Closed-loop pathways associated with clinical symptoms in children with autism spectrum disorder: a complex network analysis

Introduction

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder characterized by impairments in social communication and interaction, as well as the presence of stereotyped and repetitive behaviors (1). Despite extensive research efforts, the precise pathogenesis of ASD remains elusive, and the diagnosis still relies predominantly on medical history, neuropsychological assessment, and clinical observation, rendering the diagnostic process highly subjective. In recent years, there has been a growing need to identify specific neurobiological markers of the disorder, such as neuroimaging biomarkers (2). Resting-state functional magnetic resonance imaging (rs-fMRI) has emerged as a powerful tool for assisting in the diagnosis of various brain disorders, including ASD (3,4).

Traditional brain network models have found it difficult to identify specific biomarkers for ASD, due to their limited ability to capture the complexity of the human brain. While many network models assume direct connectivities between all brain regions, recent evidence suggests that brain networks exhibit small-world properties, with neighboring regions interacting more frequently, and distant regions relying on critical hubs to transmit and process information. However, the functional significance of each connection remains unclear, and the abnormality of a single functional connection may not be directly correlated with specific clinical indicators. To address this issue, we proposed a novel approach that highlighted the importance of closed-loop pathways in the human brain. Closed-loop pathways refer to network structures composed of multiple interconnected and interacting neurons, as well as connections between specific neural regions. These can involve structural pathways (such as white matter fiber tracts) or functional correlations (such as task-specific or state-specific interactions), with a focus on the latter in this study. Information is transmitted between neurons in a stable, directed cyclic manner, enabling the persistence of specific neural activities or behaviors. The formation of such pathways typically involves synaptic connections and interactions between neurons (5,6). Closed-loop pathways are crucial for the proper functioning of neural networks because each pathway plays a unique role, similar to how each species maintains ecological balance (7). These pathways can be detected as early as the 18−22 weeks gestational age cohort, and continue to develop as the brain matures, such as the long-distance thalamo-cortical loop (8-11). Disruptions in closed-loop pathways have been linked to the development of neurodevelopmental disorders, including ASD, which has been associated with disruptions in cortical-striatal and cortico-thalamic closed-loop pathways (12).

While previous research on neural closed-loop pathways has focused on a neuronal level, investigating these pathways at the level of brain regions is needed. To accomplish this objective, we proposed the use of complex network analysis, which has been proven effective for studying the mechanism of differentiation and integration of brain functions (13). Previous studies have shown that ASD is associated with significant changes in functional connectivity between brain regions (14,15). Thus, various brain network models have been developed to identify their potential relationships with brain diseases, including correlation brain networks and causal brain networks (16).

Causal brain networks are more suitable for closed-loop pathway analysis at the level of brain regions because they can distinguish causal relationships between brain regions and provide additional information that may be helpful in distinguishing ASD from other neurodevelopmental disorders. The present study, therefore, aimed to identify the underlying neurobiological mechanisms responsible for the characteristic symptoms of ASD.

Specifically, we sought to identify and characterize the functional closed-loop pathways in the brains of children with ASD, and determine their potential as neurobiomarkers for this disorder. By gaining a better understanding of the aberrant closed-loop pathways in the brains of individuals with ASD, we hoped to gain insight into the neurobiological basis for the concomitant manifestation of social impairments and stereotyped behaviors in this population. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-93/rc).

Methods

Participants

Participants were recruited from the specific disease cohort of ASD in the Child Mental Health Research Center, The Affiliated Brain Hospital of Nanjing Medical University, and the community. The study sample included 58 ASD patients and 57 typically developing (TD) children between the ages of 6 and 12 years. All participants were right-handed and their parents/guardians agreed to have the children participate in the study.

Inclusion criteria for the ASD group were: (I) meeting the diagnostic criteria for ASD according to the Diagnostic and Statistical Manual of Psychiatric Disorders-fifth edition (DSM-5) (17), diagnosed by at least two deputy chief physicians; (II) ages between 6 and 12 years, and (III) right-handed. Exclusion criteria were: (I) a history of neurological disorders, traumatic brain injury, or serious physical disorders; and (II) presence of a metal implant. Inclusion criteria for the TD group were: (I) ages between 6 and 12 years, and (II) right-handed. Exclusion criteria were the same as for the ASD group (Figure 1).

Figure 1 Flow diagram of study subjects. ADOS, Autism Diagnostic Observation Schedule; ASD, autism spectrum disorder; fMRI, functional magnetic resonance imaging; IQ, intelligence quotient; MRI, magnetic resonance imaging.

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The Affiliated Brain Hospital of Nanjing Medical University (No. 2022-KY022-01). All guardians of subjects provided written informed consent after being informed of the possible risks and benefits of the research.

Clinical psychological assessment

Trained psychologists and psychiatrists at the Child Psychology Research Center of The Affiliated Brain Hospital of Nanjing Medical University conducted clinical and physiological assessments. A self-developed questionnaire was administered to collect general demographic information, growth experiences, clinical histories, and genetic histories of patients. Intelligence quotient (IQ) was assessed using the fourth edition of the Wechsler Scale of Intelligence for Children (18,19). Stereotyped behavior, social interaction and communication abilities were evaluated using the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) (20,21).

Magnetic resonance imaging (MRI) acquisition

All participants refrained from taking neuropsychiatric medications 24 hours prior to the MRI, and were prohibited from carrying any metal objects into the scanning room. MRI data were collected using an 8-channel coil on a 3.0T superconducting MRI system manufactured by Siemens (Munich, Germany). During the examination, patients lay supine with cotton balls in both ears and foam pads on their heads, remaining still with eyes closed. Participants first underwent a routine T2-weighted scan for anatomical localization to eliminate intracranial lesions, followed by sagittal three-dimensional (3D) spoiled gradient-recalled echo (SPGR) sequencing and rs-fMRI. The routine MRI parameters were: repetition time (TR) =2,530 ms, echo time (TE) =3.34 ms, slice thickness =1.33, flip angle =7°, return time =1,100 ms, field of view (FOV) =256 mm × 256 mm, matrix =256 × 192, and total acquisition time of 8 mins and 7 s. The rs-fMRI parameters were: TR =2,000 ms, TE =30 ms, FOV =256 mm × 256 mm, matrix =64 × 64, flip angle =90 °, slice thickness =4 mm, gap =0 mm, and total acquisition time of 8 minutes and 6 seconds, with 30 slices and 140 time-points.

Data preprocessing

We conducted standard preprocessing procedures for fMRI data analysis. The first five volumes of each subject were discarded to minimize noise signals, and the remaining volumes were used for subsequent analysis. All functional images were registered to the T1 image and transformed into the Montreal Neurological Institute space with a resampled voxel size of 3 mm³ × 3 mm³ × 3 mm³. The Conn Toolbox 20b, a Statistical Parametric Mapping-based preprocessing pipeline, was used for outlier detection, directional segmentation and normalization, linear detrending, and functional smoothing with a Gaussian kernel of 6 mm full-width half maximum, etc. Subsequently, time series data for each brain region were extracted from the preprocessed images using the anatomical automatic labeling (AAL) (22) atlas.

We conducted standard preprocessing procedures for fMRI data analysis. The specific formulation was as follows: the Conn Toolbox 20b, a Statistical Parametric Mapping-based preprocessing pipeline, was used for outlier detection, directional segmentation and normalization, linear detrending, and functional smoothing with a Gaussian kernel of 6 mm full-width half maximum, etc. The specific preprocessing steps were as follows. (I) Because it took some time for both the machine and the subjects to enter the steady state, the first five time point images were removed and the remaining time-point data were retained for subsequent analysis. (II) Slice timing correction was performed. (III) Head motion correction: subjects with head translation >2 mm or rotation >2° were removed. (IV) Normalization: all functional images were registered to the T1 image and transformed into the Montreal Neurological Institute space with a resampled voxel size of 3 mm³ × 3 mm³ × 3 mm³. The bounding box involved [−90, −126, −72; 90, 90, 108]. (V) Filter: the filter range was 0.01−0.1 Hz to reduce the effect of physiological noise. (VI) T1 images were divided into gray matter, white matter, and cerebrospinal fluid, using the Friston24 parameter model to remove white matter and cerebrospinal fluid signals. (VII) Data: the Gaussian kernel was smoothed with a 6 mm full-width half-height. The brain was then divided into 90 brain regions (45 for left and right) according to the AAL standard partitioning template. Subsequently, time series data for each brain region were extracted from the preprocessed images using the AAL atlas.

Construction of brain causal and temporal-lag networks

As stated in the Introduction, the causal brain network provided the direction of information flow between brain regions, so it was more suitable for the identification of closed-loop pathways. In the domain of brain networks, various methods have been proposed to estimate causal relationships between brain regions. However, existing causal inference methods ignore the time delay of information transmission, leading to the construction of brain causal networks, which do not reflect the true interactions between brain regions. Therefore, in this study, we used a previously self-designed brain network modeling framework to infer causal relationships between brain regions.

Compared with previous methods, our modeling method had two main advantages. First, it could simultaneously infer the causal relationships and temporal-lag values between brain regions. Second, it was a nonlinear modeling method that had advantages over linear modeling methods. In Figure 1, we present a simple schematic diagram of the model method. The input of the modeling method was the time series data X and its first-order difference D, and the output was the brain causal network and the brain temporal-lag network (23). The implementation code for the algorithm used has been made publicly available on GitHub. The repository can be accessed at https://github.com/NJUSTxiazw/CTLN (Figure 2).

Figure 2 Schematic illustration for constructing brain causal and temporal-lag networks. The modeling method uses the preprocessed time series data X and its first-order difference D as the input, and outputs the brain causal network and the brain temporal-lag network. The causal brain network can characterize nonlinear interactions between brain regions, and the temporal-lag brain network can provide temporal-lag information between brain regions. ETLN, effective temporal lag neural network; GAN, generative adversarial network.

The closed-loop search algorithm and stable closed-loop pathways

To search for closed-loop pathways in this study, we used the simple loop search algorithm proposed by Johnson in 1975 (24), which can efficiently find all simple loops in a graph. Specifically, we provided a binarized adjacency graph as input and obtained all simple loops as output. The time complexity of Johnson’s algorithm was O((n+e)(c+1)), where n denotes the number of nodes, e denotes the number of connected edges, and c denotes the number of simple loops in the adjacency graph.

There were two strategies for obtaining all stable closed-loop pathways. The first method was to identify simple loops for all TD subjects, and then identify those high-frequency common loops among these subjects as stable closed-loop pathways. The second method was to first obtain the mean value of the binarized brain network of all TD subjects, followed by setting the connected edges with weights greater than or equal to 0.8 to 1, and those less than 0.8 to 0. Finally, the average brain network obtained in the second step was used in the Johnson algorithm to obtain all simple loops. It is worth noting that the average brain network in the second step was a necessary but not sufficient condition for the stable closed-loop pathway to be used. If 80% of TD subjects had the loop in their brain network, the edge associated with the closed-loop must be jointly owned by 80% of the subjects. However, the presence of the edge in 80% of the subjects’ brain networks did not mean that the closed-loop associated with the edge was present in 80% of the subjects. Therefore, we needed to further traverse all closed-loop pathways obtained in the previous step, and count whether 80% of the subjects co-owned the closed-loop pathways.

For the above two strategies, we adopted the second one. Although the first approach had only two steps, its search space was very large (obtaining all subjects’ simple loops as well as the matching of common loops), which was very time-consuming. The first two steps of the second approach were equivalent to performing a dimensionality reduction operation on the data, while the third and fourth steps involved searching for and verifying stable closed-loop pathways. Compared with the two strategies, the second approach was much faster. Therefore, we adopted the second approach to obtain stable closed-loop pathways of all normal subjects. However, there were two shortcomings to this strategy. (I) The strategy required execution in four sequential steps. Each step had to be completed in order and could not be performed simultaneously, which led to relatively low machine utilization. (II) To ensure reproducibility of our method, we introduced additional steps in this strategy for further validation. While this may have required extra time, we believe it was worthwhile to guarantee the reliability of the results.

Statistical analysis

SPSS 20.0 (SPSS, Chicago, IL, USA) was used to perform statistical analyses. First, we confirmed that the age and IQ distributions of both the TD and ASD groups were normal, using the Kolmogorov-Smirnov test. Sex distribution was compared using the Chi-square test. In later analyses, age, sex, and IQ were all used as covariates in the correlation analysis to eliminate the influence of these potential factors. To assess differences in the weights of connected edges within stable closed-loop pathways, we used a paired t-test. To avoid false positive closed-loop pathways, we used the false discovery rate (FDR) for multiple data corrections when comparing associations with each identified closed-loop network.

We also examined correlations between stable closed-loop pathways and social impairments and stereotyped behaviors. Specifically, there were two research schemes. First, each functional connection was separately used as an analysis object, and the correlations between them and clinical indicators were analyzed. Second, each closed-loop pathway was used as an analysis object, and the correlation between the average weights of all connected edges of the closed-loop pathway and clinical indicators was calculated.

Finally, we identified the relationships between different symptoms within the ASD group. To more accurately measure the degree of interaction between different circuits, we used the Multivariate Granger Causality Toolbox (MVGC) for statistical analysis (25). MVGC utilizes an autoregressive algorithm that eliminates potential confounding factors, enabling a more precise assessment of the influence between different closed-loop pathways. Two-tailed P values of less than 0.05 were considered statistically significant.

Results

Demographic and clinical characteristics

In this study, 104 subjects were successfully enrolled, with 51 ASD patients and 53 TD children. Eleven subjects were excluded due to incomplete imaging data. Statistical analyses were conducted to compare the demographic characteristics of the two groups. There is no significant difference (P>0.05) in age or sex between the groups. However, there were significant differences in IQ, with the TD group having a significantly higher average IQ than the ASD group (Table 1). These findings suggested that the groups were well-matched for age and sex, but differed in intellectual abilities.

Stable closed-loop pathways

The results, with the FDR for multiple data comparison corrections, showed a total of 84 stable closed-loop pathways, which were grouped into eight independent clusters (Figure 3). The results also showed that there were more connections within the same area than across areas, which was consistent with previous studies. Additionally, the brain network exhibited small-world properties, meaning that adjacent brain regions had a higher degree of interaction. The frontal lobe area had the most complex functional interactions, involving three cluster structures. These structures had interactions within the frontal lobe, but also with brain areas in the parietal and temporal lobes.

Figure 3 Identification results of stable closed-loop pathways in the TD group. There were 84 stable closed-loop pathways in total, which could be divided into eight groups of independent clusters. The same color of the nodes indicates that these brain regions originated from the same area. The brain diagram illustrates the interconnected structures of eight clusters. CAL.L, calcarine fissure and surrounding cortex left; CUN.L, cuneus left; HES.L, Heschl gyrus left; HES.R, Heschl gyrus right; IFGoperc.L, inferior frontal gyrus, opercular part left; IFGoperc.R, inferior frontal gyrus, opercular part right; IFGtriang.L, inferior frontal gyrus, triangular part left; IFGtriang.R, inferior frontal gyrus, triangular part right; INS.R, insula right; IOG.L, inferior occipital gyrus left; IPL.L, inferior parietal, but supramarginal and angular gyri left; LING.L, lingual gyrus left; LING.R, lingual gyrus right; PAL.L, lenticular nucleus, pallidum left; PAL.R, lenticular nucleus, pallidum right; PCG.R, posterior cingulate gyrus right; PCL.L, paracentral lobule left; PCL.R, paracentral lobule right; PCUN.L, precuneus left; PCUN.R, precuneus right; PoCG.R, postcentral gyrus right; PreCG.R, precentral gyrus right; PUT.L, lenticular nucleus, putamen left; PUT.R, lenticular nucleus, putamen right; ROL.L, Rolandic operculum left; ROL.R, Rolandic operculum right; SMA.R, supplementary motor area right; SOG.L, superior occipital gyrus left; SOG.R, superior occipital gyrus right; SPG.L, superior parietal gyrus left; TD, typically developing.

This study also found that many brain regions in the occipital lobe area had stable interactions, forming a large-scale stable cluster involving a total of 11 brain regions. The occipital lobe is responsible for processing visual information, and the closed-loop pathway formed in the 6th cluster may be related to hand-eye coordination and concentration. The study also found that part of the parietal lobe and subcortical brain regions were involved in the formation of stable closed-loop pathways in typical children. Overall, these results suggested that brain function depended on the integration and collaboration of multiple brain regions.

Group comparison

Based on stable closed-loop pathways, we used them as the candidate set of potential biomarkers to examine whether there were intergroup differences in the weights of the edges associated with these stable closed-loop pathways between the TD and ASD groups. The results of intergroup differences were corrected by FDR. Figure 4 shows the visualization results of the differences between groups. Detailed differences in intergroup connectivity are shown in Table S1 and Table S2.

Figure 4 Closed-loop pathways with significant differences between the ASD and TD groups. The motion direction of the ball in each arc represents the causal relationship between two brain regions (from cause to effect). ASD, autism spectrum disorder; CAL.R/L, calcarine fissure and surrounding cortex right/left; CUN.R, cuneus right; HES.R/L, Heschl gyrus right/left; INS.R, insula right; IOG.L, inferior occipital gyrus left. LING.R/L, lingual gyrus right/left; MFG.R, middle frontal gyrus right; MOG.R/L, middle occipital gyrus right/left; PAL.R/L, lenticular nucleus pallidum right/left; PCG.R, posterior cingulate gyrus right; PCL.R, paracentral lobule right; PCUN.R/L, precuneus right/left; PUT.R, lenticular nucleus, putamen right; PUT.R/L, lenticular nucleus putamen right/left; ROL.R, Rolandic operculum right; SFGdor.R, superior frontal gyrus, dorsolateral; SMA.R, supplementary motor area right; SOG.R/L, superior occipital gyrus right/left; SPG.L, superior parietal left; TD, typically developing.

In terms of regional distributions, functional abnormalities in children with ASD were mainly concentrated in the parietal and occipital lobes. From the perspective of the distribution of the left and right hemispheres, the number of brain regions with abnormal functions in the right hemisphere was more than that in the left hemisphere, which coincided with previous studies on the lateralization of brain function in children with ASD (26). In addition, we found a greater number of abnormal functional connections within the ipsilateral hemisphere, but those across the hemisphere appeared to be more stable. For example, Heschl gyrus left (HES.L)→Rolandic operculum right (ROL.R), as well as posterior cingulate gyrus right (PCG.R)→precuneus left (PCUN.L), were stably identified as abnormal connections in both brain network models.

Correlation analysis

We used two schemes to identify correlations between closed-loop pathways and clinical indicators in individuals with ASD. In the first scheme, our analysis did not reveal a significant correlation between functional connectivity and clinical indicators in the ASD group. However, the second scheme yielded promising results. We found a significant correlation between certain closed-loop pathways and clinical indicators, as shown in Figure 5. The x-axis of each correlation graph shown in Figure 5 represents the mean value of the weights of the two edges of the corresponding loop, while the y-axis represents the specific clinical index value. Our novel method showed that the average causal effect of four closed-loop pathways was significantly negatively correlated with clinical indicators. Notably, three of these closed-loop pathways were significantly negatively correlated with the total score of the combined social-communication domain in the ADOS-2, including lenticular nucleus putamen left (PUT.L)→lenticular nucleus pallidum right (PAL.R)→PUT.L (r=−0.448, P=0.001), PAL.R→lenticular nucleus putamen right (PUT.R)→PAL.R (r=−0.362, P=0.012), and insula right (INS.R)→Heschl gyrus right (HES.R)→INS.R (r=−0.345, P=0.016). Moreover, superior parietal left (SPG.L)→inferior parietal, but supramarginal and angular gyri right (IPL.R)→SPG.L (r=−0.308, P=0.03) was significantly negatively correlated with the score of the repetitive behaviors domain in the ADOS-2. Additionally, the average temporal-lag effect of one closed-loop pathway, precentral gyrus right (PreCG.R)→postcentral gyrus right (PoCG.R)→PreCG.R (r=0.031, P=0.313), was significantly positively correlated with the total score of the combined social-communication domain in the ADOS-2.

Figure 5 The relationships between closed-loop pathways and clinical indicators. The ADOS_(S-C) refers to the total score of the combined social-communication domain in the Autism Diagnostic Observation Schedule, and the vertical axis represents the score of the ADOS_Respective behaviors or ADOS_(S-C). (A) The horizontal axis in the correlation graph represents the mean value of all edge weights corresponding to the loop in the brain causal network. (B) The horizontal axis in the correlation graph represents the mean value of all edge weights corresponding to the loop in the brain temporal-lag network. ADOS, Autism Diagnostic Observation Schedule; HES.R, Heschl gyrus right; INS.R, insula right; IPL.L, inferior parietal, but supramarginal and angular gyri left; PAL.R, lenticular nucleus, pallidum right; PoCG.R, postcentral gyrus right; PreCG.R, precentral gyrus right; PUT.L, lenticular nucleus, putamen left; PUT.R, lenticular nucleus, putamen right; SPG.L, superior parietal gyrus left.

Analysis of interaction strengths among closed-loop pathways

To accurately measure the strengths of interactions between these closed-loop pathways, we used MVGC for statistical analyses. Table 2 shows the results of our analyses on the interaction strengths among the four closed-loop pathways associated with social impairment and stereotyped behavior. It is important to note that, due to a shared node (PAL.R), the interaction strength between the 2nd and 3rd closed-loop pathways could not be computed. Our findings revealed a positive correlation in the interaction strength between these closed-loop pathways. Furthermore, we observed that the impact of the 1st and 4th closed-loop pathways on the 2nd and 3rd closed-loop pathways appeared to be greater than that of the latter on the former. Additionally, our analysis suggested that the 4th closed-loop pathway had a dominant effect, with its influence on the other three pathways surpassing that of any other closed-loop pathway.

Discussion

The etiology of ASD is currently unknown. In the past, it had been suggested that ASD is primarily an abnormality in the structure of a particular brain region. More studies have found widespread cortical underconnectivity, local overconnectivity, and mixed results, suggesting disrupted brain connectivity as a potential neural signature of autism. Moreover, this abnormal brain function is closely related to social impairment and stereotyped behaviors in ASD. In functional neuroimaging studies, both fMRI and Functional Near-Infrared Spectroscopy have found abnormalities in the activation or functional connectivity of brain regions related to social cognitive functions in individuals with ASD (27). A Chinese study (28) found increased local functional connectivity in the left temporal subregion and left fusiform gyrus in ASD patients aged 6−18 years, which was correlated with higher stereotyped behavior scores. It has also been reported that the strength of functional connectivity between the striatum and the precentral cortex and the middle frontal gyrus in ASD children and adolescents was positively correlated with the severity of stereotyped behaviors (29,30). In recent years, as fMRI research methods have improved, an increasing number of researchers have proposed that ASD may be associated with abnormalities in large-scale functional networks, rather than in individual brain regions. Multiple studies (31,32) have found abnormalities in brain function in individuals with ASD, especially in the prefrontal cortex, cingulate gyrus, amygdala, and striatum, which may be involved in core symptoms. Abnormal functional connectivity in the Default Mode Network (DMN) has also been frequently reported. ASD children with social and visual engagement difficulties showed reduced functional connectivities between the DMN and visual brain networks, which were correlated with more severe social communication impairments, suggesting that abnormal functional connectivity in the DMN may be a potential mechanism underlying social impairments in ASD (33-35). The fMRI has a high spatial resolution and can measure neural activity in individuals with ASD during rest or task performance, which is beneficial for detecting abnormalities in functional connectivities or activation states. However, its limitations include susceptibility to motion artifacts caused by head movements and breathing, which can be challenging to control in ASD toddlers.

Numerous studies have reported that ASD is characterized by a reduction in the modular properties of brain networks, indicating disrupted brain function integration (36,37). Analysis of closed-loop pathways can therefore provide valuable insights into the pathogenesis of ASD, as well as potential clinical treatment strategies. Our study showed that there were a large number of abnormal neural circuits in the ASD group, which were mainly distributed in the frontal lobe, parietal lobe, and occipital lobes, and some abnormal loop interactions in these brain regions showed a significant negative correlation with social impairment and repetitive behavior in children with ASD. These findings corroborated previous studies and suggested weakened functional interactions between these brain regions in individuals with more severe symptoms (38,39).

Regarding the distributions of temporal-lag network regions, the number of abnormal functional connections within the same hemisphere and across hemispheres was higher than that of causal brain network output results, especially in the frontal and temporal lobes (similar to previous studies using Granger causality analysis and graph theory analysis) (40,41). However, the overall regions with functional abnormalities were basically consistent between the two methods, which were mainly concentrated in the frontal and occipital lobes. This also suggested that previous research methods may have involved excessive analysis, leading to some false-positive neural circuits. When considering the directed condition of “information flow”, correction to some extent could be achieved. However, larger samples and more experiments are still needed for further validation.

In recent years, brain functional studies on ASD have predominantly focused on the frontal and parietal lobes. Our study revealed a significant number of aberrant neural circuits within these brain regions in individuals with ASD. These regions have been implicated in several cognitive and perceptual processes that are commonly disrupted in individuals with ASD, including social cognition, sensory processing, and attention (3). This finding is consistent with other studies that have suggested that ASD is characterized by alterations in functional connectivities within and between different brain regions (14). These findings suggested that the pathophysiology of ASD may involve complex disruptions in neural connectivity. They also highlighted the importance of investigating brain circuits beyond just individual brain regions in this disorder (38).

It is worth noting that our study found that the occipital lobe also played an important role in ASD, which was slightly different from previous studies. The occipital lobe is mainly involved in the regulation of visual function, as well as emotional and cognitive functions. The retina shares many similarities with the brain, including its embryologic origin, anatomy, and neurophysiology (42-44). Furthermore, the retina and brain share similar mechanisms in neuroimmune defenses and responses (45,46). These similarities make the retina a promising area for studying ASD. Studies have shown that children with ASD exhibit visual abnormalities and prefer to use visual ways of thinking (47-50). Studies on ASD have found that the number of weak connections in the occipital lobe was significantly increased, and there was a significant correlation with the severity of communication and social disorders in patients (51,52). We also found that the thickness of the ganglion cell layer and inner nuclear layer (INL) was significantly increased in the ASD group, with a correlation between INL thicknesses and cognitive levels (53). The study of ASD biomarkers based on the retina therefore, has potential in future research and may provide additional insights into ASD. It can also verify the scientific and innovative nature of loop analysis.

The present study investigated the modular properties of brain networks in individuals with ASD, with a particular focus on closed-loop pathways. The results revealed a significant reduction in the functional integration mechanism of the brain in individuals with ASD, which supported previous findings (54,55). Moreover, we conducted further analyses and found significant correlations between four closed-loop pathways and clinical indicators of social impairments and stereotyped behaviors. Using the MVGC toolbox (56), we assessed the strengths of interactions between these pathways and found positive correlations in their interaction strengths, which explained the co-occurrence of stereotyped behaviors and social impairments in individuals with ASD. Studies have shown that targeted interventions can effectively alleviate patients’ clinical symptoms and improve the quality of life, as well as the prognoses of children with ASD (57), making early and accurate identification extremely important. Social impairment and stereotyped behavior are indispensable core symptoms for the diagnosis of ASD, but these symptoms are not prominent in the early stage of the disease, which presents challenges to the diagnosis of ASD. According to data from the 2011–2012 USA National Survey of Children’s Health, of the 1,496 children aged 2–17 years who were diagnosed with ASD, approximately 20% of those with ASD were misdiagnosed with ADHD prior to diagnosis (58). In the present study, we found positive interactions between closed-loop pathways associated with social impairments and those associated with stereotyped behaviors, which may explain why patients with ASD always present with multiple symptoms at the same time. In addition, changes of brain function and structure in neurodevelopmental disorders vary at different ages (59). The subjects enrolled in this study were between 6 and 12 years of age, so in future studies, we plan to expand the size and age range of the data set to further validate the rationality of the proposed methodology.

In addition, when comparing the interaction strengths between different symptoms, it was found that the effects of the 1st and 4th closed-loop pathways on the 2nd and 3rd closed-loop pathways appeared to be greater than that of the latter on the former. In other words, closed-loop pathways in peripheral regions have a greater impact on medial regions of the brain. However, we believe that the reason for the abovementioned phenomenon may be that the 2nd and 3rd closed-loop pathways are located at the junction of the parietal lobe and the frontal lobe of the brain, so they may be involved in information transmission, while the 1st and 4th closed-loop pathways, located in the edge areas, are responsible for the collection or processing of information. Thus, the lateral region (such as the 1st and 4th closed-loop pathways) had a greater influence on the medial region of the brain, which is consistent with previous findings (10,60). However, the first and fourth loops are respectively associated with repetitive behavior and social deficits, and there is no dominant relationship between these two core symptoms of ASD. At the same time, it can be seen from the strength of their interactions that although repetitive behaviors and social deficits are interrelated, their impact is not significant. This is consistent with clinical observations, which once again have proven the reliability of the loop analysis method.

Our analysis also suggested that the 4th closed-loop pathway had a dominant effect, surpassing the impact of any other pathway on the other three. These findings have important implications for understanding the complex brain interactions underlying ASD symptoms and for developing targeted treatments. Previous studies have reported that the insula and temporal lobes play an important role in human emotional regulation and social cognition (56,61), which has an important impact on ASD because of the coexistence of social deficits and repetitive behavior. In the future, further research should be conducted to include emotions related to ASD, to identify the underlying synergy among social impairment, repetitive behavior, and emotional disturbance. When comparing the results of Figure 5A and Figure 5B, it was shown that the closed-loop pathway in Figure 5A was negatively correlated with clinical symptoms, but the closed-loop pathway in Figure 5B was positively correlated with clinical symptoms. This may be due to weakening of the functional connection strength between brain regions, which further leads to a longer delay in the transmission of information between them. Therefore, the average weight of the closed-loop pathway in Figure 5B was positively correlated with clinical symptoms. These findings provided new insights into the pathogenesis of ASD and highlighted the potential of closed-loop pathway analysis in developing clinical treatment strategies.

Limitations

Several limitations should be considered in our study. First, while our study included a significant number of participants, the sample size may still be limited, and caution should be exercised in generalizing the findings to all individuals with ASD. In addition, the rs-fMRI study was mainly focused on children aged 6−12 years, with a higher proportion of male participants. Future research should confirm whether the results of this study are applicable to children of other age groups, adults, and female ASD patients. Second, one-third of ASD children have intellectual disabilities, but our sample was predominantly within the normal range of intelligence, which may limit the generalizability of the research findings. Additionally, the causal modeling method we used was relatively new, which may have limited the conclusions of the study, so further replication studies with larger sample sizes are needed. Furthermore, the complexity of closed-loop pathways makes it challenging to fully comprehend the role of each pathway with only a limited set of clinical indicators. Further research is needed to obtain a more complete understanding of the functions of these pathways in the brains of children with ASD.

Conclusions

In conclusion, our analysis of closed-loop pathways highlighted the cooperative pattern among various brain regions in children with ASD. Our findings were consistent with previous research results and identified several functional closed-loop pathways that correlated with social impairments and stereotyped behaviors. Despite these behaviors frequently co-occurring in individuals with ASD, our results suggested that they had minimal influence on one another. Furthermore, the clustered distribution of numerous abnormal neural closed-loop pathways associated with social impairments and stereotyped behaviors in ASD could serve as potential neurobiomarkers for early diagnosis and treatment. Moreover, our approach may have broad applicability to other brain diseases.

Acknowledgments

We thank International Science Editing (http://www.internationalscienceediting.com) for editing this manuscript.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-93/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-93/dss

Funding: This study was supported by the STI2030-Major Projects (Grant No. 2021ZD0204004) for data collection, Science and Technology Development Foundation, Nanjing Medical University (No. 20210225) and Medical Science and Technology Development Foundation, Nanjing Department of Health (No. YKK21115).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-93/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the Affiliated Brain Hospital of Nanjing Medical University (No. 2022-KY022-01). All guardians of subjects provided written informed consent after being informed of the possible risks and benefits of the research.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Li Z, Yang L, Chen H, Fang Y, Zhang T, Yin X, Man J, Yang X, Lu M. Global, regional and national burden of autism spectrum disorder from 1990 to 2019: results from the Global Burden of Disease Study 2019. Epidemiol Psychiatr Sci 2022;31:e33. [Crossref] [PubMed]
2. Penney AM, Greenson J, Schwartz IS, Estes AM. “On-Time Autism Intervention”: A Diagnostic Practice Framework to Accelerate Access. Front Psychiatry 2022;13:784580. [Crossref] [PubMed]
3. Uddin LQ, Dajani DR, Voorhies W, Bednarz H, Kana RK. Progress and roadblocks in the search for brain-based biomarkers of autism and attention-deficit/hyperactivity disorder. Transl Psychiatry 2017;7:e1218. [Crossref] [PubMed]
4. Kern JK, Geier DA, King PG, Sykes LK, Mehta JA, Geier MR. Shared Brain Connectivity Issues, Symptoms, and Comorbidities in Autism Spectrum Disorder, Attention Deficit/Hyperactivity Disorder, and Tourette Syndrome. Brain Connect 2015;5:321-35. [Crossref] [PubMed]
5. Lewis AS, Calipari ES, Siciliano CA. Toward Standardized Guidelines for Investigating Neural Circuit Control of Behavior in Animal Research. eNeuro 2021;8. ENEURO. [Crossref] [PubMed]
6. Chen P, Hong W. Neural Circuit Mechanisms of Social Behavior. Neuron 2018;98:16-30. [Crossref] [PubMed]
7. Bertelsen N, Landi I, Bethlehem RAI, Seidlitz J, Busuoli EM, Mandelli V, et al. Imbalanced social-communicative and restricted repetitive behavior subtypes of autism spectrum disorder exhibit different neural circuitry. Commun Biol 2021;4:574. [Crossref] [PubMed]
8. McConnell SK, Ghosh A, Shatz CJ. Subplate pioneers and the formation of descending connections from cerebral cortex. J Neurosci 1994;14:1892-907. [Crossref] [PubMed]
9. McConnell SK, Ghosh A, Shatz CJ. Subplate neurons pioneer the first axon pathway from the cerebral cortex. Science 1989;245:978-82. [Crossref] [PubMed]
10. Tau GZ, Peterson BS. Normal development of brain circuits. Neuropsychopharmacology 2010;35:147-68. [Crossref] [PubMed]
11. Supèr H, Soriano E, Uylings HB. The functions of the preplate in development and evolution of the neocortex and hippocampus. Brain Res Brain Res Rev 1998;27:40-64. [Crossref] [PubMed]
12. Courchesne E. Brainstem, cerebellar and limbic neuroanatomical abnormalities in autism. Curr Opin Neurobiol 1997;7:269-78. Erratum in: Curr Opin Neurobiol 1997;7:568. [Crossref] [PubMed]
13. Ren Shen, Li Junhua, Taya F, deSouza J, Thakor NV, Bezerianos A. Dynamic Functional Segregation and Integration in Human Brain Network During Complex Tasks. IEEE Trans Neural Syst Rehabil Eng 2017;25:547-56. [Crossref] [PubMed]
14. Sato W, Uono S. The atypical social brain network in autism: advances in structural and functional MRI studies. Curr Opin Neurol 2019;32:617-21. [Crossref] [PubMed]
15. Sakoğlu U, Pearlson GD, Kiehl KA, Wang YM, Michael AM, Calhoun VD. A method for evaluating dynamic functional network connectivity and task-modulation: application to schizophrenia. MAGMA 2010;23:351-66. [Crossref] [PubMed]
16. Ribeiro AH, Vidal MC, Sato JR, Fujita A. Granger Causality among Graphs and Application to Functional Brain Connectivity in Autism Spectrum Disorder. Entropy (Basel) 2021;23:1204. [Crossref] [PubMed]
17. Edition F. Diagnostic and statistical manual of mental disorders. Am Psychiatric Assoc 2013;21:591-643.
18. Erden G, Yiğit İ, Çelik C, Guzey M. The diagnostic utility of the Wechsler Intelligence Scale for Children-Fourth Edition (WISC-IV) in identification of gifted children. J Gen Psychol 2022;149:371-90. [Crossref] [PubMed]
19. Oliveras-Rentas RE, Kenworthy L, Roberson RB 3rd, Martin A, Wallace GL. WISC-IV profile in high-functioning autism spectrum disorders: impaired processing speed is associated with increased autism communication symptoms and decreased adaptive communication abilities. J Autism Dev Disord 2012;42:655-64. [Crossref] [PubMed]
20. Chojnicka I, Pisula E. Adaptation and Validation of the ADOS-2, Polish Version. Front Psychol 2017;8:1916. [Crossref] [PubMed]
21. Lebersfeld JB, Swanson M, Clesi CD, O’Kelley SE. Systematic Review and Meta-Analysis of the Clinical Utility of the ADOS-2 and the ADI-R in Diagnosing Autism Spectrum Disorders in Children. J Autism Dev Disord 2021;51:4101-14. [Crossref] [PubMed]
22. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O, Delcroix N, Mazoyer B, Joliot M. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 2002;15:273-89. [Crossref] [PubMed]
23. Xia Z, Zhou T, Mamoon S, Alfakih A, Lu J. A Structure-Guided Effective and Temporal-Lag Connectivity Network for Revealing Brain Disorder Mechanisms. IEEE J Biomed Health Inform 2023;27:2990-3001. [Crossref] [PubMed]
24. Johnson DB. Finding all the elementary circuits of a directed graph. SIAM Journal on Computing 1975;4:77-84.
25. Barnett L, Seth AK. The MVGC multivariate Granger causality toolbox: a new approach to Granger-causal inference. J Neurosci Methods 2014;223:50-68. [Crossref] [PubMed]
26. Conti E, Calderoni S, Gaglianese A, Pannek K, Mazzotti S, Rose S, Scelfo D, Tosetti M, Muratori F, Cioni G, Guzzetta A. Lateralization of Brain Networks and Clinical Severity in Toddlers with Autism Spectrum Disorder: A HARDI Diffusion MRI Study. Autism Res 2016;9:382-92. [Crossref] [PubMed]
27. Guo Z, Tang X, Xiao S, Yan H, Sun S, Yang Z, Huang L, Chen Z, Wang Y. Systematic review and meta-analysis: multimodal functional and anatomical neural alterations in autism spectrum disorder. Mol Autism 2024;15:16. [Crossref] [PubMed]
28. Su X. Brain structure and function study of autism spectrum disorder patients based on VBM and resting-state functional magnetic resonance [dissertation]. Xinxiang Medical University; 2023.
29. Akkermans SEA, Rheinheimer N, Bruchhage MMK, Durston S, Brandeis D, Banaschewski T, Boecker-Schlier R, Wolf I, Williams SCR, Buitelaar JK, van Rooij D, Oldehinkel MTACTICS consortium. Frontostriatal functional connectivity correlates with repetitive behaviour across autism spectrum disorder and obsessive-compulsive disorder. Psychol Med 2019;49:2247-55. [Crossref] [PubMed]
30. Abbott AE, Linke AC, Nair A, Jahedi A, Alba LA, Keown CL, Fishman I, Müller RA. Repetitive behaviors in autism are linked to imbalance of corticostriatal connectivity: a functional connectivity MRI study. Soc Cogn Affect Neurosci 2018;13:32-42. [Crossref] [PubMed]
31. Ruan L, Chen G, Yao M, Li C, Chen X, Luo H, Ruan J, Zheng Z, Zhang D, Liang S, Lü M. Brain functional gradient and structure features in adolescent and adult autism spectrum disorders. Hum Brain Mapp 2024;45:e26792. [Crossref] [PubMed]
32. Blume J, Dhanasekara CS, Kahathuduwa CN, Mastergeorge AM. Central Executive and Default Mode Networks: An Appraisal of Executive Function and Social Skill Brain-Behavior Correlates in Youth with Autism Spectrum Disorder. J Autism Dev Disord 2024;54:1882-96. [Crossref] [PubMed]
33. Odriozola P, Dajani DR, Burrows CA, Gabard-Durnam LJ, Goodman E, Baez AC, Tottenham N, Uddin LQ, Gee DG. Atypical frontoamygdala functional connectivity in youth with autism. Dev Cogn Neurosci 2019;37:100603. [Crossref] [PubMed]
34. Funakoshi Y, Harada M, Otsuka H, Mori K, Ito H, Iwanaga T. Default mode network abnormalities in children with autism spectrum disorder detected by resting-state functional magnetic resonance imaging. J Med Invest 2016;63:204-8. [Crossref] [PubMed]
35. Lombardo MV, Eyler L, Moore A, Datko M, Carter Barnes C, Cha D, Courchesne E, Pierce K. Default mode-visual network hypoconnectivity in an autism subtype with pronounced social visual engagement difficulties. Elife 2019;8:e47427. [Crossref] [PubMed]
36. Kim S, Kim YE, Song I, Ujihara Y, Kim N, Jiang YH, Yin HH, Lee TH, Kim IH. Neural circuit pathology driven by Shank3 mutation disrupts social behaviors. Cell Rep 2022;39:110906. [Crossref] [PubMed]
37. Saunders A, Kirk IJ, Waldie KE. Hemispheric Coherence in ASD with and without Comorbid ADHD and Anxiety. Biomed Res Int 2016;2016:4267842. [Crossref] [PubMed]
38. Shao L, Fu C, You Y, Fu D. Classification of ASD based on fMRI data with deep learning. Cogn Neurodyn 2021;15:961-74. [Crossref] [PubMed]
39. Li Y, Zhu Y, Nguchu BA, Wang Y, Wang H, Qiu B, Wang X. Dynamic Functional Connectivity Reveals Abnormal Variability and Hyper-connected Pattern in Autism Spectrum Disorder. Autism Res 2020;13:230-43. [Crossref] [PubMed]
40. Li D, Liu C, Huang Z, Li H, Xu Q, Zhou B, Hu C, Zhang Y, Wang Y, Nie J, Qiao Z, Yin D, Xu X. Common and Distinct Disruptions of Cortical Surface Morphology Between Autism Spectrum Disorder Children With and Without SHANK3 Deficiency. Front Neurosci 2021;15:751364. [Crossref] [PubMed]
41. Kim D, Lee JY, Jeong BC, Ahn JH, Kim JI, Lee ES, Kim H, Lee HJ, Han CE. Overconnectivity of the right Heschl’s and inferior temporal gyrus correlates with symptom severity in preschoolers with autism spectrum disorder. Autism Res 2021;14:2314-29. [Crossref] [PubMed]
42. Doron R, Lev M, Wygnanski-Jaffe T, Moroz I, Polat U. Development of global visual processing: From the retina to the perceptive field. PLoS One 2020;15:e0238246. [Crossref] [PubMed]
43. Thomas MG, Papageorgiou E, Kuht HJ, Gottlob I. Normal and abnormal foveal development. Br J Ophthalmol 2022;106:593-9. [Crossref] [PubMed]
44. Chokron S, Dutton GN. From vision to cognition: potential contributions of cerebral visual impairment to neurodevelopmental disorders. J Neural Transm (Vienna) 2023;130:409-24. [Crossref] [PubMed]
45. Oertel FC, Havla J, Roca-Fernández A, Lizak N, Zimmermann H, Motamedi S, Borisow N, White OB, Bellmann-Strobl J, Albrecht P, Ruprecht K, Jarius S, Palace J, Leite MI, Kuempfel T, Paul F, Brandt AU. Retinal ganglion cell loss in neuromyelitis optica: a longitudinal study. J Neurol Neurosurg Psychiatry 2018;89:1259-65. [Crossref] [PubMed]
46. Waldman AT, Benson L, Sollee JR, Lavery AM, Liu GW, Green AJ, Waubant E, Heidary G, Conger D, Graves J, Greenberg B. Interocular Difference in Retinal Nerve Fiber Layer Thickness Predicts Optic Neuritis in Pediatric-Onset Multiple Sclerosis. J Neuroophthalmol 2021;41:469-75. [Crossref] [PubMed]
47. Cheng N, Pagtalunan E, Abushaibah A, Naidu J, Stell WK, Rho JM, Sauvé Y. Atypical visual processing in a mouse model of autism. Sci Rep 2020;10:12390. [Crossref] [PubMed]
48. Rasmussen R, Matsumoto A, Dahlstrup Sietam M, Yonehara K. A segregated cortical stream for retinal direction selectivity. Nat Commun 2020;11:831. [Crossref] [PubMed]
49. Simmons DR, Robertson AE, McKay LS, Toal E, McAleer P, Pollick FE. Vision in autism spectrum disorders. Vision Res 2009;49:2705-39. [Crossref] [PubMed]
50. Spiteri S, Crewther D. Neural Mechanisms of Visual Motion Anomalies in Autism: A Two-Decade Update and Novel Aetiology. Front Neurosci 2021;15:756841. [Crossref] [PubMed]
51. Choi US, Kim SY, Sim HJ, Lee SY, Park SY, Jeong JS, Seol KI, Yoon HW, Jhung K, Park JI, Cheon KA. Abnormal brain activity in social reward learning in children with autism spectrum disorder: an fMRI study. Yonsei Med J 2015;56:705-11. [Crossref] [PubMed]
52. Nickl-Jockschat T, Habel U, Michel TM, Manning J, Laird AR, Fox PT, Schneider F, Eickhoff SB. Brain structure anomalies in autism spectrum disorder--a meta-analysis of VBM studies using anatomic likelihood estimation. Hum Brain Mapp 2012;33:1470-89. [Crossref] [PubMed]
53. Wang Y, Zhu Z, Wang Y, Li M, Ma X, Huang K, Gao H, Zhang Y, Chen Q, Zhang W, Ke X. Relationship between autism spectrum disorder and peripapillary intraretinal layer thickness: a pediatric retrospective cross-sectional study. Quant Imaging Med Surg 2024;14:8347-60. [Crossref] [PubMed]
54. Yang J, Lee J. Different aberrant mentalizing networks in males and females with autism spectrum disorders: Evidence from resting-state functional magnetic resonance imaging. Autism 2018;22:134-48. [Crossref] [PubMed]
55. Conti E, Mitra J, Calderoni S, Pannek K, Shen KK, Pagnozzi A, Rose S, Mazzotti S, Scelfo D, Tosetti M, Muratori F, Cioni G, Guzzetta A. Network over-connectivity differentiates autism spectrum disorder from other developmental disorders in toddlers: A diffusion MRI study. Hum Brain Mapp 2017;38:2333-44. [Crossref] [PubMed]
56. Bathelt J, Geurts HM. Difference in default mode network subsystems in autism across childhood and adolescence. Autism 2021;25:556-65. [Crossref] [PubMed]
57. Zwaigenbaum L, Bauman ML, Choueiri R, Fein D, Kasari C, Pierce K, et al. Early Identification and Interventions for Autism Spectrum Disorder: Executive Summary. Pediatrics 2015;136:S1-9. [Crossref] [PubMed]
58. Miodovnik A, Harstad E, Sideridis G, Huntington N. Timing of the Diagnosis of Attention-Deficit/Hyperactivity Disorder and Autism Spectrum Disorder. Pediatrics 2015;136:e830-7. [Crossref] [PubMed]
59. Fu L, Wang Y, Fang H, Xiao X, Xiao T, Li Y, Li C, Wu Q, Chu K, Xiao C, Ke X. Longitudinal Study of Brain Asymmetries in Autism and Developmental Delays Aged 2–5 Years. Neuroscience 2020;432:137-49. [Crossref] [PubMed]
60. Pua EPK, Thomson P, Yang JY, Craig JM, Ball G, Seal M. Individual Differences in Intrinsic Brain Networks Predict Symptom Severity in Autism Spectrum Disorders. Cereb Cortex 2021;31:681-93. [Crossref] [PubMed]
61. Qian L, Wang Y, Chu K, Li Y, Xiao C, Xiao T, Xiao X, Qiu T, Xiao Y, Fang H, Ke X. Alterations in hub organization in the white matter structural network in toddlers with autism spectrum disorder: A 2-year follow-up study. Autism Res 2018;11:1218-28. [Crossref] [PubMed]