Design and Implementation of a Novel Subject-Specific Neurofeedback Evaluation and Treatment System.

Electroencephalography (EEG)-based neurofeedback (NF) is a safe, non-invasive, non-painful method for treating various conditions. Current NF systems enable the selection of only one NF parameter, so that two parameters cannot be feedback simultaneously. Consequently, the ability to individually-tailor the treatment to a patient is limited, and treatment efficiency may therefore be compromised. We aimed to design, implement and test an all-in-one, novel, computerized platform for closed-loop NF treatment, based on principles from learning theories. Our prototype performs numeric evaluation based on quantifying resting EEG and event-related EEG responses to various sensory stimuli. The NF treatment was designed according to principles of efficient learning, and implemented as a gradual, patient-adaptive 1D or 2D computer game, that utilizes automatic EEG feature extraction. Verification was performed as we compared the mean area under curve (AUC) of the theta band of a dozen subjects staring at a wall or performing the NF. Most of the subjects (75%) increased their theta band AUC during the NF session compared with the trial staring at the wall (p = 0.041). Our system enables multiple feature selection and its machine learning capabilities allow an accurate discovery of patient-specific biomarkers and treatment targets. Its novel characteristics may allow for improved evaluation of patients and treatment outcomes.