On the classification of EEG/HEG-based attention levels via time-frequency selective multilayer perceptron for BCI-based neurofeedback system

Attention Deficit/Hyperactivity Disorder (ADHD) is a neurobehavioral disorder which leads to the difficulty on focusing, paying attention and controlling normal behavior. Globally, the prevalence of ADHD is estimated to be 6.5%. Medicine has been widely used for the treatment of ADHD symptoms, but the patient may have a chance to suffer from the side effects of drug, such as vomit, rash, urticarial, cardiac arrthymia and insomnia. In this paper, we propose the alternative medicine system based on the brain-computer interface (BCI) technology called neurofeedback. The proposed neurofeedback system simultaneously employs two important signals, i.e. electroencephalogram (EEG) and hemoencephalogram (HEG), which can quickly reveal the brain functional network. The treatment criteria are that, for EEG signals, the patient needs to maintain the beta activities (13–30 Hz) while reducing the alpha activities (7–13 Hz). Simultaneously, HEG signals need to be maintained continuously increasing to some setting thresholds of the brain blood oxygenation levels. Time-frequency selective multilayer perceptron (MLP) is employed to capture the mentioned phenomena in real-time. The experimental results show that the proposed system yields the sensitivity of 98.16% and the specificity of 95.57%. Furthermore, from the resulting weights of the proposed MLP, we can also conclude that HEG signals yield the most impact to our neurofeedback treatment followed by the alpha, beta, and theta activities, respectively.