Bridging Gaps in Neurocare: Validation of FaaS as a Brainwave Neurosensing Technology
Presentation
Overview
Overview
Description
ABSTRACT
Introduction. Electroencephalography (EEG) assessments are pivotal in the acquisition of brain wave data, providing a real-time representation of electrical activity in the brain during various states of arousal, including sleep. EEG enables the identification of distinct sleep stages and quantifies the duration spent in each stage, offering valuable insights into sleep architecture. Clinicians routinely employ EEG in sleep studies to diagnose and manage a wide array of sleep disorders, such as insomnia, narcolepsy, and sleep apnea, among others. Recent epidemiological evidence suggests that sleep disorders are becoming increasingly prevalent, affecting approximately 30% of the adult population globally, with trends indicating a steady rise in incidence rates. In recent years, innovative technologies have sought to integrate EEG with other diagnostic tools to monitor and mitigate the risks of sleep-related impairments in real-time. The purpose of this study is to validate a novel EEG prototype (Nuream, Inc.) with an array of silver-chloride transducers sewn into a pillowcase (the “pillow array”), and methods of processing non-traditional EEG data from “Fabric as a Sensor” (FaaS) technology.
Materials and Methods. 15 college-aged students were recruited to participate in the first phase of this study. (Table 1) Participants reported their age, gender, and history of traumatic brain injury (TBI) before each 3-hour trial, where they lay on a bed with their head on the pillow array. (Figure 1) A traditional adhesive EEG electrode with electrolyte gel was placed in the center of each participant’s forehead to compare with data from the pillow array, and a pair of adhesive reference electrodes were placed on either side of the forehead sensor as baseline references for the “ActiveTwo” (Biosemi, BV) data acquisition system (DAQ). During each trial, data from the 64-channel pillow array and traditional EEG transducer were sampled at 512Hz and transmitted to a computer in an adjacent room. Data was monitored and saved with a custom LabVIEW program (ActiView) developed by Biosemi for their DAQ systems. A webcam was used to monitor each trial and identify subject movements and changes in head position.
Results. Custom MATLAB routines (MathWorks, Inc.) were used for multiple steps of data processing, including pre-processing to minimize signal interference, time-frequency domain conversion, and experimental methods of comparing results. Pre-processing first involved a linear detrend to offset gradual changes in the raw data. Then, a second-order Butterworth filter with a passband between 2 and 35 Hz was applied to minimize interference from waves outside of typical EEG frequency ranges. Each 30-second interval of data was shaped with a Hamming window before conversion to the frequency domain with a fast Fourier Transform (FFT). The distribution of energy across frequencies in each interval was used to characterize trends related to sleep stages, including the strength of anticipated signals (like Alpha waves), and average frequency values. The final stage of processing involves experimental methods of minimizing the influence of non-active channels, and combining test channels for direct comparison with the reference signal.
Discussion Figure 3 demonstrates the comparison of average frequency trends between a pair of “reference” and “test” data sets. In this example, frequency data from 8 sensors in the middle of the pillow array were combined as the test set to reduce interference from less-active sensors. Frequency values for each 30-second interval are shown, along with 5-minute rolling averages. The relationship between rolling averages of the test and reference sets is represented by a correlation coefficient of R=0.36. As the project continues, this processing method will be expanded to compare multiple trials, and new processing methods will be developed, including automated selection of active channels to process, and the isolation of specific waveforms to track.
