Keywords

EEG, VO2, Convolutional Neural Network, Prediction, Data Processing, Biomedical Engineering

Abstract

In this study, three EEG datasets, created via different levels of pre-processing, were used to train VO2 prediction models whose performance was tested and compared. Understanding the link between EEG and VO2 during walking is important when working in rehabilitation or studying metabolic energy optimization. Common methods of measuring these two bio-signals tend to be bulky, uncomfortable, and inefficient, so finding a way to predict VO2 from EEG via 1D CNN model training is a worthy pursuit.  It was hypothesized that, due to a CNN needing more data to find patterns for prediction, a minimally processed EEG dataset would have a higher accuracy in predicting VO2 than a cleaned dataset where bad channels are removed, or a cleaned dataset with bad channels zeroed out instead of removed. EEG and VO2 data were collected from young adults (n = 3) in good health. A 5-minute standing baseline, followed by a randomized order of five 6-minute fixed-speed walking trials (0.8, 1.0, 1.2, 1.4, 1.6 m/s)  and a self-paced walking trial were performed on an instrumented treadmill while EEG and VO2 data was collected. As predicted, the minimally processed dataset produced the most accurate model with 3.77% higher overall accuracy than the dataset with bad channels replaced and 3.92% higher than the bad-channels-removed dataset. Due to a lack of subjects and other limitations, further research is required to fully understand the link between EEG and VO2, however this suggests that CNN models can aid in this process when trained with minimally processed data containing both biological data and motion data, such as artifacts.

Completion Date

2026

Semester

Spring

Committee Chair

Helen Huang

Degree

Master of Science (M.S.)

College

College of Engineering and Computer Science

Department

Department of Mechanical and Aerospace Engineering

Document Type

Dissertation/Thesis

Identifier

DP0053269

Download

Share

COinS
 

Accessibility Statement

This item was created or digitized prior to April 24, 2027, or is a reproduction of legacy media created before that date. It is preserved in its original, unmodified state specifically for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the University Libraries provides accessible versions of archival materials upon request. To request an accommodation for this item, please submit an accessibility request form.