2, a state typified by anticorrelation within and between DMN and VN, and state 4, a state typified by strong internetwork correlation. At the same time, eyes-closed global meta-state statistics revealed, participants with FASD exhibited increased dynamic fluidity and dynamic range (i.e. entered more states, changed from one meta-state to another more often, traveled greater distances) when compared to TDC. In an eyes-open resting-state, TDC spent significantly more time in state 1, a state typified by positive intra and interdomain connectivity with modest correlation within FN, while participants with FASD spent a larger fraction of time in a state typified by strong correlation within and between FN, ATN, SN but anticorrelation within and between VN and DMN. While there were no group differences in global meta-state statistics during an eyes-open state, there were trends for participants with FASD to enter more states and have a higher state span. When taken together, these results indicate participants with FASD tend to enter more states more often than TDC, showing increased dynamic fluidity and dynamic range, and spend a longer amount of time in states typified by anticorrelation within and between DMN and VN. We did not find evidence here that age at scan, SES, or sex impacted the dFNC findings.

Using the k-means clustering method, we found that 2 states typified dynamic eyes-open resting data, while 6 states typified dynamic eyes-closed resting data. These results show eyes closed and eyes open resting-states capture unique and different information, similar to (Agcaoglu et al., 2020, Agcaoglu et al., 2019, Allen et al., 2018, Candelaria-Cook et al., 2022b). While the Agcaoglu et al. (2019, 2020) papers restrained the group ICA and components for both resting states to a single combined open/closed analysis, we determined there were enough task differences between resting-states to warrant individual analysis, thus allowing networks within each state to be fully revealed. Our results illustrate how different eyes-open and eyes-closed data can be as evidenced by similar but slightly different networks, along with revealing important group differences within each state (i.e. clear evidence of dynamic fluidity and dynamic range differences during the eyes-closed but not eyes-open resting-state). Together this highlights the importance of capturing both resting-states, as looking at one state alone may miss or obscure important group differences.

dFNC has identified stable brain dynamics involving multiple discrete and reoccurring discrete states (Allen et al., 2014, Hansen et al., 2015, Hutchison et al., 2013, Iraji et al., 2021, Du et al., 2018). Within fMRI, dFNC patterns are robust against variations in data quality, analysis approach, grouping, and decomposition methods (Abrol et al., 2017). Furthermore, the dynamic correlations are reliable in terms of test-retest reliability in both the Human Connectome Project and Multimodal MRI Reproducibility Resource datasets (Choe et al., 2017). Although it has been shown that fMRI based dFNC is less reliable than sFNC (Zhang et al., 2018), the opposite may be true of MEG based dFNC where dFNC may be more reliable than sFNC due to the higher temporal resolution of MEG data (Dimitriadis et al., 2018). Regardless, fMRI studies recommend that each type of dFNC metric should be considered separately. For example, while there is an observed similarity of brain states across methods and data sets, metrics derived from the brain states (dwell time and number of change points) were found to be less reliable across two sessions than dFNC mean and variance measures (Choe et al., 2017). Lower reliability for brain state derived metrics may not necessarily be a surprise, but instead highlight that while certain participant traits are relatively stable (i.e. average brain states, number of brain states, connectivity patterns, characterization of network interactions producing each brain states), other participants’ traits are more variable and will naturally vary across days and within session due to physiologic factors such as arousal and attention (i.e. dwell time, number of change points). For the data presented here, a key finding was that participants with FASD enter more states more often than TDC and dwell in states typified by anticorrelation within and between DMN and VN. Since the fMRI dFNC reliability findings suggest state connectivity patterns between groups would have higher reliability than dwell time measures, more research with a test-retest design is needed to determine the MEG dFNC reliability and confirm the reported group differences with FASD. However, we believe our findings may have good reliability due to the artifact free nature of the short scan duration used (1 minute of continuous data) and the fact that shorter resting state scans have been shown to have higher reliability (Teeuw et al., 2021). Furthermore, using a dFNC approach as opposed to a sFNC approach for MEG data yield highly reliability states (Allen et al., 2018, Dimitriadis et al., 2018).

Here, we found altered network interactions in participants with an FASD. These results are in line with previous sFNC results showing children with FASD had decreased DMN and interhemispheric connectivity. While we do not have other dFNC studies in FASD to directly compare to, these findings align well with previous MRI structural and functional findings which reported decreased brain connectivity in individuals with PAE. Studies have reported decreased sFNC in individuals with PAE in the default mode network (Santhanam et al., 2011), frontal parietal network and salience network (Little et al., 2018), and attention networks (Fan et al., 2017), and increased sFNC in the sensorimotor network (Long et al., 2018). These studies have reported both differences in individuals with FASD in strongly correlated (intranetwork) regions (Ware et al., 2021, Fan et al., 2017), and weakly connected (internetwork) regions (Little et al., 2018, Ware et al., 2021). Children and adolescents with PAE have decreased structural white matter connectivity at the whole-brain level such as decreased global efficiency, degree centrality and participant coefficient, along with network alterations in attention, somatomotor, and default brain networks (Long et al., 2020). One of our main findings is that participants with FASD spend more time in states typified by anticorrelation within DMN (i.e. states with decreased DMN connectivity), similar to a study which found that within the posterior cingulate cortex, medial prefrontal cortex and inferior parietal lobules connectivity was reduced in PAE groups at rest (Santhanam et al., 2011). Along with decreased DMN connectivity, we also found participants with FASD had decreased visual network connectivity, paralleling eye movement deficits (Paolozza et al., 2015) and saccade network deficits (Stephen et al., 2013), but also spent a significant amount of time in a state typified by high internetwork connectivity, similar to sensorimotor network increases (Long et al., 2018). All these network aberrations taken together would indicate prenatal alcohol exposure is having a global effect on resting-state connectivity. While the comparisons of our findings with those from sFNC may show some similarities, there will also be many differences between sFNC and dFNC results. Large network changes may be visible in static results, but sFNC does not provide a complete picture of connectivity, as time-varying changes in connectivity captures more subtle and unique information (Agcaoglu et al., 2020).

Here, we also found participants with an FASD had disrupted global meta-state metrics during an eyes-closed resting-state or more specifically exhibited increased dynamic fluidity and dynamic range when compared to TDC. In terms of dynamic fluidity, participants with an FASD occupied more meta-states than TDC and changed from one meta-state to another more often than TDC. In terms of dynamic range, participants with an FASD traveled through more state space than TDC, as measured by both the maximal L1-distance between meta-states and total distance traveled. These dFNC meta-state metrics were first introduced by Miller et al. (2014, 2016). Evidence that these dFNC global meta-state metrics provide unique information for clinical populations has been reported in patients with schizophrenia (Miller et al., 2016, Sanfratello et al., 2019a, Sanfratello et al., 2019b) and typically developing adolescents (Agcaoglu et al., 2020). Here we found dFNC global meta-state metrics provided unique information for participants with an FASD. In contrast to the many reports of decreased sFNC from fMRI data in various networks in participants with an FASD (Santhanam et al., 2011, Little et al., 2018, Fan et al., 2017, Long et al., 2018, Long et al., 2020), the dFNC global meta-state metrics from MEG data reported here indicate increased dynamic fluidity and dynamic range in participants with an FASD. This finding is line with previous FASD studies which have found increases in global measures of cortical network connectivity such as increased characteristic path length in children with PAE (Wozniak et al., 2013, Wozniak et al., 2017), and disruptions in global network segregation and integration (i.e. averaging clustering coefficient characteristic path length and global efficiency) in adolescents the an FASD (Rodriguez et al., 2021). There are important temporal differences between MEG and fMRI resting-state data (i.e. MEG data has a higher temporal resolution and may provide connectivity patterns unique from fMRI) and distinctions between global meta-state information and network correlations (i.e. how participants travel through states versus direct network to network correlations) which should be considered when comparing findings. As such the global metrics of connectivity dynamism presented here are novel and unique, future research is needed to validate the findings.

A potential structural basis for the observed differences in functional connectivity is the weaker white matter fiber connectivity as evidenced by lower FA in FASD groups (Wozniak et al., 2011a, Lebel et al., 2008, Sowell et al., 2008) with DMN deficits resulting from lower FA in cingulum bundles connecting the posterior cingulate cortex and medial prefrontal cortex (Santhanam et al., 2011). The white matter structural differences often are present in long range tracts such as the lateral splenium of corpus callosum (Sowell et al., 2008), bilateral superior longitudinal fasciculus (Lebel et al., 2008), left inferior longitudinal fasciculus (Fan et al., 2016), right inferior longitudinal fasciculus (Green et al., 2013), corticospinal tract (Paolozza et al., 2015), and corticothalamic tracts such as anterior and posterior internal capsule (Stephen et al., 2021). When examined longitudinally, young children with PAE 2–8 years of age showed slower white matter development in frontal and temporal tracts specifically the genu of the corpus callosum, inferior fronto-occipital fasciculus, inferior longitudinal fascicular and uncinate fasciculus (Kar et al., 2022), while older children with PAE between 5–15 years of age showed faster white matter development with greater change (steeper decreases) in mean diffusivity of the inferior longitudinal fasciculus and superior fronto-occipital fasciculus (Treit et al., 2013). Given the number of white matter tracts impacted and dynamic changes during development, prenatal alcohol exposure may affect white matter globally. When individuals with PAE have increases in network connectivity, there is often lower network efficiency which requires overcompensation and recruitment of additional brain regions (Long et al., 2018, Wozniak et al., 2013). Taken together these findings suggest children with PAE demonstrate atypical white matter developmental trajectories with compensatory changes later in development and reduced brain plasticity. A complex combination of white matter alterations and structural connectivity differences, along with altered neurodevelopmental trajectories present in individuals with FASD may in part explain the differences in dynamic functional connectivity found here.

There were a few limitations of the present research which warrant consideration. First, the current study had limited insight into gamma frequency relationships. Due to dFNC analysis constraints and the complexity of MEG data, we used 60 secs of continuous data for each resting-state, down-sampled to 100 Hz with a 10 ms sampling rate. This provided an upper frequency limit of 50 Hz. The 1–50 Hz range includes delta band (1–4 Hz), theta band (5–8 Hz), alpha band (9–13 Hz), beta band (14–29 Hz), and low gamma band (30–50 Hz). Often gamma includes 30–100 Hz data, so the present study is excluding high gamma band (>50 Hz) information. However, we feel the current cutoff provides sufficient information as high gamma tends to be lower amplitude and will thus be less likely to influence the correlation value in this broadband signal. Second, due to the retrospective study design, we were unable to exclude or control for concurrent polysubstance abuse in the individuals with prenatal alcohol exposure; therefore, a confounding effect of other substances requires further research. Furthermore, while we know which participants clinically presented with an FASD diagnosis by ages 6 to 16 years, estimates of alcohol consumption and prenatal timing of alcohol consumption are unavailable. It is known that the effects of PAE on brain development can vary based on the pattern of alcohol consumption or timing of exposure (Savage et al., 2002, Stephen et al., 2021). Third, given the sample size we are unable to explore the impact of various FASD subdiagnoses on dFNC results, although previous reports suggest sub diagnoses impact functional connectivity (Santhanam et al., 2011) and spectral power (Candelaria-Cook et al., 2021) to varying extents. Despite these limitations, the current study had several strengths including an age and sex matched design, along with application of a novel approach for examining dFNC in FASD with complex MEG data.

5. Conclusions

MEG dFNC revealed important resting-state differences in children with FASD. Children and adolescents 6–16 years of age with an FASD tend to enter more states more often than TDC exhibiting increased dynamic fluidity and dynamic range and spend a longer amount of time in states typified by anticorrelation (i.e. decreased connectivity) within and between DMN and VN, and a longer amount of time in a state with high internetwork connectivity, perhaps as a compensatory mechanism. These network aberrations taken together would indicate prenatal alcohol exposure is having a global effect on resting-state connectivity. To the best of our knowledge, this is the first report examining dFNC in children and adolescents with FASD. By leveraging the higher temporal resolution of MEG data, we examined how functional networks changed over time to produce stable and discrete brain states. Our results suggest prenatal alcohol exposure broadly alters resting-state networks. Future work should aim to identify the reliability of dFNC metrics in this clinical population and examine correlations with behavioral and clinical measures. Further understanding how these dynamic meta-state differences contribute to cognitive function will aide in intervention development for individuals with an FASD.