A deep learning approach with event-related spectral EEG data in attentional deficit hyperactivity disorder

Abstract

Attention deficit hyperactivity disorder (ADHD) is a heterogeneous neurodevelopmental disorder that affects 5% of the pediatric and adult population worldwide. The diagnosis remains essentially clinical, based on history and exam, with no available biomarkers. In this paper, we describe a deep convolutional neural network (DCNN) for ADHD classification derived from the time-frequency decomposition of electroencephalography data (EEG), particularly of event-related potentials (ERP) during the Flanker Task collected from 20 ADHD adult patients and 20 healthy controls (HC). The model reaches a classification accuracy of 88%, superior to resting state EEG spectrograms and with the key advantage, compared with other machine learning approaches, of avoiding the need for manual selection of EEG spectral or channel features. Finally, through the use of feature visualization techniques, we show that the main features exciting the DCNN nodes are a decreased power in the alpha band and an increased power in the delta-theta band around 100ms for ADHD patients compared to HC, suggestive of attentional and inhibition deficits, which have been previously suggested as pathophyisiological signatures of ADHD. While confirmation with larger clinical samples is necessary, these results highlight the potential of this methodology to develop CNS biomarkers of practical clinical utility.

Investigation of Machine Learning Methods for Early Prediction of Neurodevelopmental Disorders in Children

“…Similarly, using the ABIDE dataset, which was based on rs-fMRI and deep neural network, ASD was distinguished from typically developing subjects [71]. Wireless Communications and Mobile Computing In addition, it was found that CNN algorithm was most efficient among all applied ML algorithms, and several studies have reported the rise in accuracy for ADHD diagnosis and examination by utilizing CNN with an accuracy range of between 90 ± 10 percent [55,[72][73][74][75][76][77][78]. Similarly, numerous studies were also conducted using CNN for ASD diagnosis and analyses showing a high accuracy rate > 70-90% [44,[79][80][81][82].…”

Machine learning validation of EEG+tACS artefact removal

“…In recent years there have been several investigations into the removal of this artefact in the EEG by applying post-collection signal processing to the raw EEG data [7][8][9][10][11][12][13][14][15][16][17]. There have also been investigations into the properties of the tACS artefacts [18][19][20][21][22], for example how they spread across the head [19] and how they mix with other bio-signals such as ECG and breathing patterns [18].…”