Cerebellar Structural Abnormality in Autism Spectrum Disorder: A Magnetic Resonance Imaging Study

Article information

Psychiatry Investig. 2023;20(4):334-340

Publication date (electronic) : 2023 April 20

doi : https://doi.org/10.30773/pi.2022.0254

Qifang Lu¹, Jin Chen¹, Yanming Wang², Li Huang¹, Zhoufan Jiang², Benedictor Alexander Nguchu², Shishuo Chen², Bensheng Qiu^,2, Xiaoxiao Wang^,2,3

¹School of Clinical Medicine, Anhui Medical College, Anhui, China

²Center for Biomedical Imaging, University of Science and Technology of China, Anhui, China

³Institute of Artificial Intelligence, Hefei Comprehensive National Science Center, Anhui, China

Correspondence: Xiaoxiao Wang, PhD Center for Biomedical Imaging, University of Science and Technology of China, Hefei, Anhui 230026, China Tel: +86 18956068202, E-mail: wang506@ustc.edu.cn

Correspondence: Bensheng Qiu, PhD Center for Biomedical Imaging, University of Science and Technology of China, Hefei, Anhui 230026, China Tel: +86 15056903508, E-mail: bqiu@ustc.edu.cn

Received 2022 September 21; Revised 2022 December 5; Accepted 2023 January 15.

Abstract

INTRODUCTION

Autism spectrum disorder (ASD), is a complex neurodevelopmental disorder characterized by difficulties in social interaction, repetitive, stereotyped behaviors, and a certain degree of cognitive deficits. Recent statistics have shown a growing prevalence of ASD in 0.7% of children aged 6–12 years [1]. The pathogenesis of ASD remains unclear, with some reports proposing a relationship with genetics [2], environment, neurotransmitters, and neural pathways. Due to this, ASD is associated with different levels of severity and performance in cognition, behavior, and language, which makes the diagnosis and treatment of ASD rather difficult.

In recent years, magnetic resonance imaging (MRI) is a good non-invasive imaging method and is widely used in research. Analyzing structural, functional, and metabolic MRI information would help identify objective, biological markers of ASD. And the correlation analyses would provide bases for neuropathology and render therapeutic directions for clinical treatment. Previous studies of ASD have reported alterations in focal brain regions, especially the frontal and temporal lobes, hippocampus, amygdala, and cerebellum, but the results were inconsistent [3,4], which may also be due to the heterogeneity of ASD.

Over the last 30 years, the study of functions of the cerebellum has been challenging. Available literature proposes that higher-order functions such as emotion regulation, attention control, cognition, and memory requires involvement of the cerebellum. Because of its involvement in a wide range of functions, the impaired cerebellum is associated with several psychiatric and neurological disorders [5-8]. Earlier studies have already found structural and functional anomalies in patients with ASD, which involved the cerebellum areas [9-11]. Even though the cerebellar structure is small, it is still rich in neurons, and so it has been shown to have functional topography across its subregions. These subregions have a crucial role to play in higher-order cognitive functions [12]. Even so, ASD studies on cerebellar subregions are few, which limits our understanding of neural bases of ASD.

Therefore, we aimed to explore the morphological changes related to ASD in the cerebellar lobules by including additional cerebellar lobules which were not involved in the previous studies. We surmise that additional lobules would help uncover structural changes associated with ASD, thus providing additional neural mechanisms to support the diagnosis of clinical ASD.

METHODS

Subjects

The experimental data used in this paper were from the Autism Brain Imaging Data Exchange (ABIDE), a large, international, publicly available dataset (http://fcon_1000.projects.nitrc.org/indi/abide/). The subject inclusion criteria were: 1) the amount of data at the subject’s site is greater than 30; 2) Wechsler Abbreviated Scale of Intelligence assessment of intelligence with a full scale intelligence quotient (FIQ) score >75 (if FIQ was missing, the validated Performance IQ and Verbal IQ scores were used to estimate); 3) Subject age should be 6–18 years; and 4) no chronic systemic diseases, nor use of antipsychotic medications.

A total of 172 subjects from three independent sample sets (NYU Langone Medical Center, Stanford University, and University of Leuven) from the ABIDE I dataset were included in the study. A total of 75 individuals had ASD (64 males and 11 females, aged 7–18 years) and 97 individuals were typically development (TD, 72 males and 25 females, aged 6–18 years). To allow a reliable comparison and joint analysis in multi-site studies, we applied a robust data harmonization method, Com-Bat, which adopts statistical or mathematical concepts to reduce unwanted site variability while maintaining the biological content, such as cortical thickness and age information [13]. Table 1 shows the demographic information of the participants, which includes age, gender, FIQ, and the scores of four items in the Autism Diagnostic Interview-Revised (ADI-R). The ADI-R score statistics in Table 1 are for the 54 individuals in the ASD group for whom scores were recorded. Statistical data of age, gender, or IQ did not differ significantly between ASD and TD individuals.

Table 1.

Sample data statistical information

Image processing

Both MRI structural and resting-state functional images were acquired for each subject. Considering the purpose of this study, we selected data with T1-weighted images in three orientations: cross-sectional, coronal, and sagittal, while the quality of the selected images was examined jointly by two imaging physicians with intermediate or higher titles. The highresolution structural images were acquired with a standard fast spoiled gradient-echo T1-weighted sequence (TR: 11.08 ms; TE: 4.3 ms; flip angle: 45°; FOV: 256 mm; 256×256 matrix; 180 slices; 1 mm3 resolution). Data processing was performed using Volbrain, an online MR brain volume measurement system (https://www.volbrain.upv.es/) that uses CERES (CEREbellum Segmentation) technology to segment and estimate structural measurements. CERES is a patch-based multiatlas segmentation tool that automatically segments the cerebellar lobes [14]. Each cerebellar hemisphere was segmented automatically into 12 lobules (I–II, III, IV, V, VI, Crus I, Crus II, VIIB, VIIIA, VIIIB, IX, and X). Measurements of total volume, grey matter volume, cortical thickness, and normalized cortical thickness were recorded. Figure 1 shows the cerebellar segmentation maps in Montreal Neurological Institute space as reported by Volbrain.

Figure 1.

CERES cerebellar marker segmentation map. Resulting images are located in MNI space from CERES volumetry reports. The 12 lobules (I–II, III, IV, V, VI, Crus I, Crus II, VIIB, VIIIA, VIIIB, IX, and X) of the cerebellar hemisphere are labeled on the lobules segmentation maps. CERES, CEREbellum Segmentation; MNI, Montreal Neurological Institute.

RESULTS

Comparison of normalized cortical thickness measurements of the cerebellar lobules in the ASD and TD groups

In ANOVA tests, we found significant differences in normalized cortical thickness between the ASD and TD groups (F=5.436, p=0.021). Of 24 lobules (12 from each cerebellar hemisphere), five lobules (right lobule VI, left lobule VI, right lobule Crus I, left lobule Crus I, and left lobule X) differed significantly between ASD and TD in measurements of normalized cortical thickness as tested by independent sample t-test. Specifically, the normalized cortical thickness appeared to be smaller in the ASD group than the TD group in these lobules, with p-values of 0.006, 0.017, 0.016, 0.008, and 0.01, respectively. While no other cerebellar lobules presented significant differences in normalized cortical thickness. These results were summarized in Table 2 and Figure 2.

Table 2.

Comparison of normalized cortical thickness measurements of the cerebellar lobules in the ASD and TD groups

Figure 2.

Comparative results of normalized cortical thickness of cerebellar lobules in ASD and TD groups. Among the 24 lobules of the cerebellum, the normalized cortical thickness of right lobule VI (p=0.006), left lobule VI (p=0.017), right lobule Crus I (p=0.016), left lobule Crus I (p=0.008), and left lobule X (p=0.01) in ASD were significantly reduced compared with the TD group. Error bars indicate interquartile range. *indicate an exception value. ASD, autism spectrum disorder; TD, typically developing.

Correlation between the normalized cortical thickness of cerebellar lobules and clinical symptoms in autistic patients

The correlation analyses between cerebellar lobule structures and clinical symptoms were performed with data from only 54 out of 75 patients since only these patients had ADI-R scores recorded. In combination with the results of Section 3.1, Pearson correlation analysis was used to correlate the normalized cortical thickness of the right lobule VI, left lobule VI, right lobule Crus I, left lobule Crus I, and left lobule X of the cerebellum with the scores of four items in the ADI-R (reciprocal social interaction subscore; abnormalities in communication subscore; rsetricted, repetitive, and stereotyped patterns of behavior subscore; and abnormalities of developmental evident at or before 36 months subscore). There was a positive correlation between the normalized cortical thickness of the left lobule Crus I and abnormalities of developmental evident at or before 36 months subscore (r=0.3, p=0.028), as shown in Figure 3. Nonetheless, the remaining behavioral scores did not show a significant correlation with normalized cortical thickness.

Figure 3.

Scatter plot of normalized cortical thickness of the left lobule Crus I and ADI-R score in autism patients. This findings illustrated, in ASD a positive correlation between the normalized cortical thickness of the left lobule Crus I and abnormalities of developmental evident at or before 36 months subscore (r=0.3, p=0.028). ADI-R, Autism Diagnostic Interview-Revised; ASD, autism spectrum disorder.

DISCUSSION

The cerebellum, located in the posterior cranial fossa, on the back of the pons and medulla oblongata, accounts for about 10% of the brain’s total volume but contains more neurons than any other part of the brain [15]. MRI, as a multiparametric, multimodal and non-invasive examination technique, is more frequently used in the study of cerebellar structure and function [16-19]. Buckner et al. [20] mapped cerebellar function by analyzing magnetic resonance image data from 1,000 subjects, refined to the cerebellar lobules. That is, each lateral cerebellar hemisphere is divided into three parts: the anterior cerebellar lobe (including lobules I–V), the posterior cerebellar lobe (including lobules VI, Crus I, Crus II, VIIB, VIIIA, VIIIB, and IX), and the choroidal lobule (mainly lobule X).

Based on extensive fiber connections between brain regions, the anterior cerebellar lobe and the choroidal lobule are involved in motor control and regulation of somatic balance, the posterior cerebellar lobe may be involved in memory, attention, emotional control, and even involved in social processes [21]. Thus, cerebellar damage can result in dysfunctions including language, spatial and executive functions as well as emotional dysregulation.

Several studies have reported that the cerebellum is closely related to psychiatric disorders, especially autism [22,23], attention deficit hyperactivity disorder [24], obsessive-compulsive disorder [25], and schizophrenia [26]. However, previous studies have mostly used the cerebellum or lobes as the object, suggesting that there is an increase or decrease in cerebellar or lobar volume in the patient group compared to the normal control group [27,28]. This is because early depictions of the cerebellar contours were manually segmented by experts, which is undoubtedly a tedious, time-consuming process, and large-scale, fine-grained segmentation is impractical. Subsequently, a spatially unbiased atlas template of the cerebellum and brainstem (SUIT) is an automated algorithm based on spatially unbiased mapping templates of the cerebellum and brainstem developed specifically for cerebellar segmentation that has been widely used [29,30]. In SUIT, cerebellar lobules I, II, III, and IV cannot be separated and are collectively referred to as lobules I–IV, making it impossible to study these lobules separately [31,32].

Although smaller in size, lobules I–III are still relevant for separate analysis [33]. In 2017, a patch-based multi-atlas segmentation tool, CERES, was developed to differentiate and measure the 24 lobules of the cerebellum automatically and rapidly by simply providing standard resolution MRI T1-weighted images. Its advanced results have also been shown to be superior to at least five widely used automated cerebellar segmentation methods14 and have been applied in studies of disease cerebellar alterations with good evaluation [7,33]. This is the key reason for choosing this algorithm in this study.

Considering that changes in brain structure inevitably lead to abnormalities in certain brain functions and the greater accessibility of structural MRI data, demonstrating the existence of variability between patients with ASD and healthy populations with the help of morphological alterations in the cerebellum has been the main method of research. Previous studies have found lobular hypoplasia in the cerebellum of patients with ASD. This is mainly manifested by a reduction in the volume of lobules VI and VII [34] and a reduction in the gray matter of lobules Crus I, Crus II, VIII, and IX [17,35,36].

Cortical thickness is a morphological indicators that is related to neurobiological processes [8,37,38]. The normalized cortical thickness is the result of its normalization with the cube root of the intracranial volume [14]. The present study found a significant difference between the ASD and TD groups in the overall mean values of normalized cortical thickness of the 24 cerebellar lobules, suggesting the presence of altered cerebellar structure in patients with ASD compared to typically developing individuals. Then specific to each lobule, it was found that the normalized cortical thickness of the right lobule VI, left lobule VI, right lobule Crus I, left lobule Crus I, and left lobule X of the cerebellum in the ASD group was smaller than that in the TD group, leading to the conclusion that there was cortical thinning in localized areas of the cerebellum in ASD patients. This is consistent with the findings of Hadjikhani et al. [39] who found a thinning of the cerebral cortex belonging to the region of the mirror neuron system. Wang et al. [40] compared the developmental changes in the thickness of the cerebellar lobules in 19 patients with ASD and 14 TD and found thinning of the right Crus II cortical thickness in patients with ASD. These findings are entirely inconsistent with the present study; this may be due to the heterogeneity in ASD or sample population.

There is still an ongoing debate about whether ASD is associated with thinning or thickening of cerebral cortical thickness [40-42]. In a study of cortical thickness in 266 patients with ASD between the ages of 6 and 35 years, it was found that there was a general increase in cortical thickness (mainly on the left side) from the age of 6 years in patients with ASD, but this difference became smaller in adulthood [41,43]. This led to the conjecture of delayed cortical maturation in ASD patients and validated the dynamic nature of morphological changes in the ASD brain. This also explains the complex course of ASD and the high individual heterogeneity of the disease.

Structural differences within the cerebellum are also associated with the core features of ASD deficits, and the reduction in Crus II lobule thickness was accompanied by asymmetry, both of which were associated with the severity of stereotypic behavioral symptoms [39]. This study also found a positive correlation between the normalized cortical thickness of the left lobule Crus I and abnormalities of developmental evident at or before 36 months subscore in patients with ASD, suggesting a role of the cerebellum in the development of ASD in preschool-aged children. This may suggest that a variety of interventions for a child’s preschool years can help him learn critical social, communication, functional, and behavioral skills. The behavioral correlates of abnormalities in normalized cortical thickness in specific lobules of the cerebellum may provide additional ideas for the development of biomarkers and psychiatric therapies that provide potential precise targets for clinical treatment, diagnosis, prediction, and prognostic development of ASD. However, no correlation was found between the thinned cortical thickness and clinical symptoms such as social skills, communication abnormalities, and repetitive specific behaviors in ASD patients. Considering the differences in sample sources, data collection and analysis methods among the groups made the findings not completely consistent.

In summary, the cerebellum has become one of the key brain regions affected by ASD and is gradually becoming a hot spot for research. Using an advanced, automated cerebellar lobule segmentation technique to study morphological differences in the cerebellar lobules, we found that patients with ASD have abnormal structural development of the cerebellar lobules, and correlates positively with clinical symptoms. This further reveals the role of the cerebellum in the pathogenesis of autism and provides additional neural mechanism support for the diagnosis of clinical autism.

Article information Continued