Search Results (171)

Autism Spectrum Disorder is a long-term neurodevelopmental disorder. Early diagnosis is crucial for timely rehabilitation and intervention. Recently, machine learning and deep learning techniques have been widely explored and have produced encouraging results using eye-tracking scanpath images for the early detection of ASD. However, these approaches exhibit inconsistent performance and classification error rates, as well as limited generalization, due to differences in learning approaches and architectural designs across individual models. To address these problems, we employed a weighted-average ensemble of deep learning models using eye-tracking scanpath images. In this work, two different pretrained convolutional neural networks were selected, including Xception and VGG16, based on their proven efficacy and performance. Moreover, we fine-tuned each model individually and evaluated them on a dataset containing eye-tracking scanpath images. We implemented a weighted-average ensemble technique to combine the diverse predictions of the two models. It reduces classification errors and improves the model’s generalization and overall performance. The adopted weighted-average ensemble technique achieved an accuracy of 98.18%, with a perfect recall, and a competitive Area Under the Curve (AUC) of 99.59%. These findings highlight that applying a weighted average to integrate multiple models’ predictions strengthens the generalization and reliability of ASD detection. Full article

While the cerebellar dopaminergic system is suggested to be implicated in neurodevelopmental disorders, especially autism spectrum disorder (ASD), the details of its disturbances remain unclear. We performed a comparative analysis of human (GTEx) and mouse (GSE144046, GSE144277) transcriptomes, complemented by RT-qPCR in DAT-KO rats, to identify dopaminergic gene associations in the normal cerebellum and neurodevelopmental disorder models. Pairwise dopaminergic gene correlations were generally weak, with a slight increase in interaction complexity in ASD models. However, weighted gene co-expression network analysis identified a robust gene module involving Comt, which was consistently associated with synaptic translation across mouse datasets. These associations reflect regulatory processes in the whole cerebellum, which is commonly represented in rodent studies but absent in human data, which are acquired in studies of cerebellar subregions. ASD modeling exerted contrasting effects: Cul3 haploinsufficiency increased the number of genes involved in the module with a decrease in connectivity, while Mbd5 haploinsufficiency led to module collapse. These findings confirm neurodevelopmental disorders as a heterogeneous condition where divergent backgrounds uniquely rewire cerebellar dopaminergic networks. Considering the cerebellum’s role in ASD and that some ASD medications target the dopamine system, further investigation of these identified trends may support the development of more personalized therapeutic approaches. Full article

Background: Autism spectrum disorder (ASD) is associated with immune dysregulation in a subset of individuals, though findings remain heterogeneous and poorly defined, particularly regarding immune subtypes and metabolic context. Methods: We analyzed whole-blood microarray data from GSE18123 (GPL570: ASD n = 46, controls n = 19; GPL6244: ASD n = 68, controls n = 21) using an integrated immunometabolic framework. CD4⁺ T-cell transcriptional programs were used to assign dominant immune phenotypes (TH1, TH2, TH17, Tfh, FOXP3⁺ Treg, Tr1-like). Metabolic demand was quantified via the τ-axis; execution capacity was assessed using cytosolic and mitochondrial energy compensation ratios (CECR, MECR). Induction–execution mismatch was captured by three Gap metrics (Cytosolic, Warburg, Global). Functional validation correlated these metrics with transcriptional signatures of folate transport, one-carbon metabolism, receptor-mediated micronutrient uptake (LRP2–CUBN–AMN), cobalamin processing, and vitamin D activation across both platforms. Results: Six immunometabolic CD4⁺ subtypes were identified within ASD. τ-axis discrimination was strongest for Tr1-like (AUC = 0.811) and Tfh (AUC = 0.825) states, while TH17 profiles were indistinguishable from controls. Despite variation in metabolic demand, CECR and MECR remained relatively preserved, indicating decoupling between induction and execution capacity. Global Gap values were most negative in Tfh and TH1 states and positive in TH17 and controls. Negative Gap states showed coordinated suppression of ATP-intensive micronutrient acquisition pathways, including folate transport (FOLR1/2, SLC19A1), megalin–cubilin-mediated uptake (r ≈ 0.77–0.79), and vitamin D activation (CYP27B1). Intracellular cobalamin processing was upregulated in proportion to metabolic demand (r > 0.9). Findings were directionally replicated across both datasets. Conclusions: These data demonstrate that ASD exhibits structured immunometabolic heterogeneity characterized by subtype-specific demand–capacity imbalance. The Global Gap framework provides transcriptomic evidence of energetic deficit in Tfh- and Tr1-like-dominant states. Future clinical studies should incorporate subtype-stratified assessment of micronutrient status and metabolic execution capacity. Full article

Introduction: Medical care for transgender minors is understudied, largely because these forms of care are relatively recent. The primary objective of this work was to describe the cohort of transgender adolescents who initiated follow-up at the Strasbourg University Hospital before the age of 18, whether or not they began hormone therapy prior to reaching adulthood. Method: This was an observational, retrospective, single-center, descriptive study conducted among adolescents who had attended at least one consultation in our center before the age of 18 between January 2017 and March 2024. Results: Our population consisted of 115 patients predominantly made up of transmasculine (AFAB) adolescents (68%). Compared with the general population, we observed significantly higher rates of psychiatric co-occurrences, autism spectrum disorder (ASD), and attention-deficit/hyperactivity disorder (ADHD). Only 46.1% initiated gender-affirming hormone therapy (GAHT) in our cohort, and just 34.8% before age 18. A total of 6% of adolescents received puberty blockers as monotherapy. The mean age at GAHT initiation was 16.99 years. Transition pathways appear to differ according to the adolescent’s type of schooling. The rate of retransition/treatment interruption in our sample ranged from 0% to 6.1%, depending on the criteria applied. We did not identify any adolescent who retransitioned to their sex assigned at birth after starting GAHT by the end of the data collection. Discussion: The high prevalence of psychiatric co-occurrences raises important questions regarding how to improve care for these adolescents. The predominance of AFAB adolescents similarly prompts reflection on the barriers that transfeminine adolescents may face when seeking to transition before adulthood. In addition, the substantial number of adolescents presenting with ASD or ADHD underscores the need for particular vigilance regarding their specific needs and overall well-being. Finally, the variability in retransition rates depending on the criteria applied highlights the absence of a consensual definition, which limits the comparability and validity of existing studies. Conclusions: Long-term prospective studies are needed to objectively demonstrate the effectiveness of current transition pathways. Academic research in this field should be strengthened, along with the development of larger prospective datasets, to improve the overall health of this population. Full article

Conventional CNN architectures often struggle with information loss during feature extraction, particularly in pooling and flattening layers, where spatial coherence and high-frequency details critical for tasks such as medical diagnostics are compromised. To address this, we introduce a novel integration of Hilbert curve flattening and multiscale frequency-selective wavelet pooling, which preserves diagnostically relevant features while optimizing computational efficiency. Multifrequency selective wavelet pooling improves the performance and adaptability of convolutional neural networks by preserving spatial adjacency structures and eliminating duplicate information. Here, raster flattening was replaced with a conventional Hilbert curve that organized data more efficiently, and wavelet pooling performed feature selection across frequency bands better than average pooling or max-pooling. On standard architectures (Inception, VGG16, ResNet, EfficientNet), our approach consistently produced an improved precision of 1.42% over earlier methods across all datasets and classes, including diagnosis of autism via structural MRI in a proof-of-concept dataset (38 subjects, 4 in the test set), with high precision, at 99%. Hence, validation on larger independent cohorts will be part of the future work. The synergy of Hilbert curve flattening and multiscale frequency-selective wavelet pooling mitigates signal decomposition losses and maintains spatial frequency relationships, advancing CNNs for high-stakes applications like medical imaging and remote sensing. These new strategies enhance spatial coherence and global efficiency, ensuring robustness in applications ranging from medical imaging to time-series forecasting. Full article

Pediatric neurodevelopmental disorders (NDDs) encompass a highly heterogeneous group of conditions characterized by complex interactions among genetic, molecular, developmental, and environmental factors. Growing evidence increasingly supports an important role for neuroglial dysfunction, including disturbances in astrocytic, microglial, and oligodendroglial biology, in the pathophysiology of disorders such as autism spectrum disorder, global developmental delay, intellectual disability, and rare neurogenetic syndromes. At the same time, artificial intelligence (AI)-assisted analytical approaches are becoming increasingly relevant in pediatric diagnostics through integration of multidimensional datasets, including clinical phenotypes, neuroimaging, genomic sequencing, and molecular biomarkers. This review examines the evolving intersection of neuroglial biology and AI-based analytical methods in pediatric NDDs. Current understanding of neuroglial mechanisms underlying disease vulnerability and developmental heterogeneity is discussed alongside emerging applications of machine learning, deep phenotyping platforms, radiogenomics, and large language models in diagnostic interpretation and clinical decision support. Important translational and ethical challenges, including algorithmic bias, interpretability limitations, data governance, and disparities in data accessibility, are also considered. Overall, integration of neuroglial research with AI-assisted analytical frameworks may contribute to more biologically informed interpretation of pediatric neurodevelopmental disorders and support ongoing development of increasingly individualized diagnostic approaches. Full article

Background: Autism spectrum disorder (ASD) arises from complex gene–environment interactions. While trace element abnormalities have been studied, associations with autism severity remain inconsistent. Ratios indicating detoxification balance, rather than single toxic elements, may better reflect severity. Objective: To examine the relationships between urinary trace element levels, detoxification-related element ratios, and autism severity measured by the Childhood Autism Rating Scale (CARS). Methods: In a cross-sectional study of 168 participants (103 ASD, 65 controls), thirty urinary trace elements were quantified by ICP-MS. ASD patients were stratified by CARS into subthreshold ASD (n = 29), mild–moderate ASD (n = 36), and severe ASD (n = 38). Analyses included Mann–Whitney U, Kruskal–Wallis, and Spearman correlation tests, focusing on Li/Pb, Cu/Pb, and Cr/Pb ratios. Results: Individual elements showed weak associations with CARS; lead correlated positively (ρ = 0.209, p = 0.035) and lithium inversely (ρ = −0.194, p = 0.051). In contrast, element ratios showed stronger links: Li/Pb (ρ = −0.349, p = 0.0003), Cu/Pb (ρ = −0.320, p = 0.0011), and Cr/Pb (ρ = −0.209, p = 0.035). Severe ASD exhibited modest 90th-percentile elevations for toxic elements but high heterogeneity. Conclusions: Single-element levels showed limited associations with ASD severity. Element ratios, particularly Li/Pb, showed stronger statistical associations than individual elements in this cross-sectional dataset; however, these findings should be interpreted as candidate correlates rather than causal or clinically validated biomarkers. Full article

Maternal folate receptor alpha autoantibodies (FRAA) have been associated with impaired placental folate transport, fetal cerebral folate deficiency (CFD) and neurodevelopmental risks including autism spectrum disorder (ASD); however, first-trimester-specific reference intervals remain undefined. This prospective single-center study of 534 healthy pregnant women at 10 + 0 to 15 + 6 weeks’ gestation (week 10 n = 26; week 11 n = 155; week 12 n = 203; week 13 n = 105; week 14 n = 39; week 15 n = 6) used a CE-IVDR FRAA ELISA, following CLSI EP28-A3c and IFCC C-RIDL protocols, to establish week-specific percentiles (P5, P50, P95, P99) via non-parametric estimation and log-smoothed regression with a 1-week rolling window. An internal-consistency ROC analysis was performed against the within-dataset ≥P99 designation and is therefore not interpretable as discrimination against an independent clinical outcome. Median FRAA declined from 29 ng/mL (week 10) to 25 ng/mL (week 14), with provisional clinically actionable thresholds of P95 ≈ 120 ng/mL and P99 ≈ 150 ng/mL. These data provide the first first-trimester normative percentile curves for maternal FRAA and may support prioritizing FRAA assessment before 15 weeks (onset of accelerated transplacental IgG transfer). Given the cross-sectional, single-center design, the small samples at weeks 14–15, and the absence of long-term neurodevelopmental outcome data, the proposed thresholds and any downstream clinical implications, including folinic acid intervention and ASD risk mitigation, should be considered hypothesis-generating and require external and longitudinal validation. Full article

Background/Objectives: Autism Spectrum Disorder (ASD) is a set of neurological and neurodevelopmental disorders characterized by difficulties in social communication and interaction, repetitive behaviors, and sensory processing differences. Recent studies have shown that circRNAs play a crucial role in the pathophysiology of ASD. In this study, we present an exploratory machine learning framework integrating circRNA sequence features, miRNA interactions, gene targets, and pathway enrichment analysis to investigate ASD-associated molecular signatures. Methods: Differential circRNAs were identified from human peripheral blood datasets, and informative features were selected using attribute-based filtering and Information Gain ranking. Machine learning models were developed using the WEKA platform. Results: The HyperPipes classifier achieved the highest performance (92.5% accuracy under cross-validation). Analysis using an independent ASD gene expression dataset showed consistent discriminative patterns of the derived gene-level signatures across multiple machine learning classifiers. The competitive endogenous RNA network and enriched gene pathways were also analyzed. Conclusions: Overall, this study provides a computational, preliminary framework for analyzing circRNA-associated molecular patterns in ASD. Findings should be interpreted in the context of limited sample size and dataset availability. Full article

Background: Head motion and variability in scan quality remain major methodological challenges in autism neuroimaging. Large multi-site datasets such as ABIDE-II provide a unique opportunity to systematically quantify diagnostic differences in MRI data quality and assess the influence of site-level heterogeneity. Methods: Functional MRI Quality Assessment Protocol (QAP) metrics were combined with phenotypic data from ABIDE-II. Participants were classified as autistic (ASD) or typically developing (TD). Key quality metrics—including mean framewise displacement (mFD), proportion of volumes exceeding 0.20 mm (FD > 0.20), signal-to-noise ratio (SNR), and entropy focus criterion (EFC)—were analyzed alongside age, sex, IQ, and site. Group differences were evaluated using non-parametric tests and linear mixed-effects models with site as a random factor. Additional analyses examined site-level heterogeneity and the impact of quality-control (QC) thresholds on sample composition. Results: The final sample included 1277 participants (579 ASD; 698 TD) across 14 sites. ASD participants exhibited significantly greater head motion (median mFD = 0.101 vs. 0.081 mm; p < 1 × 10⁻¹⁰) and modest reductions in signal quality (lower SNR, higher EFC). Elevated motion in ASD was observed in 12 of 14 sites, although effect sizes varied substantially. Mixed-effects models confirmed that diagnosis remained a significant predictor of motion after adjusting for covariates. In contrast, signal-quality differences were small and largely explained by site effects. Simulated QC procedures disproportionately excluded ASD participants, with exclusion rates up to 31% compared to 18% in TD. Conclusions: ASD participants show consistently higher head motion, while signal-quality differences are minimal and largely site-driven. Standard QC procedures disproportionately exclude ASD individuals, highlighting the need for improved motion handling and more balanced quality-control strategies in multi-site studies. Full article

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder caused by the interaction between genetic and environmental influences, potentially mediated by epigenetic mechanisms such as DNA methylation. Genome-wide DNA methylation profiling was performed using the Infinium MethylationEPIC v2.0 array on peripheral blood mononuclear cells (PBMCs) from 100 children with ASD and 50 typically developing controls. Differential methylation analyses were conducted by adjusting for age, sex, and estimated blood-cell-type composition as covariates. Functional enrichment, SFARI gene-overlap analysis, and cross-cohort validation were performed. We identified 3507 differentially methylated positions (DMPs) in the ASD cohort. Functional enrichment revealed pathways involved in neuronal signaling, synaptic activity, and immune regulation, suggesting coordinated neurodevelopmental and immune processes in ASD. Stratification by clinical severity demonstrated common and unique biological characteristics between the moderate and severe ASD groups. Furthermore, DMP-associated genes significantly overlapped with high-confidence ASD risk genes from the SFARI database and established transcriptomic signatures of neurodevelopmental disorders. Comparisons with independent post mortem brain tissue and peripheral blood datasets revealed partial overlap and directional concordance. However, the strength of concordance varied across datasets and was limited in the most directly comparable peripheral blood cohort. Our findings suggested that DNA methylation profiling of PBMCs provided peripheral epigenetic signatures and candidate loci for further validation in larger independent cohorts. Full article

Background: Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition marked by heterogeneous behavioral symptoms and systemic comorbidities, including immune and gastrointestinal dysfunctions. Emerging studies suggest that glycosylation—a fundamental post-translational modification regulating cellular communication and immune responses—may play a role in ASD pathophysiology, yet its contribution remains underexplored. Methods: In this study, we developed an integrative transcriptomic and network analysis framework to investigate glycosylation-related gene expression changes and their functional associations in ASD. Using publicly available datasets from bulk and single-cell RNA sequencing of brain and blood tissues, we focused on four prior-knowledge gene subsets: glycogenes, extracellular matrix glycoproteins, immune response genes, and autism risk genes. Results: Differential expression and pathway enrichment analyses revealed consistent dysregulation of glycosylation pathways, including mucin-type O-glycan biosynthesis, glycosaminoglycan metabolism, GPI-anchor formation, and sialylation, across ASD tissues. These transcriptional changes were functionally linked to altered immune signaling (e.g., IL-17, Toll-like receptor, and complement pathways) and synaptic development pathways, forming a distinct glyco-immune axis. Network analysis identified key glycogenes such as GALNT10, NEU1, LMAN2L, and CHST1 as central molecular nodes, interacting with immune and neuronal regulators. Linkage disequilibrium analysis further revealed ASD-associated SNPs influencing the expression of these glycogenes in both blood and brain tissues. Conclusions: Together, these findings support a model in which disrupted glycosylation contributes to ASD pathophysiology by mediating immune dysregulation and altered neuronal connectivity. This study offers a systems-level framework to understand the molecular complexity of ASD and highlights glycogenes as potential biomarkers and targets for future therapeutic exploration. Full article

This Data Descriptor presents an anonymized, shuffled dataset of creatinine-normalized urinary metabolite measurements from 73 Bulgarian children with autism spectrum disorder (ASD), released to support reuse in secondary analyses and cross-cohort comparisons. The public release represents a pathway-oriented 24-marker subset from a broader urinary diagnostic panel, assembled as a self-contained resource for investigators working in these metabolic domains. Spot urine results are provided as individual-level values after creatinine normalization; for trimethylamine, values below the limit of quantification (LOQ) were replaced with LOQ/2. The deposit contains measurements for 24 urinary markers grouped into three functional classes (neurotransmitters and aromatic amino acid precursors; one-carbon/methylation and vitamin-related metabolites; and energy metabolism/organic acids with microbiome-related amines). The underlying cohort comprised children aged 3–13 years, and no contemporaneous neurotypical control group was enrolled. Second-morning, midstream, acid-stabilized spot urine samples were collected within the provider’s workflow; metabolites were measured by LC–MS/MS, and spot urinary creatinine was measured enzymatically for normalization. The release includes the results table in both XLSX and CSV formats, a reference limits and units file for contextual interpretation, a data dictionary, a README, a changelog, and SHA-256 checksums for integrity verification. The public files contain de-identified analytical variables only and omit individual-level demographics, dates, standalone urinary creatinine, and richer clinical metadata to preserve anonymity. Full article

Background: The recognition of autism spectrum disorder (ASD) has been a challenge due to the heterogeneity in symptoms and complex variations in brain function. Resting-state functional magnetic resonance imaging (rs-fMRI) has become instrumental in studying these disorders by accessing underlying abnormal neural activity and connectivity. Recently, deep learning approaches have shifted the analysis of brain networks by capturing spatiotemporal information from fMRI sequences. Nonetheless, most existing studies are limited by relying on a single representational scale, typically restricting analysis to either voxel-level spatiotemporal patterns or static connectivity matrices. Additionally, the dynamic reconfiguration of functional coupling and its variations across different anatomical parcellations are often ignored, which obscures neurobiologically meaningful dynamics. Methods: In this regard, we propose a multi-atlas dynamic connectivity transformer fused with 4D spatiotemporal modeling for ASD recognition (MADCT-4D). Specifically, the framework comprises two complementary branches. The 4D spatiotemporal branch encodes raw rs-fMRI volumes to learn hierarchical representations of evolving neural activity, while the dynamic-connectivity branch models time-resolved functional connectivity sequences constructed from multiple atlases, enabling the network to capture dynamic reconfiguration at the connectome level under different parcellation granularities. Moreover, we perform late fusion by combining the branch-specific decision scores with a learnable gate, allowing the model to adaptively weight voxel-level dynamics and multi-atlas connectivity evidence for each subject. Results: Extensive experiments on the publicly available ABIDE dataset demonstrate that the proposed method achieves 90.2% accuracy for ASD recognition, outperforming multiple competitive baselines. Conclusions: The proposed framework yields interpretable biomarkers based on learned dynamic connectivity patterns that are consistent with altered functional coupling in ASD. Full article

Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental disorder that requires early diagnosis for better intervention. However, current clinical behavioural examinations are time-consuming and prone to human error. Objective and effective biomarkers are essential for the diagnosis and prognosis of the disorder. Electroencephalography (EEG) is a non-invasive and inexpensive brain-imaging technique that is widely applied in the diagnosis of ASD. Feature-based methods have shown better performance in EEG-based applications. Here, we present a prediction framework based on massive feature extraction using the highly comparative time-series analysis (HCTSA) method and a hybrid feature selection method for the classification of ASD from resting-state EEG. Machine-learning models are trained and tested on a different number of selected features. Our models demonstrated 100% accuracy with ≥50 features on a balanced dataset of 56 participants. The most discriminating EEG channels and features were used in the prediction process, as well as those using Shapley values to provide explainability of our framework. Whilst these results are promising, we acknowledge the limitations of a single small-scale dataset and emphasise the need for validation on larger independent cohorts before clinical translation. Full article