Developing a portable, customizable, single-channel EEG device for homecare and validating it against a commercial EEG device

Abstract

There are several commercial electroencephalography (EEG) devices on the market; however, affordable devices are not versatile for diverse research applications. The purpose of this project was to investigate how to develop a low-cost, portable, single-channel EEG system for a research institute that could be used for neurofeedback-related applications in homecare. A device comparison was intended to examine what system requirements such a system would need to achieve the secondary objective of developing a neurofeedback application that demonstrates the functionalities of the new device. A portable, single-channel EEG device prototype was realized that consisted of an amplifier module called EEG Click, a single-board microcontroller, an electrode cable, some disposable wet electrode pads, and a custom 3D-printed headband. Three pieces of software were developed: firmware for the prototype, two supporting computer applications for data recording, and visual neurofeedback. The neurofeedback application replayed a first-person view roller coaster video at a varying frame rate based on the theta band's mean power spectral density (PSD). The prototype was compared against a commercial device, InteraXon MUSE 2 (Muse). Technical measurements included determining the amplitude-frequency characteristics and signal quality, such as signal-to-noise ratio (SNR), spurious-free dynamic range (SFDR), and total harmonic distortion (THD). Furthermore, four physiological measurements were performed on six human test subjects, aged between 21-31 (mean: 26.0, std: 3.11), to compare the altered brain activity and induced artifacts between the two devices. The four tests were respiratory exercise, head movement exercise, eye movement exercise, and paced auditory serial addition test (PASAT), where each measurement included several epochs with various stimuli. After the recordings, PSD was calculated for each bandpass filtered epoch, then the spectra were split into theta (4-8 Hz), alpha (8-12 Hz), and beta bands (12-30 Hz).

The PSD values were averaged within each frequency band, and then these baseline-corrected mean values were the input for the repeated measures ANOVA statistical analysis. Results revealed that the amplitude-frequency characteristic of the prototype was low-pass filter-like and had a smaller slope than Muse's. The prototype's SNR, including and excluding the first five harmonics, was 6 dB higher, while SFDR and THD for the first five harmonics were roughly the same as Muse's. The two devices were comparable in detecting changes in most physiological measurements. Some differences between the two devices were that Muse was able to detect changes in respiratory activity in the beta band (F(8,16) = 2.510, p = .056), while the prototype was more sensitive to eye movement, especially lateral and circular eye movement in theta (F(2,8) = 9.144, p = .009) and alpha (F(2,8) = 6.095, p = .025) bands. A low-cost, portable EEG prototype was successfully realized and validated. The prototype was capable of performing homecare neurofeedback in the theta band. The results indicated it is worth exploring further the capabilities of the prototype. Since the sample size was too small, more complex physiological measurements with more test subjects would be more conclusive. Nevertheless, the findings are promising; the prototype may become a product once.