EEG Signal Classification for ADHD Based on Bidirectional GRU and Forward Channel Selection

Attention Deficit Hyperactivity Disorder (ADHD) is a common neurodevelopmental disorder in children, and many scholars have used Electroencephalography (EEG) to study it and achieved good results. However, current models still have the problems of too many channels and weak spatial dependence, which may increase computational complexity and affect practical applications. In this study, a new classification model is constructed, which can not only analyze the spatial relationship but also introduce the forward feature selection method. Using Pearson correlation and random forest, the independent diagnostic ability of all channels is analyzed, and the baseline model is constructed to extract temporal features. In the study, the forward feature selection strategy is used to screen a 10-channel subset from the original 19 channels. The experimental results show that the test accuracy of the 10 core electrodes is relatively high, which can significantly reduce the false positive rate in diagnosis and provide effective support for the screening of ADHD.