Complexity Analysis based on Parietal Fuzzy Entropy to Facilitate ADHD Diagnosis in Young Children

Attention deficit hyperactivity disorder (ADHD) is the most common condition affecting the development of neurons in children. Therefore, early and accurate diagnosis of ADHD in young children is of paramount importance. In this study, the 8-channel wireless wearable EEG measurement device was employed to record EEG data from 30 children diagnosed with ADHD and 30 typical development (TD) young children aged 4-7 years. The data was collected both during rest and while the children performed a Kiddie Continuous Performance Test (K-CPT). We extract relative power spectral density (PSD) unaffected by factors like skull resistance and thickness. Additionally, a range of complex entropy values based on the time domain were extracted. These included sample entropy (SaEn), permutation entropy (PeEn), singular value decomposition entropy (SvdEn), and fuzzy entropy (FuEn). We compare the performance of k-Nearest Neighbors (kNN), Support Vector Machine (SVM), and XGBoost, and utilized the sequential forward selection (SFS) feature selection method in the wrapper approach. Through this process, the study identified the most effective EEG data segments and feature subsets. The findings indicated that using a combination of resting and K-CPT EEG data yielded greater discriminability. Notably, the study found that extracting beta power from the right occipital lobe along with fuzzy entropy from the parietal lobe resulted in optimal accuracy of 90% in distinguishing between children with ADHD and TD children. These outcomes highlight the potential of relative PSD and complexity metrics to support the clinical diagnosis of early ADHD. Furthermore, these metrics may contain unique neurobiomarkers that could be valuable for devising early intervention strategies.