Keywords
"EEG", "metabolic cost", "deep learning", "Convolution neural network", "cost of transport"
Abstract
Electroencephalography (EEG) is a technique used to non-invasively record electrical signals from the brain that can be used to understand how the brain functions and controls movements. EEG, however, is susceptible to recording electrical artifact signals that do not originate from brain activity. Common artifacts that come up in EEG data are generated by eye movements, heart contractions, muscle contractions, motion artifacts, electromagnetic interference signals, and powerline noise. These artifacts are often filtered out of EEG data or reduced to clean the signals and preserve the information from the brain. However, some of these signals may provide useful information about a person’s metabolic cost or the amount of energy required to perform a task. To measure metabolic cost, individuals breathe into a mask or mouthpiece attached to a metabolic analyzer in a process called indirect calorimetry. Metabolic cost is a common measure for evaluating the effectiveness of rehabilitation assistive devices. As such, several studies have used wearable sensors such as inertial measurement units (IMUs), electromyography (EMG), and heart rate monitors to estimate metabolic cost. The purpose of this study was to explore using EEG and its artifacts to predict the metabolic cost of individuals while they walked on a treadmill at a range of fixed speeds and a self-paced speed. We used a dual-layer EEG system that simultaneously records normal EEG on the scalp and electrical signals that are not generated from the body but rather by motion or from the environment. We recorded dual-layer EEG and metabolic energy as three young adults walked on a treadmill at a range of speeds. We then trained a deep learning convolution neural network (CNN) model with two subsets of data (just the motion artifact signals, and the minimally processed traditional EEG signals recorded on the scalp) data to predict metabolic cost to determine which subset of data had more information. Our preliminary results suggest that there is potential for using EEG to predict metabolic cost.
Completion Date
2024
Semester
Fall
Committee Chair
Helen Huang
Degree
Master of Science (M.S.)
College
College of Engineering and Computer Science
Department
Mechanical and Aerospace Engineering
Degree Program
Biomedical Engineering
Format
Identifier
DP0028991
Language
English
Release Date
12-15-2024
Access Status
Thesis
Campus Location
Orlando (Main) Campus
STARS Citation
Grubb, Jordan, "Using electroencephalography to predict metabolic cost" (2024). Graduate Thesis and Dissertation post-2024. 29.
https://stars.library.ucf.edu/etd2024/29
Accessibility Status
PDF accessibility verified using Adobe Acrobat Pro Accessibility Checker