Detection of ADHD Disorder Using Dynamic Connectivity Tensors in Bidirectional Circular Reservoir Computing
Mohammadreza Bakhtyari, Sayeh Mirzaei · 2021
attention deficit hyperactivity disorder (ADHD) is a type of neurodevelopmental disorder. These disorders affect the development of the human nervous system and lead to abnormal brain function, which may influence a person's emotions, ability to learn and memory. It is also possible that the effects of neurodevelopmental disorders will continue throughout a person's life. Due to the lack of biomarkers to diagnose ADHD, the risk of misdiagnosis is high. The physician diagnoses the disorder with a description given to him by the patient or designed tests. Since biological signals such as electroencephalography (EEG) can record and measure brain function, they can help diagnose this disorder. In this study, an innovative method for extracting the characteristics of EEG signals is presented, which includes two steps. EEG signals are first converted into temporal segments, and then the spatial features are encoded using a dot product between the time frames of the channels. This data constructs the input of the model, which consists of three parts. In the first stage, the data enters reservoir computing module to extract dynamic features of data. After performing the required calculations, we use the Principal Component Analysis (PCA) technique for dimensionality reduction. In the last step, a Support Vector Machine (SVM) classify the data. This method obtains 99% accuracy, which is the highest achieved accuracy on the data used in this study.