ADHD detection from EEG signals using GCN based on multi-domain features

Ling Li ¹ Xueyang Guo ¹ Zihan Yang ¹ Yanping Zhao ^1* Xu Liu ¹ Junxian Yang ¹ Yanyan Chen ² Xinxian Peng ² Lina Han ²

• ¹ Jilin University, Changchun, China
• ² Changchun Sixth Hospital of China, Changchun, China

Attention deficit hyperactivity disorder (ADHD) is a common psychiatric disorder in children during their early school years. At present, many researchers have been devoted to ADHD detection and achieved many results. However the research of accurate, rapid and automated detection methods is still a challenge. In this study, we propose an ADHD detection method using a graph convolutional neural network (GCN) model based on multi-domain features of EEG signals. Firstly, we use the long short term memory (LSTM) and convolutional neural network (CNN) models to extract the time domain and frequency domain features of EEG signals, respectively. Secondly, we combine the phase lag index (PLI) and coherence (COH) as a fusion feature to construct a new functional connectivity matrix reflecting brain functional connectivity. Finally, we design a GCN model combining the features extracted by the LSTM and CNN models and the functional connectivity features to identify ADHD. Our novel functional connectivity matrix constructed by the fusion features can simultaneously reflect the phase synchrony and signal intensity consistency between brain regions, which is superior to the traditional functional connectivity markers. Moreover the GCN model can fully extract the topological feature of the functional connectivity matrix, and combine the three kinds of features in the time domain, frequency domain and topology to achieve more accurate and effective detection of ADHD. In order to value the effectiveness of the proposed method, we test it on two EEG datasets, resulting in achieving the average accuracy rates of 97.29% and 96.67% respectively, and the proposed method performs better performance by comparing it with the other models, including XGBoost, LightGBM, AdaBoost, and random forest. In addition, visualization experiments reveal the differences in brain connectivity distribution between ADHD patients and healthy subjects. These experimental results show that the proposed method has obvious advantages on ADHD detection, and will be used to assist neurologists in diagnosing ADHD.