Investigation of Consumer Grade EEG as a Fall Risk Assessment Tool
Loading...
Date
2017-08-29
Authors
Shea, Kaela
Journal Title
Journal ISSN
Volume Title
Publisher
University of Waterloo
Abstract
Fall prevention for geriatric populations is a growing concern among clinicians and researchers due to severe risk of morbidity and loss of independence. Emerging evidence has demonstrated that cognitive workload while walking influences gait stability and the risk of falling. Electroencephalography (EEG) presents a potential method to provide objective measures, via Theta (4-7 Hz) and Alpha (8-13 Hz) frequency band powers, of cognitive workload during daily activities. Consumer grade EEG headsets increase the accessibility of EEG signal measurement through lowered cost and easier setup protocols. The following thesis
presents a series of studies investigating the sensitivity of the Interaxon Muse headband, and the Emotiv Epoc+ to measure cognitive load changes under ambulatory conditions (i.e., while walking). While the Muse yielded no sensitivity to changes in neural activity associated with changes in cognitive load levels, the Emotiv Epoc+ yielded high sensitivity to cognitive load changes across all 14 electrodes. Further research concerning this thesis centered around the use of the Emotiv Epoc+ system to distinguish levels of cognitive load under ambulatory conditions.
To examine the impact of motion artifact on EEG signals measured by the Emotiv Epoc+, a swim cap paradigm was used to isolate noise associated with gait. Signal to noise ratio (SNR) estimates indicate that EEG signals are 8 to 20 times the power of gait-induced noise, supporting the Emotiv Epoc+ during ambulatory monitoring conditions. Applying time and spectral system identification techniques, the relationship between motion-induced artifacts and recorded inertial measurement unit (IMU) yielded a strong nonlinear response. The final study of this thesis evaluated the utility of the Emotiv Epoc+ to measure 3 levels of cognitive load using a working memory paradigm while walking on a treadmill. A quadratic support vector machine (SVM) classifier was able to classify three levels of cognitive load at an accuracy of 70.3 %. These promising initial results, coupled with the short measurement time (10 sec), support the long-term goal of assessing cognitive load in an ambulatory environment towards implementation in fall risk assessment systems.
Description
Keywords
EEG, Motion Artifact, Geriatric, Fall Prevention, Classification