Background:Individuals with acute psychiatric disorders attempt suicide or engage in self-harm even during inpatient hospitalization. Despite ongoing clinical monitoring, accurately identifying fluctuating suicide risk remains challenging in acute psychiatric inpatient settings. Recently, ecological momentary assessment (EMA) has been increasingly investigated as a promising approach for predicting suicide risk by capturing dynamic changes in patients' mental states. This study investigated whether wearable-derived passive EMA features of sleep and activity, combined with baseline clinical variables, could support daily morning triage for acute suicide risk in psychiatric inpatients admitted to a closed ward. Objective:This exploratory pilot study evaluated whether wearable-derived sleep and activity features could complement baseline clinical variables for daily morning risk triage among psychiatric inpatients in a closed ward. Methods:We conducted a prospective observational pilot study of 87 enrolled psychiatric inpatients. Of these, 84 had valid Columbia-Suicide Severity Rating Scale (C-SSRS) assessments, and 69 contributed 151 assessment-linked records with both valid C-SSRS labels and temporally aligned wearable-derived sleep/activity features. Participants wore Fitbit Sense devices throughout hospitalization to collect passive sleep and activity data. Physical activity was summarized into 14 non-overlapping 2-hour windows spanning the previous day and assessment morning, ending at the 10:00 AM C-SSRS assessment. These features were combined with baseline clinical variables, including demographics and baseline C-SSRS score, to develop an L1-penalized logistic regression (LASSO) model. Performance was evaluated using recall (sensitivity) and the F2-score. Results:The multimodal fusion model showed numerically higher recall than the conventional assessment model (0.560 vs 0.289) and a higher F2-score (0.548 vs 0.300). However, the 95% confidence intervals overlapped substantially across models; therefore, these findings should be interpreted as exploratory and hypothesis-generating rather than as confirmatory evidence of model superiority. The fusion model identified C-SSRS-positive records from patients with low admission scores using wearable-derived activity-pattern features, particularly blunted morning activity (08:00-10:00) and nocturnal hyperactivity (22:00-24:00). Conclusion:These exploratory findings suggest that wearable-derived sleep and activity features may provide complementary information for daily morning risk triage in psychiatric inpatients. Larger studies are needed to validate whether this approach can support routine clinical review beyond baseline clinical variables.

The COVID-19 outbreak has highlighted the importance of mathematical epidemic models like the Susceptible-Infected-Recovered (SIR) model, for understanding disease spread dynamics. However, enhancing their predictive accuracy complicates parameter estimation. To address this, we proposed a novel model that integrates traditional mathematical modeling with deep learning which has shown improved predicted power across diverse fields. The proposed model includes a simple artificial neural network (ANN) for regional disease incidences, and a graph convolutional neural network (GCN) to capture spread to adjacent regions. GCNs are a recent deep learning algorithm designed to learn spatial relationship from graph-structured data. We applied the model to COVID-19 incidences in Spain to evaluate its performance. It achieved a 0.9679 correlation with the test data, outperforming previous models with fewer parameters. By leveraging the efficient training methods of deep learning, the model simplifies parameter estimation while maintaining alignment with the mathematical framework to ensure interpretability. The proposed model may allow the more robust and insightful analyses by leveraging the generalization power of deep learning and theoretical foundations of the mathematical models.

Amyloid-β accumulation is a pivotal factor in Alzheimer’s disease (AD) progression. As treatment for AD has not been successful yet, the most effective approach lies in early diagnosis and the subsequent delay of disease progression. Hence, this study introduces a deep learning model to predict amyloid-β accumulation in the brain. We mathematically modeled the diffusion of amyloid-β based on its biological traits, encompassing generation, clearance, and diffusion. We converted the model into a deep learning framework with multi-layer perceptron (MLP) and graph convolutional neural network (GCN) (Kipf et al., 2016) to forecast the accumulation of the protein. We extracted the necessary information from various neuroimage data, including T1 structural magnetic resonance (MR) images, 18 F-Florbetapir positron emission tomography (PET) scans, and diffusion weighted MR images (DWI), to simulate the diffusion of the protein (Figure 1). We used longitudinal data of 146 subjects, incorporating 436 data points. The proposed model accurately predicted amyloid-β after 2 years (Figure 2), showing a high correlation in the test dataset (median = 0.8273, IQR = [0.7708, 0.8692]), outperforming the previous model (average 0.58) (Kim et al., 2019). We examined generation and clearance terms, mapping top 30% ROIs onto the brain by averaging each term across subjects (Figure 3). The regions with early AD amyloid-β accumulation are believed to be related to the default mode network and prefrontal network (Palmqvist et al., 2017) supported by Figure 3a. The effectiveness of amyloid-β clearance may be influenced by brain activity (Mergenthaler et al., 2013; Ullah et al., 2023). Earlier studies reported diminished metabolism in specific regions during the early AD (Chételat et al., 2020; Kantarci et al., 2021). The proposed model identified high clearance regions (Figure 3b), aligning with regions showing normal metabolism. We introduced a deep learning model that simulates the diffusion of amyloid-β with strong predictive performance and interpretation. While parameters were optimized for the entire group, accuracy varied for some subjects. Also, further investigation is needed to interpret each term comprehensively. Despite the need for individual optimization and additional interpretative analysis, the model may contribute to the diagnosis of AD.

BACKGROUND:Fatigue is a common symptom in myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD), but its underlying mechanisms are not well understood. OBJECTIVE:To investigate MRI volumetric changes associated with fatigue in MOGAD patients. METHODS:Forty-three MOGAD patients and 29 healthy controls (HC) were recruited from January 2020 to December 2022. Brain volumes, including the brainstem and its subregions (midbrain, superior cerebellar peduncle [SCP], pons, and medulla), were analyzed using FreeSurfer v7.1. Volumetric measures were compared between groups, adjusted for age, sex, and intracranial volume. Correlations between Fatigue Severity Scale (FSS) scores and brain volumes were examined, and predictors of fatigue severity were further explored using multiple linear regression analysis. RESULTS:Fatigue was present in 30.2 % of MOGAD patients. Fatigued patients had significantly reduced medulla volume compared to HC (p = 0.004) and exhibited higher Expanded Disability Status Scale scores, greater depression, and poorer sleep quality, and lower quality of life compared to non-fatigued patients. Higher fatigue scores correlated with smaller brainstem volumes, including the medulla (r = -0.368, p = 0.019), pons (r = -0.331, p = 0.037), and SCP (r = -0.338, p = 0.033). In multiple regression analysis, reduced medulla (B = -0.014, p = 0.019) and SCP volumes (B = -0.188, p = 0.008) were significant predictors of fatigue, while other clinical factors were not. CONCLUSIONS:Fatigue in MOGAD is associated with reduced brainstem volumes, particularly in the medulla, suggesting a role of the brainstem in the pathophysiology of fatigue.

Post-stroke complex regional pain syndrome (CRPS) is a challenging complication that impairs recovery during stroke rehabilitation, particularly during the subacute phase. Despite its clinical significance, the neural substrates underlying post-stroke CRPS, specifically following subcortical stroke, remain unclear. This retrospective observational study included 40 patients with first-ever subcortical stroke diagnosed with CRPS via a three-phase bone scan, and 40 propensity score-matched controls without CRPS. White matter tract involvement was analyzed using atlas-based lesion mapping and voxel-based lesion overlap analysis in patients with available clinical scores and imaging findings (magnetic resonance imaging or computed tomography). Between-group comparisons of white matter tract involvement were conducted with false discovery rate (FDR) correction. Clinical characteristics were similar between groups, except for fewer CRPS patients with a shoulder flexor manual muscle test score ≥3. Lesion overlap with the cingulum in the cingulate gyrus was significantly greater in the CRPS group (F = 5.197, FDR-adjusted p = 0.040). Although the forceps minor showed marginal significance before correction, it was non-significant after adjustment. These findings raise the possibility that the cingulate cortex, particularly the cingulum, may contribute to post-stroke CRPS pathophysiology. However, further confirmation in larger prospective studies is needed.

Background: Multiple sclerosis (MS) and neuromyelitis optica spectrum disorders (NMOSD) are both chronic inflammatory demyelinating diseases of the central nervous system, but they have distinct pathophysiological features that distinguish their clinical phenotypes and lesion characteristics. The objective of this study was to investigate the correlation between white matter (WM) lesion location and previous relapse activity, as well as disease duration, in these two diseases. Methods: This study included 64 patients with relapsing-remitting MS and 49 with NMOSD. We used the voxel-based lesion-symptom mapping (VLSM) method to determine the correlations of the presence of WM lesions with the number of attacks and the disease duration in each disease group. Results: We found that WM lesions were correlated with the number of attacks; the deep WM of the right parietotemporal region including parts of the superior longitudinal fasciculus in MS patients and the right superior corona radiata, corticospinal tract, and inferior fronto-occipital fasciculus in NMOSD patients. However, there were no specific locations associated with disease duration in patients with either disease. Conclusion: Our VLSM analysis confirmed that prior relapse activity was associated with distinct WM lesion locations in MS and NMOSD, respectively. In contrast, disease duration, independent of disease activity, showed no association with specific lesion patterns in either disease.

BACKGROUND Optic neuritis is a common manifestation of myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD), which displays distinctive features that differ from other demyelinating diseases. OBJECTIVES In this study, we investigated the alterations in white matter (WM) connectivity post-optic neuritis in patients with MOGAD. METHODS We enrolled patients with MOGAD presenting with optic neuritis (MOGAD-ON) (N=18) and healthy controls (HC) (N=35). Structural connections between any pair of 90 cortical and subcortical regions were established using diffusion tensor imaging and graph theory. Network-based statistics were employed to identify disrupted patterns in WM networks. Additionally, we investigated the association between WM network integrity and visual acuity. RESULTS Analysis of WM networks revealed that MOGAD-ON patients exhibited reduced total strength, global efficiency, and local efficiency compared to HCs (all, p<0.05). At the nodal level, MOGAD-ON patients showed more regions with altered network topologies than HCs, some of which were associated with visual acuity and visual functional scores. NBS identified the most discriminative connectivity changes in the left hippocampus, thalamus, inferior temporal gyrus, fusiform gyrus, and middle temporal occipital gyrus. CONCLUSION This study indicates that MOGAD-ON patients experience significant disruptions in WM networks, particularly in the left temporal and occipital lobes.

Alzheimer's disease (AD) is a severe neurodegenerative disease characterized by ongoing brain tissue decline. Since amyloid- $\beta$ accumulation is supposed as a primary cause of AD progression, mathematical models have been recently employed to simulate amyloid- $\beta$ diffusion in our brain. These models require parameter estimation through observed data and an iterative process to update parameters, although increased complexity can hinder their optimization. To overcome these difficulties, we converted the mathematical model of amyloid- $\beta$ diffusion into a deep learning model, combining multi-layer perceptron (MLP) and graph convolutional neural network (GCN). The proposed model well predicted the change of accumulation level of amyloid- $\beta$ with a high accuracy.

Understanding the morphology of amyloid fibrils is crucial for comprehending the aggregation and degradation mechanisms of abnormal proteins implicated in various diseases, such as Alzheimer's disease, Parkinson's disease, type II diabetes, and various forms of amyloidosis. Atomic force microscopy (AFM) stands as the most representative method for studying amyloid fibril morphology. However, obstacles in AFM images, including noise, salt, and amorphous aggregates, often impede accurate sample quantification. In this study, we developed denoising software employing a U-Net deep learning architecture to address this issue. The software efficiently eliminated various impediments that interfere with fibril analysis in noisy AFM images, thereby facilitating precise quantification of amyloid fibrils. We also developed automated fibril analysis technologies using the denoised AFM images, leading to quicker, more precise, and more objective assessments of fibril morphology. Furthermore, we presented a method for fibril stiffness extraction from a modulus image through mask creation based on a denoised height image. Our approach secures time efficiency and precision in analyzing amyloid morphology, and we believe it will significantly advance the currently stagnant research on amyloid-related diseases.

BACKGROUND:Autonomy support, which involves providing individuals the ability to control their own behavior, is associated with improved motor control and learning in various populations in clinical and non-clinical settings. This study aimed to investigate whether autonomy support combined with an information technology (IT) device facilitated success in using the more-affected arm during training in individuals with stroke. Consequently, we examined whether increased success influenced the use of the more-affected arm in mild to moderate subacute to chronic stroke survivors. METHODS:Twenty-six participants with stroke were assigned to the autonomy support or control groups. Over a 5-week period, training and test sessions were conducted using the Individualized Motivation Enhancement System (IMES), a device developed specifically for this study. In the autonomy support group, participants were able to adjust the task difficulty parameter, which controlled the time limit for reaching targets. The control group did not receive this option. The evaluation of the more-affected arm's use, performance, and impairment was conducted through clinical tests and the IMES. These data were then analyzed using mixed-effect models. RESULTS:In the IMES test, both groups showed a significant improvement in performance (p < 0.0001) after the training period, without any significant intergroup differences (p > 0.05). However only the autonomy support group demonstrated a significant increase in the use of the more-affected arm following the training (p < 0.001). Additionally, during the training period, the autonomy support group showed a significant increase in successful experiences with using the more-affected arm (p < 0.0001), while the control group did not exhibit the same level of improvement (p > 0.05). Also, in the autonomy support group, the increase in the use of the more-affected arm was associated with the increase in the successful experience significantly (p = 0.007). CONCLUSIONS:Combining autonomy support with an IT device is a practical approach for enhancing performance and promoting the use of the more-affected upper extremity post-stroke. Autonomy support facilitates the successful use of the more-affected arm, thereby increasing awareness of the training goal of maximizing its use. TRIAL REGISTRATION:The study was registered retrospectively with the Clinical Research Information Service (KCT0008117; January 13, 2023; https://cris.nih.go.kr/cris/search/detailSearch.do/23875 ).

Automatic V th extraction with ML is proposed. Various V th extraction methods have been suggested for different types of FETs. It takes lots of time and effort to find an appropriate extraction method for each device, and V th can be differently estimated depending on the insight of researcher. These problems are solved through an objective evaluation method using ML. The final model is represented by kNN regression with normalized both V G and I D .

Background: Motivation to use the more-affected arm is an essential indicator of recovery in stroke survivors. This study aimed to investigate whether personal mastery experience via intensive repetitive reaching movements with autonomy support may increase self-efficacy and thus increase performance and use of the more-affected arm in mild-to-moderate subacute to chronic stroke patients. Methods: Twenty-six participants with stroke were divided into two groups: a motivation group (with autonomy support) and a control group (without autonomy support). Five weeks of training and test sessions were administered using the individualized motivation enhancement system that we developed. The task difficulty parameter modulated the time limit for attaining targets to provide autonomy support. We analyzed various clinical and behavioral measures using mixed-effect models. Results: Successful experiences did not change in the control group (p = .129),but dramatically increased in the motivation group (p < .0001). Performance significantly improved in the retention test for both groups (p < .0001), without any group differences (p = .329). However, the motivation group exhibited a dramatic increase in the use of the more-affected arm (p < .0001), whereas the control group did not (p Conclusions: The successful experience of personal mastery accomplished by autonomy support increased the use of the affected arm. Autonomy support in the motivation group may make a participant aware of the training goal: to use the more-affected arm as much as possible or make the affected arm use more habitual. Trial registration: The study was registered with The Clinical Research Information Service (CRIS), KCT0008117. Registered retrospectively on January 13, 2023, at https://cris.nih.go.kr/cris/search/detailSearch.do/23875

Fluid-attenuated inversion recovery imaging (FLAIR) is a magnetic resonance (MR) method that is frequently utilized to diagnose brain lesions. However, it usually provides transverse section imaging in high resolution but with very low resolution in the other axis. This non-isotropy huddles its wide utilization in research including machine learning. In this study, we applied a deep-learning based super resolution (SR) technique to convert non-isotropic FLAIR images into isotropic one. We proposed a multi-input 3D ResUnet that uses both FLAIR and T1-weighted MR images as its inputs. As a result, our proposed model successfully reconstructed isotropic FLAIR images (SSIM: 0.9947 $\pm 0.0009)$ , and out-performed the FLAIR only single-input 3D ResUnet.

A fast and precise threshold voltage (Vth) extraction method is required for the process design of electronic systems using metal–oxide–semiconductor field‐effect transistors (MOSFETs) and its immediate on‐site analysis during fabrication. The selection of a suitable Vth extraction method is a complicated task because it involves a trade‐off between accuracy and simplicity according to the device scheme. Herein, an automatic‐prediction method of the MOSFET Vth using machine learning (ML) is proposed. The ML model is trained with Vth, extracted using different methods (2nd derivative, constant current, and Y‐function) and from various kinds of FETs (finFET, 2D FET, and metal–oxide thin‐film transistors). The concept of threshold ratio (Rth) for universal Vth prediction, which considers the normalized Vth within certain VG ranges, is suggested. The precision and accuracy of ML models are statistically verified by calculating the root mean square error (RMSE), mean absolute error, and mean coefficients of determination (R2) values. The universal ML model (k‐nearest neighbor (kNN)) achieves 1.35% of RMSE and 0.98 of R2 for the best score. The ML model eliminates the ambiguity in Vth extraction and provides objective Vth prediction for most FET schemes used in the semiconductor industry and research field.

Multiple sclerosis (MS) and neuromyelitis optica spectrum disorder (NMOSD) are two representative chronic inflammatory demyelinating disorders of the central nervous system. We aimed to determine and compare the alterations of white matter (WM) connectivity between MS, NMOSD, and healthy controls (HC). This study included 68 patients with relapsing-remitting MS, 50 with NMOSD, and 26 HC. A network-based statistics method was used to assess disrupted patterns in WM networks. Topological characteristics of the three groups were compared and their associations with clinical parameters were examined. WM network analysis indicated that the MS and NMOSD groups had lower total strength, clustering coefficient, global efficiency, and local efficiency and had longer characteristic path length than HC, but there were no differences between the MS and NMOSD groups. At the nodal level, the MS group had more brain regions with altered network topologies than did the NMOSD group when compared with the HC group. Network alterations were correlated with Expanded Disability Status Scale score and disease duration in both MS and NMOSD groups. Two distinct subnetworks that characterized the disease groups were also identified. When compared with NMOSD, the most discriminative connectivity changes in MS were located between the thalamus, hippocampus, parahippocampal gyrus, amygdala, fusiform gyrus, and inferior and superior temporal gyri. In conclusion, MS patients had greater network dysfunction compared to NMOSD and altered short connections within the thalamus and inferomedial temporal regions were relatively spared in NMOSD compared with MS.

Prism Adaptation (PA) is used to alleviate spatial neglect. We combined immersive virtual reality with a depth-sensing camera to develop virtual prism adaptation therapy (VPAT), which block external visual cues and easily quantify and monitor errors than conventional PA. We conducted a feasibility study to investigate whether VPAT can induce behavioral adaptations by measuring after-effect and identifying which cortical areas were most significantly activated during VPAT using functional near-infrared spectroscopy (fNIRS). Fourteen healthy subjects participated in this study. The experiment consisted of four sequential phases (pre-VPAT, VPAT-10°, VPAT-20°, and post-VPAT). To compare the most significantly activated cortical areas during pointing in different phases against pointing during the pre-VPAT phase, we analyzed changes in oxyhemoglobin concentration using fNIRS during pointing. The pointing errors of the virtual hand deviated to the right-side during early pointing blocks in the VPAT-10° and VPAT-20° phases. There was a left-side deviation of the real hand to the target in the post-VPAT phase, demonstrating after-effect. The most significantly activated channels during pointing tasks were located in the right hemisphere, and possible corresponding cortical areas included the dorsolateral prefrontal cortex and frontal eye field. In conclusion, VPAT may induce behavioral adaptation with modulation of the dorsal attentional network.

BackgroundUnilateral spatial neglect (USN) is common and associated with poor motor and cognitive outcomes as well as impaired quality of life following stroke. Traditionally, the neural substrates underlying USN have been thought to be cortical areas, such as the posterior parietal cortex. However, patients with stroke involving only subcortical structures may also present with USN. While only a few studies have reported on USN in subcortical stroke, the involvement of white matter tracts related to brain networks of visuospatial attention is one possible explanation for subcortical neglect. Therefore, this study aimed to investigate which specific white matter tracts are neural substrates for USN in patients with subcortical stroke.MethodsTwenty-two patients with subcortical stroke without cortical involvement who were admitted to the Department of Rehabilitation Medicine at Seoul National University Bundang Hospital were retrospectively enrolled. Nine subjects were subclassified into a “USN(+)” group, as they had at least two positive results on three tests (the Schenkenberg line bisection test, Albert's test, and house drawing test) and a score of 1 or higher on the Catherine Bergego scale. The remaining 13 subjects without abnormalities on those tests were subclassified into the “USN(–)” group. Stroke lesions on MRI were manually drawn using MRIcron software. Lesion overlapping and atlas-based analyses of MRI images were conducted. The correlation was analyzed between the overlapped lesion volumes with white matter tracts and the severity of USN (in the Albert test and the Catherine Bergego scale).ResultsLesions were more widespread in the USN(+) group than in the USN(–) group, although their locations in the right hemisphere were similar. The atlas-based analyses identified that the right cingulum in the cingulate cortex, the temporal projection of the superior longitudinal fasciculus, and the forceps minor significantly overlapped with the lesions in the USN(+) group than in the USN(–) group. The score of the Catherine Bergego scale correlated with the volume of the involved white matter tracts.ConclusionIn this study, white matter tracts associated with USN were identified in patients with subcortical stroke without any cortical involvement. Our study results, along with previous findings on subcortical USN, support that USN may result from damage to white matter pathways.

Background: Elderly patients with late-life depression (LLD) often report mild cognitive impairment (MCI), so Alzheimer's disease (AD) is hard to identify in these patients. We aimed to identify the structural and functional differences between prodromal AD and LLD-related MCI. Methods: We performed voxel-based morphometry and functional connectivity (FC) analyses in elderly patients with both LLD and MCI to compare alterations between those with cerebral amyloidopathy and those without. We subdivided patients into subthreshold depression (STD) and major depressive disorder (MDD) groups. Using florbetaben positron emission tomography (PET), we compared volume and connectivity between healthy controls and four STD and MDD groups with or without amyloid deposition(A): STD-MCI-A(+), MDD-MCI-A(+), STD-MCI-A(-), and MDD-MCI-A(-). Results: Subjects with MDD or amyloid deposition showed greater volume reduction in the left middle temporal gyrus. MDD groups had lower FC than STD groups in the frontal, cortical, and limbic areas. The STD-MCI-A(+) group showed greater FC reduction than the MDD-MCI-A(-) and STD-MCI-A(-) groups, particularly in the hippocampus, parahippocampus, and frontal and temporal cortices. The functional differences associated with amyloid plaques were more evident in the STD group than in the MDD group. Limitations: Limitations include disproportional sex ratios, inability to determine the longitudinal effects of amyloidopathy in large populations. Conclusions: Regional gray matter loss and alterations in brain networks may reflect impairments caused by amyloid deposition and depression. Such changes may facilitate the detection of prodromal AD in elderly patients with both depression and cognitive dysfunction, allowing earlier intervention and more appropriate treatment.

In neurotypical individuals, arm choice in reaching movements depends on expected biomechanical effort, expected success, and a handedness bias. Following a stroke, does arm choice change to account for the decreased motor performance, or does it follow a preinjury habitual preference pattern? Participants with mild-to-moderate chronic stroke who were right-handed before stroke performed reaching movements in both spontaneous and forced-choice blocks, under no-time, medium-time, and fast-time constraint conditions designed to modulate reaching success. Mixed-effects logistic regression models of arm choice revealed that expected effort predicted choices. However, expected success only strongly predicted choice in left-hemiparetic individuals. In addition, reaction times decreased in left-hemiparetic individuals between the no-time and the fast-time constraint conditions but showed no changes in right-hemiparetic individuals. Finally, arm choice in the no-time constraint condition correlated with a clinical measure of spontaneous arm use for right-, but not for left-hemiparetic individuals. Our results are consistent with the view that right-hemiparetic individuals show a habitual pattern of arm choice for reaching movements relatively independent of failures. In contrast, left-hemiparetic individuals appear to choose their paretic left arm more optimally: that is, if a movement with the paretic arm is predicted to be not successful in the upcoming movement, the nonparetic right arm is chosen instead.NEW & NOTEWORTHY Although we are seldom aware of it, we constantly make decisions to use one arm or the other in daily activities. Here, we studied whether these decisions change following stroke. Our results show that effort, success, and side of lesion determine arm choice in a reaching task: whereas left-paretic individuals modified their arm choice in response to failures in reaching the target, right-paretic individuals showed a pattern of choice independent of failures.

Previous studies have reported varying findings regarding the association of brain connectivity in autism spectrum disorder (ASD) with overconnectivity, underconnectivity, or both. Despite the emerging understanding that ASD is a developmental disconnection syndrome, very little is known about structural brain networks in preschool-aged children with low-functioning ASD. We aimed to investigate the structural brain connectivity of low-functioning ASD using diffusion magnetic resonance imaging and graph theory to examine alterations in different brain network topologies and identify any correlations with the clinical severity of ASD in preschool-aged children. Fifty-two preschool-aged children (28 with ASD and 24 with typical development) were included in the analysis. Graph-based network analysis was performed to examine the global and local structural brain networks. Nodal network measures exhibited increased nodal strength in the right Heschl's gyrus, which was positively associated with all autistic clinical symptoms (Autism Diagnostic Observation Schedule and Childhood Autism Rating Scale [CARS]). The nodal strength of the right inferior temporal gyrus showed a moderate correlation with the CARS score. Using network-based statistics, we identified a subnetwork with increased connections encompassing the right Heschl's gyrus and the right inferior temporal gyrus in preschool-aged children with ASD. The asymmetric value in the inferior temporal gyrus exhibited right dominance of nodal strength in children with ASD compared to that in typically developing children. Our findings support the theory of aberrant brain growth and overconnectivity as the underlying mechanism of ASD and provides new insights into potential regional biomarkers that can detect low-functioning ASD in preschool-aged children. LAY SUMMARY: This study supports the theory of aberrant brain growth and overconnectivity as an explanation for ASD. Measuring the right HG and inferior temporal gyrus provides new insights of potential regional biomarkers underpinning ASD in preschool-aged children.