Aug 2026· Multidimensional systems and signal processing· Vol 37· 0 citations· 37 references
Computer Science
TL;DR
It is shown that subject-level validation, preprocessing, and architectural decisions materially influence EEG-based ADHD classification, and EEG-MSCNet contributes both a model and a systematic evaluation framework.
Abstract
Attention-Deficit/Hyperactivity Disorder (ADHD) is a common neurodevelopmental condition in children. Electroencephalography (EEG) may support objective assessment, but subject-independent ADHD classification remains challenging because EEG varies across participants and recording segments. We developed the EEG Multiscale Convolutional Network (EEG-MSCNet), a multiscale convolutional architecture for pediatric EEG classification, and evaluated it using subject-level leave-one-subject-out (LOSO) validation on a public EEG dataset. The study also examined preprocessing choices, architectural ablations, channel-region subsets, computational complexity, and repeated-run stability. The final EEG-MSCNet evaluation achieved an Accuracy of 87.6%, F1-score of 87.6%, and area under the receiver operating characteristic curve (AUROC) of 88.2%. Artifact subspace reconstruction (ASR) combined with Fully Automated Statistical Thresholding for EEG artifact Rejection (FASTER) emerged as the strongest preprocessing configuration. The multiscale architecture showed stronger observed Accuracy and F1-score than reduced architectural variants, and the full montage provided the strongest channel configuration. These findings show that subject-level validation, preprocessing, and architectural decisions materially influence EEG-based ADHD classification. EEG-MSCNet contributes both a model and a systematic evaluation framework, supported by source code, configurations, logs, and the public dataset reference for reproducible comparison.
Attention-deficit/hyperactivity disorder (ADHD) affects millions globally, yet current diagnostic approaches rely on subjective behavioral assessments without objective neurophysiological markers. While machine learning on electroencephalogram (EEG) data shows promise for automated ADHD risk screening, current methods focus only on brain signals and ignore socioeconomic factors that strongly affect neurodevelopment and ADHD risk. We introduce a novel multimodal deep learning architecture integrating three complementary streams: temporal EEG dynamics via one-dimensional convolutional-recurrent networks, spectro-temporal patterns via two-dimensional convolutional networks with spatial and channel attention, and socioeconomic context via feedforward processing, combined through an attention-based fusion mechanism. Using the Cognitive Electrophysiology in Socioeconomic Context dataset, we evaluate performance across four cognitive tasks with 5-fold stratified cross-validation, ablation studies and benchmarking against a state-of-the-art EEG classification model. Under epoch-level cross-validation, the multimodal approach outperforms EEG-only baselines across all four tasks, achieving accuracy improvements of 2.1–5.9% and sensitivity gains up to 12.2%, with strong positive-class F1-scores (96.4–99.8%). Results showed higher epoch-level performance when socioeconomic context was incorporated alongside neurophysiological signals, a pattern that held across diverse cognitive paradigms. Leave-One-Subject-Out Cross-Validation across all four tasks yielded accuracy of 0.86–0.91 for the EEG-only model and 0.92–0.96 for the multimodal model, with sensitivity of 0.71–0.96 and specificity of 0.95–1.00 for the multimodal model. These subject-independent estimates are more modest than the epoch-level figures and McNemar’s test on paired predictions did not reach significance on any task. The EEG backbone, evaluated without modification on an independent paediatric dataset, also achieved 80.4% subject-independent accuracy, outperforming the prior benchmark. Labels derive from a validated self-report screening instrument rather than clinical diagnosis; this model should be understood as a proof-of-concept for ADHD risk screening, not a diagnostic tool. This work suggests the feasibility of context-aware ADHD risk screening that accounts for environmental influences on neurodevelopment alongside neurophysiological signals.
A comparative temporal deep learning framework for EEG-based ASD classification by evaluating principal component analysis (PCA) and multiscale principal component analysis (MS-PCA) as feature representations combined with a recurrent neural network with bidirectional long short-term memory (RNN-BiLSTM) and a temporal convolutional network with self-attention (TCN-SA).
M. Muliyadi, Melinda, Yuwaldi Away et al.· Journal of Electronics Elect...· 0 citations
The proposed electroencephalography-based ASD detection method was evaluated against previous approaches using the same datasets, showing a substantial improvement in performance, and the need to evaluate the proposed approach on larger datasets to further ensure model generalization.
B. S. Falih, Mohannad K. Sabir, A. Aydın· Brain Topography· 0 citations
Electroencephalography (EEG) provides a non-invasive means of supporting Attention-Deficit/Hyperactivity Disorder (ADHD) assessment. Nonetheless, existing pipelines often rely on handcrafted descriptors, segment-wise decisions, or deep architectures with limited subject-level generalization. This work introduces Hidden Markov Model-Induced Stationary RKHS Distance Learning (HIS), a probabilistic framework that represents each subject by a Hidden Markov Model with Gaussian-mixture emissions trained directly from frontal EEG recordings. Rather than vectorizing model parameters, each HMM is mapped to its induced stationary observation distribution and embedded into a Reproducing Kernel Hilbert Space (RKHS), where pairwise subject similarities are computed through a closed-form Hilbert embedding distance. These similarities are subsequently exploited by precomputed-kernel classifiers for subject-level prediction. The proposed method was evaluated against the Probability Product Kernel baseline using both a controlled synthetic EEG benchmark and a public pediatric ADHD dataset under progressively more rigorous validation protocols, culminating in repeated nested cross-validation with bootstrap confidence intervals and permutation testing. On the synthetic benchmark, HIS achieved 95.0% held-out accuracy and consistently outperformed the baseline across classifiers. On a real EEG dataset with 121 subjects, the primary evaluation protocol yielded a balanced accuracy of 73.5% (95% CI: 69.8–77.0%), an AUC of 79.6%, and an MCC of 0.483 (permutation p < 0.001) using an SVM with compact subject-specific HMMs. Complementary hyperparameter analyses and t-SNE visualizations demonstrated that HIS induces more stable and discriminative subject representations than the baseline. These results establish stationary RKHS embeddings of subject-specific HMMs as a leakage-aware framework for EEG-based ADHD decision support and underscore the critical influence of statistically rigorous evaluation protocols on reported classification performance.
Leonardo Lopez-Ortiz, Cristhian K. Valencia-Marin, J. Gil-González et al.· Italian National Conference...· 0 citations
A Deep Hybrid Neural Network framework that combines Convolutional Neural Networks (CNN) with the Aquila Optimizer (AO) for the automatic detection of epileptic seizures utilizing EEG data in MATLAB is introduced.
Swati Chowdhuri, Tiyasha Mondal· International Journal of Eng...· 0 citations
Although clinical diagnosis of childhood Attention-Deficit/Hyperactivity Disorder (ADHD) relies heavily on subjective behavioral assessments, electroencephalography (EEG) offers a low-cost, noninvasive objective alternative. This study proposes an approach employing 1D texture analysis of EEG signals for ADHD classification. Specifically, 1D Local Binary Pattern (1D-LBP), 1D Local Gradient Pattern (1D-LGP), and 1D Local Neighbor Descriptive Pattern (1D-LNDP) capture local amplitude relationships across 19 electrodes in 2-second frames, yielding a combined 570-dimensional feature space based on binary transitions. To optimize this, the minimum Redundancy Maximum Relevance (mRMR) method is used to select the 190 most discriminative features (Hybrid-190). System performance was then evaluated using various classifiers, with hyperparameters tuned via Bayesian optimization within a 5-fold cross-validation scheme. Tested on a benchmark dataset of 61 children with ADHD and 60 healthy controls, the framework achieved a peak segment-level accuracy of 99.63%, a specificity of 99.65%, an F1-score of 99.67%, and an Area Under the Curve (AUC) of 0.9987 using a k-Nearest Neighbor (KNN) classifier. Under a subject-independent protocol with per-participant aggregation, it attained a participant-level accuracy of 78.50% and an AUC of 0.8600 with a Random Forest classifier. A topographic accuracy peak was also observed in the right temporal-occipital region near the T6 electrode. Among individual descriptors, 1D-LGP achieved the highest standalone segment-level accuracy at 99.26%. Ultimately, this framework provides an interpretable and computationally efficient EEG-based approach whose subject-independent performance supports its potential for clinical decision-support systems.
Ferdi Ozbilgin· IEEE Access· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
