Participants
In this study, we recruited a total of 107 participants (69 ASD/38 TD) aged 1.5 to 6.5 years from the Foshan Fosun Chancheng Hospital, Foshan, China between November 2021 and May 2023. All participants completed the Gesell Development Schedule (GDS), which measures various developmental domains including fine motor, gross motor, personal-social, language, and adaptive behavior [26]. The parents or guardians of all participants completed the Autism Behavior Checklist (ABC), a widely used questionnaire for assessing autistic behaviors and symptom severity in individuals with ASD [27]. Although the ABC was not utilized for diagnosing ASD, it is notable that all TD children had an ABC score < 44, indicating the absence of ASD. Conversely, children with ASD had ABC scores ≥ 53. All participants diagnosed with ASD met the DSM-V criteria for ASD through clinical interview and underwent either the Autism Diagnostic Observation Schedule (ADOS Module 1 or 2) or the Childhood Autism Rating Scale (CARS), administered by the same clinician. TD participants had a total score > 85 on the GDS, indicating normal development. All children were native Mandarin or Cantonese speakers with normal hearing and no family history of mental or psychiatric disorders. This study was approved by the Foshan Fosun Chancheng Hospital. Informed consent was obtained from parents or guardians of all participants.
Two participants were excluded from the analysis due to poor MRI data quality (1 ASD/1 TD). Given the diverse range of language abilities observed within the remaining ASD sample, we stratified the ASD participants into two subgroups based on the median of their language scores obtained from the GDS. Specifically, ASD participants were classified as ASD with moderate language deficits (ASDmoderate, n = 34) if they had a language score ≥ 44.4, while those with a language score < 44.4 were categorized as ASD with severe language deficits (ASDsevere, n = 34). The detailed demographic, clinical, and behavioral information of TD group and two ASD subgroups is summarized in Table 1.
Table 1. Demographic details and clinical and behavioral testing scores.
|
TD
(n = 37)
|
ASDmoderate
(n = 34)
|
ASDsevere
(n = 34)
|
|
Mean±SD
|
Range
|
Mean±SD
|
Range
|
Mean±SD
|
Range
|
Age (years)
|
3.39±1.49
|
1.5-5.87
|
3.31±1.33
|
1.52-6.53
|
3.039±0.98
|
2-5.57
|
Gender (M/F)
|
32/5
|
32/2
|
25/9
|
Gesell subscale scores*
|
Gross motor
|
97.42±6.1
|
81.1-107
|
77.76±8.7
|
47.8-91
|
71.79±8.34
|
52-87.5
|
Fine motor
|
96.73±5.34
|
86-108
|
73.31±9.39
|
54.1-90.4
|
64.25±10.21
|
45.9-84.1
|
Personal-social
|
93.76±4.52
|
80-103
|
60.46±7.4
|
31-70.2
|
48.16±7.78
|
30.2-64.2
|
Language
|
91.47±5.55
|
75.7-101
|
55.07±7.76
|
44.6-70.7
|
37.11±5.91
|
23.6-44.2
|
Adaptive behavior
|
93.79±5.39
|
80-107
|
66.18±8.76
|
40.9-82
|
55.07±13.33
|
31.6-93.7
|
Total
|
94.34±4.22
|
85.1-100.8
|
66.72±5.61
|
49.1-79.4
|
55.25±7.11
|
39.3-68.74
|
ABC
|
25.3±8.56
|
8-43
|
66.65±14.26
|
53-107
|
74.97±19.81
|
54-130
|
ADOS scores#
|
ADOS SA
|
|
|
12.41±3.6
|
7-20
|
13.04±3.26
|
8-22
|
ADOS RRB
|
|
|
1.24±0.91
|
0-4
|
1.52±1.05
|
0-4
|
ADOS Total
|
|
|
13.66±4.08
|
8-23
|
14.56±3.91
|
9-24
|
CARS Total##
|
14.33±1.15
|
13-15
|
31.92±2.02
|
30-38
|
35±4.97
|
25-45
|
Abbreviations: ABC, Autism, Behavior Checklist; RRB, Restricted and Repetitive Behavior; ADOS, Autism Diagnostic Observation Schedule; CARS,
# ADOS was administrated in 54 ASD children (ASDmoderate: n = 29; ASDsevere: n = 25).
## CARS was administrated in 8 TD children and 60 ASD children (ASDmoderate: n = 26; ASDsevere: n = 34).
MRI data acquisition
Before MRI scanning, all of the participants were administered 0.5% chloral hydrate 0.5 ml/kg (maximum dose 10 ml) orally to induce and maintain sleep. All participants continued sleeping during scanning. All the structural MRI data were collected on a 3.0T SEMENS Skyra at the Foshan Chancheng Hospital, Foshan, China using a T1-weighted MPRAGE sequence (TE = 2.98 ms, TR = 2300 ms, resolution = 1.0 x 1.0 x 1.0 mm3, space gap=0, slice thickness = 1 mm, flip angle =9°, 144 slices, a total of 5 min 9 s).
Imaging data preprocessing
Prior to preprocessing, MR images were visually inspected and then normalized to standard AC/PC orientations. To extract grey matter maps, MRI data were processed with the Voxel-based morphometry (VBM) pipeline using the Computational Anatomy Toolbox (CAT 12; https://neuro-jena.github.io/cat/) for Statistical Parametric Mapping (SPM12; http://www.fil.ion.ucl.ac.uk), running in Matlab R2020a (MathWorks, Natick, MA, USA). Here, to minimize the potential confounds introduced by the different brain sizes and tissues between young children and adults [28], customized pediatric tissue probability maps and the diffeomorphic anatomical registration through exponentiated lie algebra (DARTEL) templates were created with the Cerebromatic (COM) Toolbox [29, 30] which provides regression parameters modeled with 1914 healthy participants aged 13 months to 75 years. Specifically, the COM toolbox can be used to generate the custom tissue probability maps that matches sample demographics to parameters that influence brain structure using a flexible non-parametric approach: multivariate adaptive regression splines [29]. The custom DARTEL templates can also be created using the COM toolbox, which matches sample demographics to a second set of regression parameters derived from 1919 participants in the same databases [30]. Here, the age and sex of each participant, and the field strength were entered into the COM toolbox to create the custom tissue probability maps and DARTEL templates, separately.
For the VBM analysis, MRI images were segmented into grey matter, white matter, and cerebrospinal fluid (CSF). Following segmentation, the grey matter images were affine registered to pediatric tissue probability maps previously generated, and then they were spatially normalized to a study-specific pediatric template using DARTEL registration. Subsequently, the grey matter images were modulated with Jacobian determinants from the normalization process to preserve regional volumes. Quality control measures were implemented to ensure sample homogeneity, with no outlier images identified. The grey matter images underwent smoothing using an 8 mm full-width at half-maximum (FWHM) kernel. Processed grey matter images had a voxel size of 1.5 mm × 1.5 mm × 1.5 mm. Finally, total grey matter, white matter, CSF, and intracranial volume (TIV) measurements were extracted for each participant.
Statistical analysis
Group differences in demographic and clinical data
Statistical analyses for demographic and clinical data were performed using R software (version 4.1.2). Specifically, differences between the TD group and the two ASD subgroups in demographic information (i.e., age) and behavioral testing (total and domain scores of GDS) were assessed using two-sample t-tests, while differences in gender were assessed using the Chi-square test. Comparisons of clinical (e.g., ADOS, CARS, ABC) scores were conducted only between the two ASD subgroups. Results for age and total and domain scores of GDS were corrected for multiple comparisons using the false discovery rate (FDR) approach.
Comparisons of global volumes between TD and two ASD subgroups
We analyzed the group differences among the TD group and the ASD subgroups in volumes of grey matter, white matter, CSF using one-way analysis of covariance (ANCOVA), while controlling for age, gender, and TIV. Additionally, a group comparison of TIV was conducted while controlling for age and gender. The results were corrected for multiple comparisons using the FDR correction (p < 0.05). For significant group differences, post-hoc t-tests were conducted to further explore the differences between TD, ASDmoderate, and ASDsevere, with corrections for multiple testing using the FDR approach.
Group differences in regional grey matter volume
The whole brain voxel-wise analyses were performed to examine the group differences between TD and two ASD subgroups in grey matter volume using the ‘y_TTest2_Image’ function in DAPBI (a toolbox for Data Processing & Analysis for Brain Imaging; https://rfmri.org/dpabi) [31], with age, gender, and TIV as covariates. We used a binary grey matter mask for the group analysis that was created with all the grey matter images across participants. Results were corrected at the cluster-level using Gaussian random field (GRF) theory, in which the t map was converted to the z map (|Z| ≥ 3.29, voxel-wise p < 0.001, cluster size > 800 voxels, cluster-wise p < 0.05, two-tailed).
Correlations between grey matter volume and language ability
Further, we examined the correlations between whole-brain grey matter volume and language ability in TD and two ASD subgroups separately, using the ‘y_Correlation_Image’ function in DPABI, while controlling for age, gender, and TIV. The resulting correlation maps were converted to the z maps and corrected for multiple comparisons at the cluster-level using the GRF correction (|Z| ≥ 3.29, voxel-wise p < 0.001, cluster size > 650 voxels, cluster-wise p < 0.05, two-tailed). Finally, all the significant clusters were visualized with the BrainNet Viewer (http://www.nitrc.org/ projects/bnv/).