The classification of ADHD using machine learning and EEG data : a systematic review

Attention deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder that spans a diverse range of symptoms such as inattention, impulsiveness, and hyperactivity. The current gold standard diagnostic process for ADHD can pose challenges due to the variations in symptoms between individuals and a shortage of trained clinicians. An alternative approach to improve the efficiency and accuracy of diagnosing individuals with ADHD could be the integration of artificial intelligence (AI) or machine learning (ML) with the current best practice diagnostic guidelines. This study aimed to systematically review the efficacy and accuracy of using EEG data in conjunction with ML to objectively classify individuals with and without ADHD. We reviewed 57 studies in total that used EEG data and various ML classifiers to categorise ADHD diagnoses. We assessed the classification accuracy, sensitivity, and specificity of the results from each study to evaluate the current effectiveness of using ML approaches to classify ADHD versus non-ADHD presentations. Furthermore, each study underwent a thorough methodological quality assessment to inform the robustness and reliability of each included publication. Although numerous studies demonstrated high classification accuracy, a large number of the publications failed to meet the Mixed Methods Appraisal Tool (MMAT) quality assessment standards. Therefore, caution should be exercised when generalising the findings to the broader efficacy of ML in ADHD classification. Future research should integrate psychological research methods to facilitate the translation of research findings into clinical applications.