Attention deficit and hyperactivity disorder classification with EEG and machine learning

Understanding the abnormalities in brain function aids differential diagnosis of various brain disorders; the complexity of this process initiates with the shared symptoms and increases with the subjectivity of the psychiatric diagnosis. Quantitative EEG analysis is a powerful tool to discard neurological abnormalities in neuropsychiatric diagnosis and remarkably used to find neurophysiological or psychiatric biomarkers. Attention deficit and hyperactivity disorder (ADHD) is one of the most common neuropsychiatric disorders in childhood that can remain in adulthood, characterized by behavior symptoms. For ADHD, biomarkers search has been pointed to monitor treatment effects (pharmacological and neurofeedback), to compare the effectivity of clinical criteria, and to track changes at different age of development. As comorbidities are common in ADHD, EEG is also used to determine the pharmacological treatment. With machine learning models, all valuable information for decision-making from several medical experts can converge in the optimal available solution or find patterns in the complexity of data; hence, there is hope for the understanding of underlying causes of the disorder and the effects of psychotropic drugs. Most of the reports of ADHD classification with machine learning use datasets from ADHD children and healthy controls, however, there is a lack of studies comparing data from other disorders. In this sense, it has been widely reported that ADHD has comorbidities, therefore more research needs to be carried out for differential diagnosis, not only to identify abnormalities but also to differentiate among the various disorders or disabilities that are found with ADHD. Finally, any validated biomarker of ADHD based on neuroimaging data should be included in clinical practice as part of the regular assessment of the expert in an intelligent decision-making support system; such a method has been assumed to be able to improve specificity, reducing overdiagnosis and, hopefully, underdiagnosis, thereby avoiding undesirable consequences in adulthood.