Test-Retest Reliability of Time-Domain EEG Features to Assess Cognitive Load Using a Wireless Dry-Electrode System.

Human Machine Interfaces (HMIs) can provide critical support and improve daily task functionality for prosthesis users or social interaction for patients with locked-in syndrome using an assistive communication device. One goal in the development of sophisticated HMIs is to reduce the cognitive load (CL) they place on the user to promote the use of the technology. Electroencephalogram (EEG)-derived measures collected with wired wet-electrode systems have been used to assess CL in laboratory environments and have demonstrated acceptable test-retest reliability. Assessment of CL during real-world unconstrained HMI operation, however, requires the use of a wireless dry-electrode EEG system which provides easier electrode application and untethered movement. However, the test-retest reliability of wireless dry-electrode systems to quantify CL has not been explored. Ensuring the consistent capture of CL-related signals across multiple sessions is critical if these devices are to be used to assess how improvements in HMIs affect CL. Therefore, the current study used a wireless dry-electrode EEG system to compare Evoked Response Potential (ERP) features of a simple auditory oddball task to measure CL during two separate testing sessions a week apart. ERPs of 11 subjects were recorded while participants performed a virtual task at two difficulty levels. A significant correlation was found between the P300 component of the ERPs and subjective ratings of CL during both testing sessions. Furthermore, there was a statistically significant test-retest reliability for this same ERP feature and similar signal-to-noise ratios (SNRs) across sessions.Clinical Relevance- This is an initial step in validating wireless dry-electrode EEG systems to assess cognitive load across multiple sessions. The evidence presented is critical if dry-wireless EEG systems are to be used to identify aspects of HMIs that reduce CL in clinical and real-life environments. Assessing CL in unconstrained environments can better inform clinicians and technology developers in their design of future HMIs.