Skip to main content
NCBI home page
Human Brain Mapping logoLink to Human Brain Mapping
. 2024 Sep 4;45(13):e70017. doi: 10.1002/hbm.70017

Abnormal multimodal neuroimaging patterns associated with social deficits in male autism spectrum disorder

Long Wei 1, Xin Xu 1, Yuwei Su 2, Min Lan 2, Sifeng Wang 2, Suyu Zhong 2,✉
PMCID: PMC11372822  PMID: 39230055

Abstract

Atypical social impairments (i.e., impaired social cognition and social communication) are vital manifestations of autism spectrum disorder (ASD) patients, and the incidence rate of ASD is significantly higher in males than in females. Characterizing the atypical brain patterns underlying social deficits of ASD is significant for understanding the pathogenesis. However, there are no robust imaging biomarkers that are specific to ASD, which may be due to neurobiological complexity and limitations of single‐modality research. To describe the multimodal brain patterns related to social deficits in ASD, we highlighted the potential functional role of white matter (WM) and incorporated WM functional activity and gray matter structure into multimodal fusion. Gray matter volume (GMV) and fractional amplitude of low‐frequency fluctuations of WM (WM‐fALFF) were combined by fusion analysis model adopting the social behavior. Our results revealed multimodal spatial patterns associated with Social Responsiveness Scale multiple scores in ASD. Specifically, GMV exhibited a consistent brain pattern, in which salience network and limbic system were commonly identified associated with all multiple social impairments. More divergent brain patterns in WM‐fALFF were explored, suggesting that WM functional activity is more sensitive to ASD's complex social impairments. Moreover, brain regions related to social impairment may be potentially interconnected across modalities. Cross‐site validation established the repeatability of our results. Our research findings contribute to understanding the neural mechanisms underlying social disorders in ASD and affirm the feasibility of identifying biomarkers from functional activity in WM.

Keywords: autism spectrum disorder, multimodal neuroimaging, social impairments

This work revealed the multimodal brain patterns (the salience network and limbic system) to explain the mechanism of autism spectrum disorder social impairment in gray matter (GM) structure and white matter (WM) function information, and WM functional activity was more sensitive to multiple social impairments than GM.

graphic file with name HBM-45-e70017-g003.jpg

Practitioner Points.

  • The salience network and limbic system were consistently associated with multiple social impairments in autism spectrum disorder (ASD).

  • White matter functional activity is more sensitive to multiple social impairments in ASD.

  • Brain regions relating to social impairments existed interconnection across modalities from multimodal perspectives.

1. INTRODUCTION

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder characterized by pervasive social deficits and restricted repetitive behaviors (Hirota & King, 2023; Levy et al., 2009) with significantly higher prevalence in males compared to females (Baio et al., 2018). Currently, the clinical diagnosis of ASD is based on social behavior observation and interview record scales, such as Autism Diagnostic Observation Schedule (ADOS) (Lord et al., 2000) and Social Responsiveness Scale (SRS) (Constantino et al., 2003). However, these scales, which are mainly based on the limited behavioral findings in clinical performance, hardly be used to explain the neuropathogenesis basis and brain representation of impaired social ability (Rafiee et al., 2022). It consequently holds paramount importance to use other methods for discerning the atypical brain patterns related with social impairment, which may provide key insights into the pathogenesis of ASD.

Recently, magnetic resonance imaging (MRI), as a noninvasive method to reveal macroscopic aspects of human brain anatomy and function, could potentially bridge gaps between the behavior and brain imaging representation of social impairments in ASD. Studies based on functional MRI (fMRI) uncovered the atypically functional activity in regions involved in social cognition, for example, speech perception (Redcay & Courchesne, 2008). Besides, there were socialization‐related structural changes in ASD at both gray matter level and white matter (WM) level. For example, Rojas et al. (2006) identified brain regions whose gray matter volume (GMV) associated with socialization and repetitive behavior symptoms in ASD. Vinçon‐Leite et al. (2024) evaluated the correlation between social perception (measured with eye‐tracking) and WM microstructure at the individual scale based on diffusion MRI, and identified some tracts related to social interaction, such as the superior longitudinal tracts. Notably, most of these studies mainly focused on single imaging modality to investigate the brain representation associated with deficient behavior in ASD, while understanding of cross‐multimodal information linking social behavior remains limited (Calhoun & Sui, 2016).

By capturing covariance information between modalities, multimodal fusion provides an attractive opportunity for a more comprehensive reflection of brain status and realizing complementary advantages of different modalities. In recent years, multimodal fusion analysis has been extensively implemented on neurological disorders, such as schizophrenia (Geenjaar et al., 2023; Qi et al., 2018; Yao et al., 2021), major depression disorder (Chen et al., 2023; Maglanoc et al., 2020), and other diseases (Guo et al., 2021; Wang et al., 2023). In particular, multimodal fusion has also been used to extract joint brain information for ASD classification (Du et al., 2020; Ingalhalikar et al., 2014; Pan et al., 2022; Saponaro et al., 2024) or for brain patterns comparison between ASD and health control (Oblong et al., 2023; Peterson et al., 2022). However, there remain two nebulous points. First, most current fusion methods are unsupervised without any prior information (i.e., social score) (Qi et al., 2018), leading to a wide gap between social clinical performance and advanced MRI techniques. Second, even though a few studies have tried to exploit the supervised multimodal fusion methods (Li et al., 2019; Qi et al., 2020) to explore atypical brain patterns of social impairment in ASD, the pattern based on WM function information is unknown. Research has demonstrated that blood oxygenation level dependent signals of WM are also related to brain neural activity (Ding et al., 2018) and exhibit structure‐specific temporal correlations along WM tracts (Ding et al., 2016). These signals contain rich and robust spatiotemporal information that helps to establish the cerebral functional organization more effectively than diffusion tensor imaging (Wang et al., 2022). Hence, there is increasing interest to uncover the brain patterns related with social deficits in ASD from a multi‐view perspective, fusing the function of WM and structural aberration in gray matter, as well as considering the clinical social interaction scores as the prior information.

Here, a two‐way supervised multimodal MRI feature fusion model was guided by SRS multiple domain scores in the public cohort: ABIDE (Autism Brain Imaging Data Exchange, https://fcon_1000.projects.nitrc.org/indi/abide/) (Di Martino et al., 2014). We first explored the aberrant social interaction brain patterns in each modal imaging. Here, six social domains and two multimodal MRI brain features were used. The six social domains are SRS total, SRS awareness, SRS cognition, SRS communication, SRS motivation, and SRS mannerisms. To detect changes in regional spontaneous activity and effectively suppress nonspecific signaling components in resting‐state fMRI, we used fractional amplitude of low‐frequency fluctuations (fALFF), an improved approach over ALFF, as described by Zou et al. (2008). To fuse the changes in gray matter morphology and WM function, GMV based on structural MRI (sMRI) and fALFF of WM signal (WM‐fALFF) from rs‐fMRI were used as multimodal brain features. Then, a model called “MCCAR + jICA” was adopted to identify potential multimodal brain areas associated with social impairment in ASD. Considering the differentiated relationship between specific modal imaging and social domain, we then investigated domain‐specific brain pattern and observed the sensitivity of modal imaging for each social domain. Finally, a leave‐one‐site‐out (LOSO) strategy was adopted as a validation procedure to substantiate the reliability of our results.

2. MATERIALS AND METHODS

2.1. Participants

This study utilized the ABIDE I/II project (Di Martino et al., 2014) as the MRI dataset, comprising 1060 ASD and 1166 healthy control (HC) participants. According to the following inclusion criteria, we finally included 699 subjects (ASD = 343, HC = 356) from 11 sites. The inclusion criteria in the present study were as follows: (1) only male subjects with available SRS scores, sMRI and rs‐fMRI were included, due to the higher ASD incidence in males than females (Baio et al., 2018; Li et al., 2019); (2) individuals with higher full‐scale IQ (FIQ) scores (>70) were included to ensure normal cognition (Koyama et al., 2007); (3) individuals with good segmentation quality and head motion (eliminated 59 participants); and (4) only sites with a sample size of ASD or HC greater than 10 were included to reduce site effects (134 participants were eliminated from 6 sites). Written informed consent was obtained from all participants under protocols approved by the Institutional Review Boards at each study site. Demographic information and symptomatic scores for participants at each site are summarized in Table 1.

TABLE 1.

Demographic and clinical information of participants of each site.

Site Num Age FIQ SRS total score
ASD HC ASD HC p‐Value ASD HC p‐Value ASD HC p‐Value
Mean ± std Mean ± std Mean ± std Mean ± std Mean ± std Mean ± std
ABIDEII—BNI 24 29 35.96 ± 15.94 39.59 ± 15.09 .400 106.92 ± 13.44 112.41 ± 12.07 .123 110.54 ± 28.89 27.97 ± 17.88 <.05
ABIDEII—GU 33 25 11.23 ± 1.53 10.82 ± 1.70 .335 120.94 ± 14.88 121.56 ± 12.07 .865 86.70 ± 36.67 19.72 ± 15.27 <.05
ABIDEII—KKI 32 93 10.56 ± 1.54 10.42 ± 1.29 .606 102.94 ± 14.00 114.59 ± 10.83 <.05 90.81 ± 27.70 17.11 ± 11.17 <.05
ABIDEII—NYU 40 27 9.77 ± 4.72 9.23 ± 1.89 .573 103.95 ± 17.63 115.81 ± 14.90 <.05 83.98 ± 31.41 23.63 ± 13.07 <.05
ABIDEII—OHSU 27 22 12.30 ± 1.88 10.23 ± 1.74 <.05 106.81 ± 17.03 118.18 ± 10.36 <.05 94.04 ± 26.09 20.32 ± 15.32 <.05
ABIDEII—SDSU 25 21 12.82 ± 3.21 13.69 ± 3.11 .358 100.40 ± 13.48 102.90 ± 10.52 .492 105.12 ± 25.62 17.76 ± 10.35 <.05
ABIDEII—TCD 18 21 14.76 ± 3.34 15.61 ± 3.12 .421 110.39 ± 15.40 118.48 ± 13.19 .086 93.89 ± 22.69 20.10 ± 14.86 <.05
ABIDEI—LEUVEN 14 15 21.86 ± 4.11 23.27 ± 2.91 .294 109.43 ± 13.09 114.80 ± 12.86 .275 76.93 ± 21.04 43.60 ± 21.67 <.05
ABIDEI—NYU 65 50 14.01 ± 6.59 12.50 ± 3.08 .137 108.58 ± 16.60 111.54 ± 13.53 .308 90.85 ± 30.47 21.98 ± 13.25 <.05
ABIDEI—USM 52 42 22.44 ± 7.75 21.23 ± 7.69 .450 101.35 ± 15.75 114.86 ± 13.82 <.05 93.15 ± 33.81 15.57 ± 12.79 <.05
ABIDEI—YALE 13 11 13.07 ± 3.02 13.10 ± 3.21 .984 102.15 ± 22.03 105.82 ± 17.37 .660 93.54 ± 24.45 13.55 ± 13.80 <.05
ALL 343 356 15.84 ± 9.37 15.41 ± 9.87 .553 106.64 ± 16.62 114.10 ± 13.03 <.05 92.36 ± 30.44 20.59 ± 14.87 <.05
Open in a new tab

Note: Site: site name; Num: number of subjects; Age: subjects age at scan.

Abbreviations: ASD, autism spectrum disorder; FIQ, full‐scale intelligence quotient; HC, healthy control.; SRS, social responsiveness scale.

2.2. Multimodal image preprocessing and metrics spatial map computation

For each participant, a GMV spatial map was calculated based on 3D T1‐weighed sMRI. The individual T1 imaging was first spatially normalized to Montreal Neurological Institute (MNI) space. Then, the MNI‐spaced T1 imaging was segmented into GM, WM, and cerebrospinal fluid (CSF). Finally, GM segmentations were smoothed using a full‐width half‐maximum (FWHM) of 6 mm Gaussian filter. All these steps were performed using Computational Anatomy Toolbox 12 (https://neuro-jena.github.io/cat/index.html, CAT12) (Gaser et al., 2024) and Statistical Parametric Mapping 12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/, SPM12).

WM‐fALFF spatial maps were obtained from resting‐state fMRI (rs‐fMRI). The preprocessing of the resting‐state fMRI included the following steps: discarding the first 10 volumes, slice timing correction, motion correction, spatial normalization to the MNI space with a voxel size of 3 × 3 × 3 mm3 and spatial smoothing using an FWHM of 6 mm Gaussian kernel. To retain the WM signal, CSF signals and 24 rigid body motion parameters were regressed out. WM‐fALFF spatial maps were computed as the sum of the amplitude values in the 0.01–0.08 Hz low‐frequency range divided by the sum of the amplitudes over the entire power spectrum (Zou et al., 2008). Notably, we excluded subjects who had mean framewise displacements (FDs) exceeding 1 mm (Muschelli et al., 2014) or maximal translation exceeding 5.0 mm (in any direction of x, y, or z) or maximal rotation exceeding 5.0° throughout the course of scanning. There were no significant group differences in mean FD (ASD: mean ± std = 0.13 ± 0.1; HC: mean ± std = 0.12 ± 0.09; two‐sample t test: t = 1.78, p = .08). More details were described in Supplementary Note 1 and Supplementary Table 1. The rs‐fMRI were processed by using Data Processing Assistant for rs‐fMRI (http://rfmri.org/DPARSF, DPARSF) (Yan et al., 2016).

In order to reduce the impact of modality mixing and prepare for the next analysis, we need to perform the data masking and matrices extraction for the above processed multimodal MRI features respectively. For data masking, to minimize mixing the GM and WM signals and potential partial volume effects from nearby structures, all data were retained voxel data by using mask files created from setting 0.7 as the threshold for tissue probability map file from SMP12 (Figure 1a). Regarding matrices extraction, masked brain images of each subject were first converted into Z‐score maps and flattened into one‐dimensional vectors, followed by stacking all vectors into a matrix. To minimize the impact of other factors, age, FIQ, mean FD, and site were regressed out for each modality (Qi et al., 2020).

FIGURE 1.

FIGURE 1

Open in a new tab

Flowchart of social‐directed multimodal fusion and validation. (a) Flowchart of multimodal fusion including multimodal image processing, processing matrices and scores vectors, multimodal fusion and calculating the correlation values between loadings vectors. (b) Leave‐one‐site‐out repeatability. (c) Association with other SRS subdomains scores. IC, independent component; SRS, Social Responsiveness Scale; TMP, tissue probability map.

2.3. Multimodal fusion analysis with reference

Studies have shown that by combining multimodal canonical correlation analysis and joint independent component analysis method (MCCA + jICA) can successfully capture multimodal interactions and high‐precision spatial components for studying brain diseases (Sui et al., 2013). However, this model is unsupervised and cannot focus on studying brain patterns related to a specific measure. In order to identify components which were significantly associated with social behavioral measurement and significantly group‐discriminating in multimodal imaging, a previously proposed “MCCAR + jICA” (MCCA with reference + jICA) analysis model was used (Li et al., 2019; Qi et al., 2018; Sui et al., 2018). Unlike “MCCA + jICA” (Sui et al., 2013), the “MCCAR + jICA” is a supervised model adding prior information (such as social scores) as guiding reference, which can maximize not only the covariance between modalities but also the correlation between the covariation matrix and the reference information as shown in Equation (1).

max∑k,j=12corrAkAj2+2λ·corrAkref22 (1)

where ref is a N × 1 vector, N is the sample size. λ is a hyperparameter used to balance the contribution of reference data in the model (details in Supplementary Note 2 and Figure S1). Ak is a subject‐by‐component loadings mixing matrix that represents the contribution weight for each subject in the corresponding component. The number of components as a hyperparameter determines how many components decomposed from the data matrix. We used the minimum description length criterion (Li et al., 2007) to estimate optimum components number for GMV and WM‐fALFF. More specifically, to maintain the between‐modalities linkages of potential target components and maximize their spatial independence, joint ICA was performed to obtain the potential independent components (ICs) and their subject‐wise loadings. Then the Pearson correlation coefficient between the loadings of each component and the reference was calculated to quantify the relationship between the target component and social impairment. In particular, the p‐values derived from a permutation test as FDR corrected for the correlation between component loading and SRS scores (details provided in Supplementary Note 3 and Figure S2). Meanwhile, we conducted a two‐sample t test on the loadings of each component between ASD and HC. Finally, one target IC is denoted as ICref, balancing significant correlation with social impairment and group differences for all modalities.

2.4. Multimodal networks associated with SRS total score

In order to uncover the neural mapping of social impairments for ASD from a multimodal perspective, two brain features (GMV from sMRI and WM‐fALFF from rs‐fMRI) were combined by fusion analysis model and SRS total score was adopted as reference. Higher SRS total score indicates higher severity of ASD clinical symptoms (Patti et al., 2021). After joint decomposition, SRS‐related joint components were identified (denoted as ICSRS_Total) and their subject‐wise loadings vector were derived for each modality (Sui et al., 2018). For each brain modality, the relationship between the component loadings and SRS total score was quantified by the Pearson correlation coefficient. Moreover, a two‐sample t test was performed on the component loadings between ASD and HC to explore the group differences for each brain modality.

2.5. Multimodal networks association with SRS subdomains score

In order to explore the convergence and divergence in neural among social subdomains, five social reactiveness subdomains were also used as references and the same multimodal fusion analysis was performed for all subdomains. The five social reactiveness subdomains included awareness, cognition, communication, mannerisms and motivation. There were significant group differences between ASD and HC in any social reactiveness subdomains (Supplementary Table 2), indicating that ASD is severely socially impaired (Constantino et al., 2003). Similar to the multimodal fusion analysis with the social total score as reference, the Pearson correlation coefficient of the components and a two‐sample t test of component loadings were performed between subdomains for each modality. Notably, 466 participants (ASD = 213, HC = 253) who had five subdomain scores were used for social subdomain analysis (Figure 1c).

2.6. Repeatability across sites

To validate the repeatability of the results, the LOSO validation strategy (Figure 1b) (Okamoto & Akama, 2021) was used. Specifically, data from one site was dropped at each time and repeated the previous steps to obtain a new IC spatial map for each modality. Based on these new spatial maps, we synthesized a voxel‐based overlapping map for each modality by counting the number of occurrences for each voxel and we call these the validation maps. As an indicator of measuring repeatability, correlation values were calculated between the validation map and the discovery map from our main results above: t higher correlation values indicate better repeatability. Further details are provided in Supplementary Note 4 and Figure S3.

3. RESULTS

3.1. Multimodal joint networks associated with SRS total score

For SRS total score, we identified a target joint component, of which the subject‐wise loadings were most significant differences between groups and most significant correlation with SRS total scores. Figure 2a showed the spatial pattern of this component. The red/blue regions indicate positive/negative correlation between subject‐wise loadings of the component and SRS total score. Multimodal joint patterns related to the SRS total score include a positive correlation network of GMV and WM‐fALFF, and a negative correlation network of GMV and WM‐fALFF. The positive correlation network of GMV was mainly located in the bilateral insula, bilateral middle temporal gyrus, right superior frontal gyrus and left calcarine. The negative correlation network of GMV was mainly located in the bilateral caudate nucleus, bilateral hippocampus, bilateral middle occipital gyrus, bilateral anterior cingulate cortex, right inferior occipital gyrus, as well as the left parahippocampal gyrus. The positive correlation network of WM‐fALFF included regions in bilateral anterior corona radiata, bilateral anterior limb of internal capsule, left external capsule, body and genu of corpus callosum, the negative contribution of WM‐fALFF was in the splenium of corpus callosum and left posterior corona radiata. All regions were summarized in Supplementary Note 5 and Supplementary Tables 3 and 4.

FIGURE 2.

FIGURE 2

Open in a new tab

The identified joint component of SRS total score. (a) The spatial maps are visualized at | z | > 1.5 thresholds, where the red/blue regions indicate positive/negative contributions for the correlation between loadings of the identified components and SRS total score in GMV or WM‐fALFF. (b) The left chart is the scatterplot for the correlation between GMV loadings of IC and SRS total score; the right chart is violin plot for the two‐sample t test on GMV loadings. (c) The left chart is the 27 result of correlation between WM‐fALFF loadings of IC and SRS total score, the right chart is result of a two‐sample t test on WM‐fALFF loadings. * means p < .005.

Figure 2b,c showed that correlation between the component loadings and the SRS total scores for both GMV and WM‐fALFF (GMV: r = .14, p = 2.00 × 10−4; WM‐fALFF: r = −.15, p = 3.00 × 10−4; all passed FDR correction at p < .05). For GMV, the higher the component loadings, the worse the social function. WM‐fALFF had the opposite relationship. Two‐sample t test indicated significant group differences in the loadings of the components for both modalities. More specifically, ASD group was higher (t = 3.73, p = 2.09 × 10−4) in components loadings for GMV and lower in that for WM‐fALFF (t = −5.04, p = 6.01 × 10−7).

3.2. Common and unique brain pattern across multiple social domains

We identified target joint components of communication, mannerisms and motivation subdomains, of which the subject‐wise loadings were most significant differences between groups and most significant correlation with subdomains scores for two modalities. Figure 3a,b showed the result of Pearson correlation and the spatial pattern of these component for GMV and WM‐fALFF, respectively (regions listed in Supplementary Tables 3 and 4). Awareness and cognition were not significantly correlated with the identified component loadings.

FIGURE 3.

FIGURE 3

Open in a new tab

Comparison of multimodal identified independent components. (a) The three subdomain spatial maps visualized at | z | > 1.5 thresholds and the results of correlation between loadings of IC and scores in GMV. (b) The three subdomain spatial maps visualized at | z | > 1.5 thresholds and the results of correlation between loadings of IC and scores in WM‐fALFF. * means p < .005. (c) The correlation matrix between three subdomains components and total component in GMV and WM‐fALFF.

We conducted a comparison of all spatial maps within each modality to identify the common and unique regions. For GMV, the salience network (SAN, including insular, caudate and anterior cingulate cortex) and limbic system (i.e., hippocampus and parahippocampal gyrus) were identified across all social domains. The middle cingulate and paracingulate gyri were uniquely associated with mannerisms. Superior and middle frontal gyrus were detected in communication and motivation. Regarding WM‐fALFF, projection white tracts (i.e., anterior corona radiata, internal capsules, and external capsules) and commissural fibers such as corpus callosum were common across all domains. Long association tracts (i.e., cingulum and superior longitudinal fasciculus) and superior corona radiate were detected uniquely depending on the different domains. For instance, cingulum was common in all domains except motivation. The superior corona radiate was common in all domains except communication. The superior longitudinal fasciculus uniquely identified in mannerisms and motivation.

For communication, the positive correlation network of GMV was mainly located in bilateral insula, bilateral inferior frontal gyrus, bilateral middle temporal gyrus, right superior, and middle frontal gyrus. The negative correlation network of GMV was mainly located in bilateral caudate, bilateral hippocampus, bilateral anterior cingulate cortex, bilateral middle occipital gyrus, right inferior occipital gyrus, and left parahippocampal gyrus. The positive contribution of WM‐fALFF was in splenium and body of corpus callosum, bilateral anterior limb of internal capsule, bilateral posterior limb of internal capsule, bilateral cingulum and left external capsule. The negative correlation network of WM‐fALFF was mainly located in the bilateral anterior corona radiata and genu of corpus callosum.

For mannerisms, the positive correlation network of GMV was mainly located in the bilateral insula, bilateral middle temporal gyrus, and bilateral inferior frontal gyrus. The negative correlation network of GMV was mainly located in the bilateral caudate, bilateral hippocampus, bilateral middle occipital gyrus, right inferior occipital gyrus, right middle cingulate gyrus, and left parahippocampal gyrus. The positive contribution of WM‐fALFF was in the bilateral anterior corona radiata, bilateral superior corona radiata, bilateral superior longitudinal fasciculus, bilateral anterior limb of internal capsule, right posterior limb of internal capsule, and genu of corpus callosum. The negative correlation network of WM‐fALFF was mainly located in the splenium and body of corpus callosum, bilateral cingulum, and left posterior corona radiata.

For motivation, the positive correlation network of GMV was mainly located in the bilateral insula, bilateral middle temporal gyrus, bilateral inferior frontal gyrus, right superior and middle frontal gyrus, and left inferior temporal gyrus. The negative correlation network of GMV was mainly located in the bilateral caudate, bilateral hippocampus, bilateral middle occipital gyrus, bilateral anterior cingulate cortex, left parahippocampal gyrus, and left calcarine. The positive contribution of WM‐fALFF was in the bilateral superior corona radiata, bilateral superior longitudinal fasciculus, bilateral posterior limb of internal capsule and external capsule. The negative correlation network of WM‐fALFF was mainly located in the whole corpus callosum, the bilateral anterior limb of internal capsule, bilateral tapetum, left posterior corona radiata, and anterior corona radiata.

Pairwise cross‐domain similarity in the spatial patterns of the components was quantified by Pearson correlation for GMV and WM‐fALFF. As illustrated in Figure 3c, the darker blue denotes a higher correlation. All results of pairwise cross‐domain similarity showed significantly correlated in both modalities. In comparison, GMV exhibits a more consistent pattern across all domains (the minimum r value is .73), while WM‐fALFF has greater divergence (lower r value but still significant) among social‐related subdomains. These patterns might indicate that WM‐fALFF is more sensitive to social domains differences than GMV.

Additionally, correlations were also calculated between subdomain ICs loadings in each modality with the referred SRS subdomain scores. As illustrated in Table 2, ICs loading of subdomains significantly correlated with the social subdomain scores across all modalities (communication: GMV r = .13, p = 4.00 × 10−3, WM‐fALFF r = .16, p = 1.2 × 10−3; mannerisms: GMV r = .11, p = 1.55 × 10−2, WM‐fALFF r = −.12, p = 9.1 × 10−3; motivation: GMV r = .09, p = 5.85 × 10−2, WM‐fALFF r = .11, p = 1.80 × 10−2). Particularly, all subdomains positively correlated with social subdomain scores in both modalities, excepting for a negative correlation between mannerisms scores and the loadings of the WM‐fALFF component.

TABLE 2.

The correlation and group differences of SRS subdomains.

Subdomains GMV WM‐fALFF
r‐Value p‐Value t‐Value p‐Value r‐Value p‐Value t‐Value p‐Value
Awareness .08 .0989 3.03 .003 −.07 .1162 −2.83 .005
Cognition .06 .1647 2.96 .003 −.07 .1221 −3.01 .003
Communication .13 .0040 3.20 .001 .16 .0012 4.28 <.001
Motivation .09 .0585 3.02 .003 .11 .0180 2.53 .012
Mannerisms .11 .0155 3.27 .001 −.12 .0091 −3.66 <.001
Open in a new tab

Note: t‐Value means two sample t tests on subdomain ICs loadings between ASD patients and HC, r‐value means correlation between loadings of the identified components and score in all modalities.

3.3. Repeatability across sites

Figure 4 showed the repeatability of our findings. The results of GMV and WM‐fALFF showed significant correlation between the validation map and the main discovery map in SRS total, communication, mannerisms and motivation (GMV: r = .81*, .32*, .51*, .44*; WM‐fALFF: r = .50*, .53*, .16*, .26*. * means p < .005). In particular, the r value of the SRS total in GMV reached a maximum of .81. Notably, the higher correlation value, the higher similarity within two maps.

FIGURE 4.

FIGURE 4

Open in a new tab

Similarity patterns of the identified multimodal spatial maps between cohorts. Reference is a subdomain score vector. Column of “discovery” is main result of spatial maps visualized at | z | > 1.5 thresholds for all modalities. Column of “validation” is validation maps for all modalities.

4. DISCUSSION

In this study, a supervised multimodal MRI fusion analysis guided by social responsiveness scores was performed to explore the neural patterns associated with social impairment in ASD patients on gray matter structure and WM function. Our results uncovered the multimodal spatial patterns related to the SRS multiple scores (total, communication, mannerisms, and motivation). The pairwise cross‐domain correlations suggested GMV exhibited a consistent brain pattern in all patterns associated with social impairments, with SAN and limbic system being commonly identified. The differentiated WM‐fALFF results explored more divergent brain patterns relating to different social deficits, which revealed that abnormal WM functional activity is more sensitive to ASD's complex social impairment. Moreover, brain regions related to social disorders may potentially interconnect across modalities in GMV and WM‐fALFF, which means that multimodal fusion can prove closely association with the brain information of one modality and the information of remote but connected brain regions in another modality. The results from validation experiments showed significant correlation across all modalities and subdomains indicating the robustness and generalizability of our findings.

4.1. The importance of SAN and limbic system in the social impairment of ASD

It should be noted that SAN and limbic system, as the vulnerable regions of our GMV results, were important to uncover multiple social impairments in ASD. Further, this relatively consistent brain patterns across social domains provide us with evidence to speculate that SAN and limbic system in GM may play a crucial role in the structural basis of social responsiveness. According to previous research findings, the SAN modulates the activities between default mode network and central executive network (Goulden et al., 2014; Sridharan et al., 2008). Abnormal SAN can damage brain activity and the behavior caused by it may become the characterization of neurological diseases. Studies have revealed that abnormal functional activity of SAN is significantly associated with sensory and responsiveness deficits in ASD (Di Martino et al., 2009; Green et al., 2015; Uddin et al., 2013), which may help understand the underlying causes of ASD social disorder (Green et al., 2016). As the key hub of SAN, the insula has been proven to be a major hub of dysfunctional social cognition in ASD (Di Martino et al., 2009). And lesion studies provide the bulk of information that hints at the role of anterior cingulate cortex in social interactions (Larson, 1962). Animal experiments also have shown that anterior cingulate cortex plays a role in regulating social behavior in ASD mice, suggesting that anterior cingulate cortex dysfunction may be related to social disorders in ASD (Guo et al., 2019). Furthermore, postmortem reports have proved that ASD patients have cellular abnormalities in the limbic system, which is bound up with the core social and communication impairment in ASD (Buitelaar & Willemsen‐Swinkels, 2000). MRI studies have shown that as an important component of the GM limbic system, abnormalities in the hippocampus and parahippocampal gyrus often co‐occur with social deficits and social withdrawal in ASD (Cooper et al., 2017; Endo et al., 2007; Mouga et al., 2022). Together with previous consistent findings, it is highlighted that SAN and limbic system in GM have crucial responsibility for collaborative processing of social responsiveness and cognition, whose abnormalities would affect the social behavior in ASD.

In addition, we found some unique brain regions related to specific social symptoms. The superior and middle frontal gyrus were identified in communication and motivation, and the median cingulate and paracingulate gyri were uniquely involved in mannerisms. Abnormal structural connections in the frontal cortex (Catani et al., 2016) and median cingulate gyrus (Chien et al., 2021) have been reported to be associated with social disorders in ASD. The middle frontal gyrus was identified in the reward‐related motivation network base on the hypothesis that motivation network abnormality would be specific to these relatively high‐level social‐cognitive processes (Assaf et al., 2013). Moreover, several research groups have begun to suggest that autism may be characterized by a fundamental disturbance in the motivational and mannerism process (Dawson et al., 1998; Mundy, 1995). Therefore, the unique brain regions related to specific clinical social impairments in our study may contribute to revealing the pathological mechanisms of ASD from social perspective.

4.2. Diversified WM functional activity abnormalities are more sensitive to complex ASD social disorders

The pairwise cross‐domain correlations in our study suggested WM‐fALFF exhibited diversified brain patterns relating to different social deficits, indicating a stronger sensitivity of WM‐fALFF to differences in social impairments in ASD. Projection WM tracts (including anterior corona radiata, internal and external capsules) and commissural fibers such as corpus callosum were commonly detected in our multiple social domains results. The corpus callosum as a main commissural channel connects left and right cerebral hemispheres and plays an important role in integrating information. There is growing literature documenting social and linguistic impairments exist in the agenesis of corpus callosum cohort (Paul et al., 2014; Turk et al., 2010) and overlap with the diagnostic criteria for autism (Paul et al., 2007). The similarities in the clinical features of agenesis of corpus callosum and ASD suggest that corpus callosum may be one important WM tract related to social disorders in ASD (Valenti et al., 2020). Additionally, as an important WM projection tract, anterior corona radiata also engages in social regulation and understanding social nuance (Burke et al., 2023). Reduced microstructural organization of anterior corona radiata has been shown to associate with symptom severity in some neurologic diseases, such as schizophrenia (Holleran et al., 2020). Similarly, WM internal and external capsule abnormalities in ASD have been reported, which are related to social interaction (Ameis & Catani, 2015).

Additionally, we also found that there are widely diverse WM fibers in different brain patterns related to social subdomain. Most of these unique tracts could be regarded as the long association fibers, such as cingulum, superior longitudinal fasciculus. Among them, superior longitudinal fasciculus has a broad neuroanatomic range, serving an important role in linking all the components of brain structures involved in social responsiveness. Some research has suggested that widespread abnormalities in association fibers may contribute to social symptoms of ASD (Im et al., 2018). Particularly, almost all WM fibers we found are long‐distance tracts, which have been demonstrated to regulate integrated activity across an extended social processing network (Ameis & Catani, 2015). Therefore, the abnormal patterns of WM functional activity, especially the long‐distance WM tracts in our results, may be also an important pathogenic factor leading to core social disorders in ASD.

4.3. Brain regions interconnect across modalities in GMV and WM‐fALFF

The common brain tracts relating to impaired social interaction from WM‐fALFF can potentially connect to the regions identified from GMV, and vice versa. For instance, the anterior corona radiata is a part of the limbic system circuitry and includes projection tracts from the internal capsule to the cortex that have been associated with impaired top‐down social responsiveness. Similarly, the identified anterior cingulate cortex in GM also forms a large region around corpus callosum that is termed the anterior executive region associated with affective behaviors and response selection. These results clearly indicate multimodal fusion and mutually connected regions proves to be a powerful tool for revealing covariance information and associations between modalities. We observe a certain degree of interconnection between the identified brain regions and WM function activity across the modalities. The anterior corona radiata and corpus callosum we identified in WM‐fALFF are part of the limbic‐thalamo‐cortical circuitry (Catani et al., 2002), which was detected in our GMV results. Abnormal microstructure of WM in limbic system is related to the core symptoms of ASD, which is closely related to genetic (Fu et al., 2022). As hubs of motor and sensory conduction fibers, the internal capsule and external capsule in WM‐fALFF link to caudate nucleus of GMV. These GWV‐fALFF interconnection, based on the multimodal joint analysis, show that the information between different modalities is not in isolation. As a powerful tool, multimodal fusion can mine the complementary information between modalities, which also highlights the potential ability to explore the pathogenesis of ASD (Sui et al., 2018).

Our work still has some limitations. First, ABIDE database provides a variety of social symptom scales, including ADOS, known as the gold standard in the ASD field (Huerta & Lord, 2012). In order to explore the intergroup differences in ASD social impairment, SRS scores were chosen as reference in our study, as ADOS only provides social score for ASD. In future work, we plan to use multiple social scale scores such as ADIR‐S as reference to increase the reliability of the results and further explore the mechanisms of social disorders. Second, additional multimodal data can be added to this model, especially involving WM and functional connectivity (i.e., dynamic functional network connectivity) to fully utilize the advantages of multimodal fusion. Thirdly, considering the significant gender disparity in ASD prevalence, females have been overlooked in this research. As a result, the findings have limited generalizability. The brain mechanisms underlying the social deficits in females with ASD should be explored in future studies. Due to the significant impact of gender on the brain patterns of ASD patients (Chaddad et al., 2017; Walsh et al., 2021), it is meaningful to extend these results to female participants in future work.

5. CONCLUSION

In summary, the present study revealed the multimodal brain patterns that can be used to explain the mechanism of ASD social impairment in gray matter structure and WM function information by using supervised two‐way multimodal fusion analysis. More consistent patterns were found in GWV, in which the SAN and limbic system are associated with multiple social disorders. WM‐fALFF is more sensitive to social impairments due to the divergent patterns. This work is helpful in understanding the neural mechanism of social impairment in ASD and confirming the potential of finding multimodal biomarkers from the multimodal imaging.

AUTHOR CONTRIBUTIONS

Long Wei performed the data analyses, wrote and reviewed the manuscript. Xin Xu performed the data analyses, result analyses and wrote the manuscript draft. Yuwei Su and Min Lan performed the data analyses. Sifeng Wang performed the result analysis. Suyu Zhong designed the study and reviewed the manuscript. All authors contributed to and have approved the final manuscript.

FUNDING INFORMATION

This work is supported in part by the National Science Foundation of China (81701783 to S.Z.), STI 2030—Major Projects (2021ZD0200500), Shandong Provincial Natural Science Foundation (ZR2021MH160) and New Talent Project of Beijing University of Posts and Telecommunications (2021RC40 and 2023RC59).

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to declare.

Supporting information

DATA S1: Supporting Information.

HBM-45-e70017-s001.docx (797.7KB, docx)

ACKNOWLEDGMENTS

The authors would like to thank the centers involving in our study of Autism Brain Imaging Data Exchange (ABIDE) I&II project.

Wei, L. , Xu, X. , Su, Y. , Lan, M. , Wang, S. , & Zhong, S. (2024). Abnormal multimodal neuroimaging patterns associated with social deficits in male autism spectrum disorder. Human Brain Mapping, 45(13), e70017. 10.1002/hbm.70017

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in Autism Brain Imaging Data Exchange at https://fcon_1000.projects.nitrc.org/indi/abide/. These data were derived from the following resources available in the public domain: —ABIDE, https://fcon_1000.projects.nitrc.org/indi/abide/abide_I.html —ABIDE, https://fcon_1000.projects.nitrc.org/indi/abide/abide_II.html.

REFERENCES

  1. Ameis, S. H. , & Catani, M. (2015). Altered white matter connectivity as a neural substrate for social impairment in autism spectrum disorder. Cortex, 62, 158–181. [DOI] [PubMed] [Google Scholar]
  2. Assaf, M. , Hyatt, C. J. , Wong, C. G. , Johnson, M. R. , Schultz, R. T. , Hendler, T. , & Pearlson, G. D. (2013). Mentalizing and motivation neural function during social interactions in autism spectrum disorders. NeuroImage: Clinical, 3, 321–331. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baio, J. , Wiggins, L. , Christensen, D. L. , Maenner, M. J. , Daniels, J. , Warren, Z. , Kurzius‐Spencer, M. , Zahorodny, W. , Robinson, C. , Rosenberg, W. T. , Durkin, M. S. , Imm, P. , Nikolaou, L. , Yeargin‐Allsopp, M. , Lee, L.‐C. , Harrington, R. , Lopez, M. , Fitzgerald, R. T. , Hewitt, A. , … Dowling, N. F. (2018). Prevalence of autism spectrum disorder among children aged 8 years—Autism and developmental disabilities monitoring network, 11 sites, United States, 2014. MMWR Surveillance Summaries, 67, 1–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Buitelaar, J. K. , & Willemsen‐Swinkels, S. H. N. (2000). Autism: Current theories regarding its pathogenesis and implications for rational pharmacotherapy. Pediatric Drugs, 2, 67–81. [DOI] [PubMed] [Google Scholar]
  5. Burke, T. , Holleran, L. , Mothersill, D. , Lyons, J. , O'Rourke, N. , Gleeson, C. , Cannon, D. M. , McKernan, D. P. , Morris, D. W. , Kelly, J. P. , Hallahan, B. , McDonald, C. , & Donohoe, G. (2023). Bilateral anterior corona radiata microstructure organisation relates to impaired social cognition in schizophrenia. Schizophrenia Research, 262, 87–94. [DOI] [PubMed] [Google Scholar]
  6. Calhoun, V. D. , & Sui, J. (2016). Multimodal fusion of brain imaging data: A key to finding the missing link(s) in complex mental illness. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, 1, 230–244. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Catani, M. , Dell'Acqua, F. , Budisavljevic, S. , Howells, H. , Thiebaut De Schotten, M. , Froudist‐Walsh, S. , D'Anna, L. , Thompson, A. , Sandrone, S. , Bullmore, E. T. , Suckling, J. , Baron‐Cohen, S. , Lombardo, M. V. , Wheelwright, S. J. , Chakrabarti, B. , Lai, M.‐C. , Ruigrok, A. N. V. , Leemans, A. , Ecker, C. , … Murphy, D. G. M. (2016). Frontal networks in adults with autism spectrum disorder. Brain, 139, 616–630. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Catani, M. , Howard, R. J. , Pajevic, S. , & Jones, D. K. (2002). Virtual in vivo interactive dissection of White matter fasciculi in the human brain. NeuroImage, 17, 77–94. [DOI] [PubMed] [Google Scholar]
  9. Chaddad, A. , Desrosiers, C. , & Toews, M. (2017). Multi‐scale radiomic analysis of sub‐cortical regions in MRI related to autism, gender and age. Scientific Reports, 7, 45639. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Chen, T. , Hong, R. , Guo, Y. , Hao, S. , & Hu, B. (2023). MS2‐GNN: Exploring GNN‐based multimodal fusion network for depression detection. IEEE Transactions on Cybernetics, 53, 7749–7759. [DOI] [PubMed] [Google Scholar]
  11. Chien, Y.‐L. , Chen, Y.‐C. , & Gau, S. S.‐F. (2021). Altered cingulate structures and the associations with social awareness deficits and CNTNAP2 gene in autism spectrum disorder. NeuroImage: Clinical, 31, 102729. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Constantino, J. N. , Davis, S. A. , Todd, R. D. , Schindler, M. K. , Gross, M. M. , Brophy, S. L. , Metzger, L. M. , Shoushtari, C. S. , Splinter, R. , & Reich, W. (2003). Validation of a brief quantitative measure of autistic traits: Comparison of the social responsiveness scale with the autism diagnostic interview‐revised. Journal of Autism and Developmental Disorders, 33, 427–433. [DOI] [PubMed] [Google Scholar]
  13. Cooper, R. A. , Richter, F. R. , Bays, P. M. , Plaisted‐Grant, K. C. , Baron‐Cohen, S. , & Simons, J. S. (2017). Reduced hippocampal functional connectivity during episodic memory retrieval in autism. Cerebral Cortex, 27, 888–902. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Dawson, G. , Meltzoff, A. N. , Osterling, J. , Rinaldi, J. , & Brown, E. (1998). Children with autism fail to orient to naturally occurring social stimuli. Journal of Autism and Developmental Disorders, 28, 479–485. [DOI] [PubMed] [Google Scholar]
  15. Di Martino, A. , Ross, K. , Uddin, L. Q. , Sklar, A. B. , Castellanos, F. X. , & Milham, M. P. (2009). Functional brain correlates of social and nonsocial processes in autism spectrum disorders: An activation likelihood estimation meta‐analysis. Biological Psychiatry, 65, 63–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Di Martino, A. , Yan, C.‐G. , Li, Q. , Denio, E. , Castellanos, F. X. , Alaerts, K. , Anderson, J. S. , Assaf, M. , Bookheimer, S. Y. , Dapretto, M. , Deen, B. , Delmonte, S. , Dinstein, I. , Ertl‐Wagner, B. , Fair, D. A. , Gallagher, L. , Kennedy, D. P. , Keown, C. L. , Keysers, C. , … Milham, M. P. (2014). The autism brain imaging data exchange: Towards a large‐scale evaluation of the intrinsic brain architecture in autism. Molecular Psychiatry, 19, 659–667. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Ding, Z. , Huang, Y. , Bailey, S. K. , Gao, Y. , Cutting, L. E. , Rogers, B. P. , Newton, A. T. , & Gore, J. C. (2018). Detection of synchronous brain activity in white matter tracts at rest and under functional loading. Proceedings of the National Academy of Sciences, 115, 595–600. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Ding, Z. , Xu, R. , Bailey, S. K. , Wu, T.‐L. , Morgan, V. L. , Cutting, L. E. , Anderson, A. W. , & Gore, J. C. (2016). Visualizing functional pathways in the human brain using correlation tensors and magnetic resonance imaging. Magnetic Resonance Imaging, 34, 8–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Du, Y. , Li, B. , Hou, Y. , & Calhoun, V. D. (2020). A deep learning fusion model for brain disorder classification: Application to distinguishing schizophrenia and autism spectrum disorder. In: Proceedings of the 11th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics. Virtual Event USA: ACM. pp 1–7. 10.1145/3388440.3412478 [DOI] [PMC free article] [PubMed]
  20. Endo, T. , Shioiri, T. , Kitamura, H. , Kimura, T. , Endo, S. , Masuzawa, N. , & Someya, T. (2007). Altered chemical metabolites in the amygdala‐hippocampus region contribute to autistic symptoms of autism spectrum disorders. Biological Psychiatry, 62, 1030–1037. [DOI] [PubMed] [Google Scholar]
  21. Fu, L. , Li, C. , Li, Y. , Cheng, X. , Cui, X. , Jiang, J. , Ding, N. , Fang, H. , Tang, T. , & Ke, X. (2022). Heritability of abnormalities in limbic networks of autism spectrum disorder children: Evidence from an autism spectrum disorder twin study. Autism Research, 15, 628–640. [DOI] [PubMed] [Google Scholar]
  22. Gaser, C. , Dahnke, R. , Thompson, P. M. , Kurth, F. , Luders, E. , & Alzheimer's Disease Neuroimaging Initiative . (2024). CAT: a computational anatomy toolbox for the analysis of structural MRI data. GigaScience, 13, 1–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Geenjaar, E. P. T. , Lewis, N. L. , Fedorov, A. , Wu, L. , Ford, J. M. , Preda, A. , Plis, S. M. , & Calhoun, V. D. (2023). Chromatic fusion: Generative multimodal neuroimaging data fusion provides multi‐informed insights into schizophrenia. Human Brain Mapping, 44, 5828–5845. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Goulden, N. , Khusnulina, A. , Davis, N. J. , Bracewell, R. M. , Bokde, A. L. , McNulty, J. P. , & Mullins, P. G. (2014). The salience network is responsible for switching between the default mode network and the central executive network: Replication from DCM. NeuroImage, 99, 180–190. [DOI] [PubMed] [Google Scholar]
  25. Green, S. A. , Hernandez, L. , Bookheimer, S. Y. , & Dapretto, M. (2016). Salience network connectivity in autism is related to brain and behavioral markers of sensory overresponsivity. Journal of the American Academy of Child and Adolescent Psychiatry, 55, 618–626.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Green, S. A. , Hernandez, L. , Tottenham, N. , Krasileva, K. , Bookheimer, S. Y. , & Dapretto, M. (2015). Neurobiology of sensory overresponsivity in youth with autism spectrum disorders. JAMA Psychiatry, 72, 778–786. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Guo, B. , Chen, J. , Chen, Q. , Ren, K. , Feng, D. , Mao, H. , Yao, H. , Yang, J. , Liu, H. , Liu, Y. , Jia, F. , Qi, C. , Lynn‐Jones, T. , Hu, H. , Fu, Z. , Feng, G. , Wang, W. , & Wu, S. (2019). Anterior cingulate cortex dysfunction underlies social deficits in Shank3 mutant mice. Nature Neuroscience, 22, 1223–1234. [DOI] [PubMed] [Google Scholar]
  28. Guo, X. , Liu, L. , Li, T. , Zhao, Q. , Li, H. , Huang, F. , Wang, Y. , Qian, Q. , Cao, Q. , Wang, Y. , Calhoun, V. D. , Sui, J. , & Sun, L. (2021). Inhibition‐directed multimodal imaging fusion patterns in adults with ADHD and its potential underlying “gene‐brain‐cognition” relationship. CNS Neuroscience & Therapeutics, 27, 664–673. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Hirota, T. , & King, B. H. (2023). Autism spectrum disorder: A review. JAMA, 329, 157. [DOI] [PubMed] [Google Scholar]
  30. Holleran, L. , Kelly, S. , Alloza, C. , Agartz, I. , Andreassen, O. A. , Arango, C. , Banaj, N. , Calhoun, V. , Cannon, D. , Carr, V. , Corvin, A. , Glahn, D. C. , Gur, R. , Hong, E. , Hoschl, C. , Howells, F. M. , James, A. , Janssen, J. , Kochunov, P. , … Donohoe, G. (2020). The relationship between white matter microstructure and general cognitive ability in patients with schizophrenia and healthy participants in the ENIGMA consortium. The American Journal of Psychiatry, 177, 537–547. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Huerta, M. , & Lord, C. (2012). Diagnostic evaluation of autism spectrum disorders. Pediatric Clinics of North America, 59, 103–111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Im, W. Y. , Ha, J. H. , Kim, E. J. , Cheon, K.‐A. , Cho, J. , & Song, D.‐H. (2018). Impaired White matter integrity and social cognition in high‐function autism: Diffusion tensor imaging study. Psychiatry Investigation, 15, 292–299. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Ingalhalikar, M. , Parker, W. A. , Bloy, L. , Roberts, T. P. L. , & Verma, R. (2014). Creating multimodal predictors using missing data: Classifying and subtyping autism spectrum disorder. Journal of Neuroscience Methods, 235, 1–9. [DOI] [PubMed] [Google Scholar]
  34. Koyama, T. , Tachimori, H. , Osada, H. , Takeda, T. , & Kurita, H. (2007). Cognitive and symptom profiles in Asperger's syndrome and high‐functioning autism. Psychiatry and Clinical Neurosciences, 61, 99–104. [DOI] [PubMed] [Google Scholar]
  35. Larson, S. J. (1962). The connections of the cingulate gyrus and insula in the monkey. Northwestern University. [Google Scholar]
  36. Levy, S. E. , Mandell, D. S. , & Schultz, R. T. (2009). Autism. The Lancet, 374, 1627–1638. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Li, T. , Fu, Z. , Liu, X. , Qi, S. , Calhoun, V. D. , & Sui, J. (2019). Multimodal neuroimaging patterns associated with social responsiveness impairment in autism: A replication study. In: 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019). Venice, Italy: IEEE. pp. 409–413. Retrieved from https://ieeexplore.ieee.org/document/8759460/
  38. Li, Y. , Adalı, T. , & Calhoun, V. D. (2007). Estimating the number of independent components for functional magnetic resonance imaging data. Human Brain Mapping, 28, 1251–1266. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Lord, C. , Risi, S. , Lambrecht, L. , Cook, E. H., Jr. , Leventhal, B. L. , DiLavore, P. C. , Pickles, A. , & Rutter, M. (2000). The autism diagnostic observation schedule—Generic: A standard measure of social and communication deficits associated with the spectrum of autism. Journal of Autism and Developmental Disorders, 30, 205–223. [PubMed] [Google Scholar]
  40. Maglanoc, L. A. , Kaufmann, T. , Jonassen, R. , Hilland, E. , Beck, D. , Landrø, N. I. , & Westlye, L. T. (2020). Multimodal fusion of structural and functional brain imaging in depression using linked independent component analysis. Human Brain Mapping, 41, 241–255. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Mouga, S. , Duarte, I. C. , Café, C. , Sousa, D. , Duque, F. , Oliveira, G. , & Castelo‐Branco, M. (2022). Parahippocampal deactivation and hyperactivation of central executive, saliency and social cognition networks in autism spectrum disorder. Journal of Neurodevelopmental Disorders, 14, 9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Mundy, P. (1995). Joint attention and social‐emotional approach behavior in children with autism. Development and Psychopathology, 7, 63–82. [Google Scholar]
  43. Muschelli, J. , Nebel, M. B. , Caffo, B. S. , Barber, A. D. , Pekar, J. J. , & Mostofsky, S. H. (2014). Reduction of motion‐related artifacts in resting state fMRI using aCompCor. NeuroImage, 96, 22–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Oblong, L. M. , Llera, A. , Mei, T. , Haak, K. , Isakoglou, C. , Floris, D. L. , Durston, S. , Moessnang, C. , Banaschewski, T. , Baron‐Cohen, S. , Loth, E. , Dell'Acqua, F. , Charman, T. , Murphy, D. G. M. , Ecker, C. , Buitelaar, J. K. , Beckmann, C. F. , The EU‐AIMS LEAP Group , Ahmad, J. , … Forde, N. J. (2023). Linking functional and structural brain organisation with behaviour in autism: A multimodal EU‐AIMS Longitudinal European Autism Project (LEAP) study. Molecular Autism, 14, 32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Okamoto, N. , & Akama, H. (2021). Extended invariant information clustering is effective for leave‐one‐site‐out cross‐validation in resting state functional connectivity modeling. Frontiers in Neuroinformatics, 15, 709179. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Pan, J. , Lin, H. , Dong, Y. , Wang, Y. , & Ji, Y. (2022). MAMF‐GCN: Multi‐scale adaptive multi‐channel fusion deep graph convolutional network for predicting mental disorder. Computers in Biology and Medicine, 148, 105823. [DOI] [PubMed] [Google Scholar]
  47. Patti, M. A. , Newschaffer, C. , Eliot, M. , Hamra, G. B. , Chen, A. , Croen, L. A. , Fallin, M. D. , Hertz‐Picciotto, I. , Kalloo, G. , Khoury, J. C. , Lanphear, B. P. , Lyall, K. , Yolton, K. , & Braun, J. M. (2021). Gestational exposure to phthalates and social responsiveness scores in children using quantile regression: The EARLI and HOME studies. International Journal of Environmental Research and Public Health, 18, 1254. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Paul, L. K. , Brown, W. S. , Adolphs, R. , Tyszka, J. M. , Richards, L. J. , Mukherjee, P. , & Sherr, E. H. (2007). Agenesis of the corpus callosum: Genetic, developmental and functional aspects of connectivity. Nature Reviews. Neuroscience, 8, 287–299. [DOI] [PubMed] [Google Scholar]
  49. Paul, L. K. , Corsello, C. , Kennedy, D. P. , & Adolphs, R. (2014). Agenesis of the corpus callosum and autism: A comprehensive comparison. Brain, 137, 1813–1829. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Peterson, B. S. , Liu, J. , Dantec, L. , Newman, C. , Sawardekar, S. , Goh, S. , & Bansal, R. (2022). Using tissue microstructure and multimodal MRI to parse the phenotypic heterogeneity and cellular basis of autism spectrum disorder. Journal of Child Psychology and Psychiatry, 63, 855–870. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Qi, S. , Calhoun, V. D. , Van Erp, T. G. M. , Bustillo, J. , Damaraju, E. , Turner, J. A. , Du, Y. , Yang, J. , Chen, J. , Yu, Q. , Mathalon, D. H. , Ford, J. M. , Voyvodic, J. , Mueller, B. A. , Belger, A. , McEwen, S. , Potkin, S. G. , Preda, A. , Jiang, T. , & Sui, J. (2018). Multimodal fusion with reference: Searching for joint Neuromarkers of working memory deficits in schizophrenia. IEEE Transactions on Medical Imaging, 37, 93–105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Qi, S. , Morris, R. , Turner, J. A. , Fu, Z. , Jiang, R. , Deramus, T. P. , Zhi, D. , Calhoun, V. D. , & Sui, J. (2020). Common and unique multimodal covarying patterns in autism spectrum disorder subtypes. Molecular Autism, 11, 90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Rafiee, F. , Rezvani Habibabadi, R. , Motaghi, M. , Yousem, D. M. , & Yousem, I. J. (2022). Brain mri in autism spectrum disorder: Narrative review and recent advances. Journal of Magnetic Resonance Imaging, 55, 1613–1624. [DOI] [PubMed] [Google Scholar]
  54. Redcay, E. , & Courchesne, E. (2008). Deviant functional magnetic resonance imaging patterns of brain activity to speech in 2–3‐year‐old children with autism spectrum disorder. Biological Psychiatry, 64, 589–598. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Rojas, D. C. , Peterson, E. , Winterrowd, E. , Reite, M. L. , Rogers, S. J. , & Tregellas, J. R. (2006). Regional gray matter volumetric changes in autism associated with social and repetitive behavior symptoms. BMC Psychiatry, 6, 56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Saponaro, S. , Lizzi, F. , Serra, G. , Mainas, F. , Oliva, P. , Giuliano, A. , Calderoni, S. , & Retico, A. (2024). Deep learning based joint fusion approach to exploit anatomical and functional brain information in autism spectrum disorders. Brain Informatics, 11, 2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Sridharan, D. , Levitin, D. J. , & Menon, V. (2008). A critical role for the right fronto‐insular cortex in switching between central‐executive and default‐mode networks. Proceedings of the National Academy of Sciences, 105, 12569–12574. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Sui, J. , He, H. , Pearlson, G. D. , Adali, T. , Kiehl, K. A. , Yu, Q. , Clark, V. P. , Castro, E. , White, T. , Mueller, B. A. , Ho, B. C. , Andreasen, N. C. , & Calhoun, V. D. (2013). Three‐way (N‐way) fusion of brain imaging data based on mCCA+jICA and its application to discriminating schizophrenia. NeuroImage, 66, 119–132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. 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, 3028. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Turk, A. A. , Brown, W. S. , Symington, M. , & Paul, L. K. (2010). Social narratives in agenesis of the corpus callosum: Linguistic analysis of the Thematic Apperception Test. Neuropsychologia, 48, 43–50. [DOI] [PubMed] [Google Scholar]
  61. Uddin, L. Q. , Supekar, K. , Lynch, C. J. , Khouzam, A. , Phillips, J. , Feinstein, C. , Ryali, S. , & Menon, V. (2013). Salience network‐based classification and prediction of symptom severity in children with autism. JAMA Psychiatry, 70, 869–879. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Valenti, M. , Pino, M. C. , Mazza, M. , Panzarino, G. , Di Paolantonio, C. , & Verrotti, A. (2020). Abnormal structural and functional connectivity of the corpus callosum in autism spectrum disorders: A review. Review Journal of Autism and Developmental Disorders, 7, 46–62. [Google Scholar]
  63. Vinçon‐Leite, A. , Saitovitch, A. , Lemaître, H. , Rechtman, E. , Boisgontier, J. , Fillon, L. , Philippe, A. , Rio, M. , Desguerre, I. , Fabre, A. , Aljabali, K. , Boddaert, N. , & Zilbovicius, M. (2024). Identifying interindividual variability of social perception and associated brain anatomical correlations in children with autism spectrum disorder using eye‐tracking and diffusion tensor imaging MRI (DTI‐MRI). Cerebral Cortex, 34, bhad434. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Walsh, M. J. M. , Wallace, G. L. , Gallegos, S. M. , & Braden, B. B. (2021). Brain‐based sex differences in autism spectrum disorder across the lifespan: A systematic review of structural MRI, fMRI, and DTI findings. NeuroImage: Clinical, 31, 102719. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Wang, P. , Wang, J. , Michael, A. , Wang, Z. , Klugah‐Brown, B. , Meng, C. , & Biswal, B. B. (2022). White matter functional connectivity in resting‐state fMRI: Robustness, reliability, and relationships to gray matter. Cerebral Cortex, 32, 1547–1559. [DOI] [PubMed] [Google Scholar]
  66. Wang, S. , Zheng, K. , Kong, W. , Huang, R. , Liu, L. , Wen, G. , & Yu, Y. (2023). Multimodal data fusion based on IGERNNC algorithm for detecting pathogenic brain regions and genes in Alzheimer's disease. Briefings in Bioinformatics, 24, bbac515. [DOI] [PubMed] [Google Scholar]
  67. Yan, C.‐G. , Wang, X.‐D. , Zuo, X.‐N. , & Zang, Y.‐F. (2016). DPABI: Data processing & analysis for (resting‐state) brain imaging. Neuroinformatics, 14, 339–351. [DOI] [PubMed] [Google Scholar]
  68. Yao, C. , Hu, N. , Cao, H. , Tang, B. , Zhang, W. , Xiao, Y. , Zhao, Y. , Gong, Q. , & Lui, S. (2021). A multimodal fusion analysis of pretreatment anatomical and functional cortical abnormalities in responsive and non‐responsive schizophrenia. Frontiers in Psychiatry, 12, 737179. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Zou, Q.‐H. , Zhu, C.‐Z. , Yang, Y. , Zuo, X.‐N. , Long, X.‐Y. , Cao, Q.‐J. , Wang, Y.‐F. , & Zang, Y.‐F. (2008). An improved approach to detection of amplitude of low‐frequency fluctuation (ALFF) for resting‐state fMRI: Fractional ALFF. Journal of Neuroscience Methods, 172, 137–141. [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.

Supplementary Materials

DATA S1: Supporting Information.

HBM-45-e70017-s001.docx (797.7KB, docx)

Data Availability Statement

The data that support the findings of this study are available in Autism Brain Imaging Data Exchange at https://fcon_1000.projects.nitrc.org/indi/abide/. These data were derived from the following resources available in the public domain: —ABIDE, https://fcon_1000.projects.nitrc.org/indi/abide/abide_I.html —ABIDE, https://fcon_1000.projects.nitrc.org/indi/abide/abide_II.html.

Articles from Human Brain Mapping are provided here courtesy of Wiley

RESOURCES