Dynamic Spatio-Temporal Graph Neural Network for Early Detection of Pornography Addiction in Adolescents Based on Electroencephalogram Signals

Achmad Ardani Prasha; Clavino Ourizqi Rachmadi; Sabrina Laila Mutiara; Hilman Syachr Ramadhan; Chareyl Reinalyta Borneo; Saruni Dwiasnati

cs.LGFeb 28, 2026

Dynamic Spatio-Temporal Graph Neural Network for Early Detection of Pornography Addiction in Adolescents Based on Electroencephalogram Signals

Authors: Achmad Ardani Prasha, Clavino Ourizqi Rachmadi, Sabrina Laila Mutiara, Hilman Syachr Ramadhan, Chareyl Reinalyta Borneo, Saruni Dwiasnati

Organizations: Faculty of Computer Science, Universitas Mercu Buana, West Jakarta, Indonesia · Faculty of Psychology, Universitas Mercu Buana, West Jakarta, Indonesia

Abstract

Adolescent pornography addiction requires early detection based on objective neurobiological biomarkers because self-report is prone to subjective bias due to social stigma. Conventional machine learning has not been able to model dynamic functional connectivity of the brain that fluctuates temporally during addictive stimulus exposure. This study proposes a state-of-the-art Dynamic Spatio-Temporal Graph Neural Network (DST-GNN) that integrates Phase Lag Index (PLI)-based Graph Attention Network (GAT) for spatial modeling and Bidirectional Gated Recurrent Unit (BiGRU) for temporal dynamics. The dataset consists of 14 adolescents (7 addicted, 7 healthy) with 19-channel EEG across 9 experimental conditions. Leave-One-Subject-Out Cross Validation (LOSO-CV) evaluation shows F1-Score of 71.00% $\pm$ 12.10% and recall of 85.71%, a 104% improvement compared to baseline. Ablation study confirms temporal contribution of 21% and PLI graph construction of 57%. Frontal-central regions (Fz, Cz, C3, C4) are identified as dominant biomarkers with Beta contribution of 58.9% and Hjorth of 31.2%, while Cz-T7 connectivity is consistent as a trait-level biomarker for objective screening.

Figures & tables

Figure 2: Dynamic Spatio-Temporal Graph Neural Network (DST-GNN) Architecture

Figure 3: Electrode placement on scalp marked in red color (Source: Kang et al. (2021) )

No.	Task Name	File Name	Duration (s)	Samples $\times$ Channels
1	Eyes Closed	EC.csv	60	15,000 $\times$ 19
2	Eyes Open	EO.csv	60	15,000 $\times$ 19
3	Happy	H.csv	60	15,000 $\times$ 19
4	Calm	C.csv	60	15,000 $\times$ 19
5	Sad	S.csv	60	15,000 $\times$ 19
6	Fear	F.csv	60	15,000 $\times$ 19

Table 1: EEG signal recording details per experimental condition (Source: Kang et al. (2021) )

Figure 4: Dataset folder structure (Source: Kang et al. (2021) )

No.	Subject ID	Gender	Label
1	S1	Male	Addicted
2	S2	Female	Not Addicted
3	S3	Female	Not Addicted
4	S4	Male	Not Addicted
5	S5	Male	Addicted
6	S6	Male	Addicted

Table 2: Participant details (Source: Kang et al. (2021) )

Figure 5: Data distribution at subject level (left) and window level (right)

Figure 6: Comparison of filtered EEG signals between healthy and addicted subjects

Figure 7: Global PSD (Power Spectral Density) comparison between groups

Figure 8: Scalp topomaps band power per frequency: healthy (top), addicted (middle), difference (bottom)

Figure 9: EEG graph structure: connection topology (left), PLI matrix (center), node features (right)

Figure 10: Temporal evolution of brain connectivity on window sequence

Figure 11: Brain connectivity pattern comparison: healthy (left, 66 edges) vs addicted (right, 48 edges)

Model	Accuracy	Precision	Recall	F1-Score	ROC-AUC
Logistic Regression	62.14%	50.00%	29.29%	34.73%	50.00%
SVM (RBF)	58.33%	50.00%	28.57%	34.37%	50.00%
Random Forest	44.52%	42.86%	16.43%	20.63%	50.00%
XGBoost	49.29%	50.00%	20.71%	26.04%	50.00%
MLP	51.90%	50.00%	26.19%	32.46%	50.00%

Table 3: Baseline model evaluation results (LOSO-CV)

Figure 12: Baseline model performance comparison with error bars

Seed	Accuracy	Precision	Recall	F1-Score	ROC-AUC
42	57.14%	54.55%	85.71%	66.67%	59.18%
123	78.57%	77.78%	100.00%	87.50%	89.80%
456	57.14%	50.00%	71.43%	58.82%	44.90%
Mean $\pm$ SD	64.29 $\pm$ 15.43%	60.77 $\pm$ 12.17%	85.71 $\pm$ 11.66%	71.00 $\pm$ 12.10%	64.63 $\pm$ 18.73%

Table 4: DST-GNN multi-seed LOSO-CV evaluation results

Figure 13: Confusion matrix (left) and DST-GNN performance metrics (right)

Figure 14: DST-GNN ROC curves for each random seed

Configuration	Accuracy	F1-Score	F1 Change
Full DST-GNN	64.29%	71.00%	Baseline
Spatial-Only (without BiGRU)	42.86%	50.00%	$-$ 21.00%
Fully-Connected (without PLI)	14.29%	14.29%	$-$ 56.71%

Table 5: DST-GNN ablation study results

Figure 15: Ablation study results comparison between model configurations

Figure 16: DST-GNN channel importance topomap showing frontal-central region dominance

Figure 17: DST-GNN feature importance: Beta and Hjorth Activity as main features

Figure 18: wPLI connectivity differences (Addicted - Healthy) on 9 experimental conditions

Figure 19: Band power analysis per condition category: baseline, emotional, cognitive, executive

Figure 20: Brain region discriminability between tasks (left) and overall ranking (right)

Figure 21: Global connectivity metrics between conditions and task categories

Figure 22: Integrated Gradients (IG): feature importance (left), top channels (center), channel $\times$ feature (right)

Figure 23: Edge importance matrix (left) and top 15 important connections (right)

Figure 24: PLI vs wPLI comparison: scatter plot, correlation distribution, and connectivity per task

Spatiotemporal GNNs

Spatiotemporal graph neural networks model how signals and relationships evolve across connected locations, agents, or physical systems.

See all 147 papers

D-GATNet: Interpretable Temporal Graph Attention Learning for ADHD Identification Using Dynamic Functional Connectivity
Mar 27, 2026Qurat Ul Ain, Alptekin Temizel, Soyiba JawedDynamic Functional Connectivity Classification
Department of Computer and Software Engineering, College of Electrical and Mechanical Engineering, National University of Sciences and Technology, Islamabad 44000, Pakistan · Graduate School of Informatics, Middle East Technical University, 06800, Ankara, Turkey
NeuroBRIDGE: Behavior-Conditioned Koopman Dynamics with Riemannian Alignment for Early Substance Use Initiation Prediction from Longitudinal Functional Connectome
Mar 31, 2026Badhan Mazumder, Sir-Lord Wiafe, Vince D. Calhoun +1Human Behavior Prediction Functional Connectivity
Department of Computer Science, Georgia State University · Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS), Georgia State University, Georgia Institute of Technology, and Emory University
MPP-GNN: Subject-Adaptive Community Detection for fMRI-Based Alzheimer's Disease Classification
Jul 29, 2026Yang Zhang, Xiao Zhou, Jonathan Warrell +3Graph Neural Networks Graph Pooling
Program in Computational Biology and Bioinformatics, Yale University, New Haven, CT 06520 USA · Department of Psychiatry, Robert Wood Johnson Medical School, and the Brain Health Institute, Rutgers University, Piscataway, NJ 08854 USA · School of Design, Pratt Institute, New York, NY 11205 USA

Mar 27, 2026J/K move · Enter open · S save

D-GATNet: Interpretable Temporal Graph Attention Learning for ADHD Identification Using Dynamic Functional Connectivity

Qurat Ul Ain, Alptekin Temizel, Soyiba Jawed

Department of Computer and Software Engineering, College of Electrical and Mechanical Engineering, National University of Sciences and Technology, Islamabad 44000, Pakistan · Graduate School of Informatics, Middle East Technical University, 06800, Ankara, Turkey

Dynamic Functional Connectivity Classification Neuroimaging Graph Attention Networks

Open paper AskPDF arXiv

Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder whose neuroimaging-based diagnosis remains challenging due to complex time-varying disruptions in brain connectivity. Functional MRI (fMRI) provides a powerful non-invasive modality for identifying functional alterations. Existing deep learning (DL) studies employ diverse neuroimaging features; however, static functional connectivity remains widely used, whereas dynamic connectivity modeling is comparatively underexplored. Moreover, many DL models lack interpretability. In this work, we propose D-GATNet, an interpretable temporal graph-based framework for automated ADHD classification using dynamic functional connectivity (dFC). Sliding-window Pearson correlation constructs sequences of functional brain graphs with regions of interest as nodes and connectivity strengths as edges. Spatial dependencies are learned via a multi-layer Graph Attention Network, while temporal dynamics are modeled using 1D convolution followed by temporal attention. Interpretability is achieved through graph attention weights revealing dominant ROI interactions, ROI importance scores identifying influential regions, and temporal attention emphasizing informative connectivity segments. Experiments on the Peking University site of the ADHD-200 dataset using stratified 10-fold cross-validation with a 5-seed ensemble achieved 85.18% +_5.64 balanced accuracy and 0.881 AUC, outperforming state-of-the-art methods. Attention analysis reveals cerebellar and default mode network disruptions, indicating potential neuroimaging biomarkers.

Abstract

Spatiotemporal GNNs

Explore similar work

D-GATNet: Interpretable Temporal Graph Attention Learning for ADHD Identification Using Dynamic Functional Connectivity

NeuroBRIDGE: Behavior-Conditioned Koopman Dynamics with Riemannian Alignment for Early Substance Use Initiation Prediction from Longitudinal Functional Connectome

MPP-GNN: Subject-Adaptive Community Detection for fMRI-Based Alzheimer's Disease Classification