Research Report on Potential Biomarkers for Early Detection of ADHD: Phase-Based Functional Brain Connectivity and Graph Theory Analysis

This is a research report titled “Potential Biomarkers for Early Detection of ADHD: Using Phase-Based Functional Brain Connectivity and Graph Theory Analysis”. This study was conducted by Farhad Abedinzadeh Torghabeh, Seyyed Abed Hosseini, and Yeganeh Modaresnia, and was published in Physical and Engineering Sciences in Medicine (2023), Volume 46, pages 1447-1465. The paper was published online on September 5, 2023. The academic background, research methods, experimental results, and scientific value will be elaborated in detail.

Academic Background and Research Questions

Attention Deficit Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder characterized by symptoms of inattention and hyperactivity/impulsivity that significantly affect the daily lives of children. According to meta-analysis data, the prevalence of ADHD among children and adolescents worldwide is approximately 5.3%, with a male-to-female ratio of about 2:1. Scientists believe that ADHD is primarily caused by dysfunctions in specific brain regions that control attention and focus, and genetics also play an important role in its development. Children with ADHD often face social difficulties, academic problems, self-esteem issues, and a negative impact on family life. Additionally, these children may face challenges in specific cognitive abilities such as language and audiovisual memory, motor coordination, working memory, vigilance, impulsivity and inhibitory control, programming, and action sequences. Therefore, early detection and intervention are crucial and can enable effective intervention before complex factors emerge.

The importance of functional connectivity and its changes connecting different brain regions has been widely studied in various neurological conditions. Graph Theory (GT) is an effective method for evaluating the functional and effective connectivity of brain networks, depicting the structure of brain networks through mathematical representations of nodes and edges. Studies of brain connectivity typically utilize various methods including Electroencephalography (EEG), Magnetoencephalography (MEG), and Functional Magnetic Resonance Imaging (fMRI). Among these, EEG is widely used as a diagnostic tool for brain connectivity research due to its high temporal resolution, cost-effectiveness in data acquisition, and broad frequency band.

Research Source and Author Information

This study was written by Farhad Abedinzadeh Torghabeh, Seyyed Abed Hosseini, and Yeganeh Modaresnia, all from the Islamic Azad University, Mashhad branch, Iran. The study was received on April 27, 2023, accepted on July 24, 2023, and published online on September 5, 2023, in Physical and Engineering Sciences in Medicine.

Research Workflow

Database and Preprocessing

This study used publicly available datasets including EEG recordings of 61 children with ADHD and 60 healthy children (Healthy Control, HC) with no history of psychological disorders. All subjects were right-handed, with an average age of 9.73 ± 1.76 years. During a visual attention task, EEG signals were recorded from 19 electrodes at a sampling rate of 128 Hz according to the international 10-20 system.

EEG signals were preprocessed using the EEGLAB toolbox (version 2022.1). Continuous EEG signals were first filtered using a bandpass finite impulse response (FIR) filter of 1-48 Hz to eliminate power line noise. Subsequently, the Clean Rawdata plugin was used to automatically remove visible artifacts caused by electrode displacement. The re-referencing procedure was then performed to re-reference the average reference value of all channels. EEG signals were decomposed using Independent Component Analysis (ICA) to remove artifacts such as muscle patterns and eye blink/movement. These components were automatically identified and removed by the ICLabel plugin. Subsequently, the time series were filtered into conventional EEG frequency bands and then segmented.

Phase Measurement Methods

Phase-Lag Index (PLI)

This study measured the functional connectivity of all 19 pairs of nodes in each frequency band and each segment using PLI. This method addresses common issues in phase-based connectivity measurements, typically ignoring zero-phase lag connectivity. The theory of PLI posits that when a false connection appears due to volume conduction, the phase angle difference will be dispersed around 0 radians. On the other hand, if the connection is not due to volume conduction, the phase angles will primarily display a positive or negative distribution. Therefore, a symmetrical distribution may indicate a false connection, while a deviation from symmetrical distribution suggests a source-dependent connection.

Inter-Site Phase Clustering (ISPC)

ISPC is a phase synchronization measurement commonly used in EEG connectivity analysis, calculating phase angle consistency between two signals at different time points. ISPC computes the average phase angle and shows the differences between channel pairs, indicating the range of EEG connectivity from zero to one.

Graph Theory Analysis and Classification

EEG connectivity analysis was conducted using Graph Theory (GT), extracting seven commonly used local features including Clustering Coefficient (CC), Local Efficiency (LE), Louvain Community (LC), Node Strength (NS), Node Degree (ND), Node Betweenness Centrality (NBC), and Subgraph Centrality (SC). Six classification methods were then used, including k-Nearest Neighbors (KNN), Linear Discriminant Analysis (LDA), Naive Bayes (NB), Decision Tree (DT), Linear Support Vector Machine (SVM), and a three-layer Artificial Neural Network (ANN).

Statistical Tests

Statistical tests determined the presence of significant differences between the two groups by calculating p-values. T-statistical tests were conducted on the graph features to test the hypotheses and provide reliable biomarkers.

Research Results

Classification Results of ISPC and PLI Connectivity Measurements

The classification of ISPC and PLI connectivity measurements achieved average classification accuracies of 99.174% and 98.347%, respectively. The specific classification accuracies under particular models and frequency bands are also listed in the paper.

Statistical analysis showed that ADHD patients had higher feature values in their δ, θ, and α frequency bands and their frontal (F3) and central (C3) channels.

Conclusion

This study proposed potential biomarkers for the early detection of ADHD through analysis of EEG data using phase-based functional connectivity and graph theory. The research demonstrated that SC features in PLI connectivity in the β frequency band had high classification accuracy, and NBC and ISPC features in the δ and θ frequency bands also showed strong differentiation capabilities. These biomarkers showcased great potential in the diagnosis of ADHD and could help identify effective intervention strategies, thus improving the quality of life for ADHD patients.