Peer-reviewed article
Subject-independent classification of ADHD from pediatric EEG: a leakage-aware deep learning framework with interpretable neurophysiological mapping
M. Geethapriya, Kalyanaraman Pattabiraman ·
Summary
Written by AI, not yet AICS's view · claude-sonnet-5-5This study looks at how well computer models can tell children with ADHD from other children using EEG (brain-wave) recordings. It shows that many earlier high accuracy claims may be inflated, because data from the same child can appear in both the training and testing sets. The authors tested six models on 121 children and 4,173 EEG epochs, using two validation methods. When the test children were completely unseen, accuracy dropped sharply for models using hand-built spectral features, but much less for a convolutional neural network (CNN) that learned from raw EEG. The authors recommend strict subject-wise validation in EEG research.
Key points
- With segment-wise validation, the SVM using power spectral density features reached 90.0% accuracy (AUC 0.96).
- With subject-wise validation, the same SVM fell to 65.6% accuracy (AUC 0.72), a 24.4 percentage point drop (p < 0.001).
- The stand-alone CNN lost only 2.5 percentage points, reaching 72.7% accuracy (AUC 0.72, recall 88.0%) on unseen children.
- SHAP analysis gave the highest relative contributions to frontal (F3, Fz, F4) and central (Cz) electrodes, which the authors say fits established ADHD neurophysiology.
- The authors recommend strict subject-wise validation as standard practice for translational EEG research.
Limitations
The abstract describes a single dataset of 121 children, and even the best subject-independent accuracy (about 73%) is modest and not a validated clinical tool. The SHAP results show plausibility only and do not establish causes. The abstract does not state where the children were recruited.
Abstract
Background The use of electroencephalography (EEG) in automated screening of Attention-Deficit/Hyperactivity Disorder (ADHD) has created a large body of literature that suggests promising outcomes, with reported classification accuracies often above 90%. A major methodological issue suggests that a substantial part of these results are actually due to data leakage, whereby EEG epochs pertaining to the same subject end up in both training and testing sets, allowing classifiers to capitalize on features specific to the individual rather than on genuine disorder-related biomarkers. Methods We assessed the performance of three classical machine learning algorithms, namely, Logistic Regression (LR), Random Forest (RF), and Support Vector Machine (SVM), as well as three deep learning models: Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and CNN-BiLSTM, on a standard pediatric EEG dataset containing 121 children and 4,173 epochs. The performance of each model was compared in terms of both segment-wise cross-validation (Protocol A) and subject-wise cross-validation (without any data leakage, Protocol B).To evaluate neurotopographic plausibility without making unsupported causal assertions, we applied Deep SHapley Additive exPlanations (SHAP) to the best-generalizing architecture. Results The performance discrepancy was substantial. Under segment-wise evaluation, the SVM operating on power spectral density (PSD) features reached 90.0% accuracy (95% CI: [89.1%, 90.9%], AUC = 0.96). However, the same model dropped to 65.6% (95% CI: [62.1%, 69.1%], AUC = 0.72) under subject-wise evaluation—a statistically significant 24.4 percentage point collapse ( p < 0.001, Wilcoxon signed-rank). In contrast, deep architectures learning directly from raw EEG were resilient: the stand-alone CNN lost only 2.5 percentage points, achieving 72.7% accuracy (95% CI: [66.0%, 79.4%], AUC = 0.72, recall = 88.0%) on completely unseen participants. Post-hoc SHAP attribution identified the highest relative contributions in frontal (F3, Fz, F4) and central (Cz) electrodes, aligning with established ADHD neurophysiology. Conclusion Hand-engineered spectral features are highly vulnerable to identity leakage, whereas end-to-end convolutional models generalize across subjects with realistic, uninflated performance. Strict subject-wise validation is strongly recommended as standard practice for translational EEG research.