ORCID ID
https://orcid.org/0000-0002-1095-0654
Date Awarded
2024
Document Type
Dissertation
Degree Name
Doctor of Philosophy (Ph.D.)
Department
Computer Science
Advisor
Gang Zhou
Committee Member
Ye Gao
Committee Member
Huajie Shao
Committee Member
Denys Poshyvanyk
Committee Member
Meiyi Ma
Abstract
Ubiquitous sensing has been increasingly explored as a viable to means collect patient health data to improve how doctors can diagnose neurological diseases. Additionally, machine learning has been slowly adopted by healthcare researchers to automatically detect and treat neurological diseases such as Parkinson's disease (PD). In this dissertation, I investigate how wearable computing and machine learning can be used to automatically detect and treat freezing of gait (FoG) symptoms of PD in both real-time and offline settings. First, we designed a FoG detection and treatment system, FoG-Finder, which combines ubiquitous sensing, mobile computing, and machine learning to detect and treat FoG in real-time. We generated time and frequency domain features from sensor data, and used a multi-input convolutional neural network (CNN) model on a smartphone for real-time FoG detection. To evaluate FoG-Finder, we collected data from 11 PD patients consisting of 716 FoG events. FoG-Finder beat the previous state-of-the-art, obtaining a 13.4% higher F1-score and 10.7% higher overall accuracy compared with other validated real-time FoG detection and treatment systems in a leave-one-subject-out (LOSO) setting. FoG-Finder also obtained an average treatment latency of 427ms and 615ms for subject-dependent and LOSO settings, respectively. Second, we improved real-time FoG detection by identifying that turns are the most common aspect of non-FoG gait falsely identified as FoG by FoG detection systems. To address this revelation, we designed the Gait-Guard wearable FoG detection and treatment system. The system uses hand-crafted features derived from domain knowledge from PD clinicians and uses a light-weight multivariate time-series transformer model capable of real-time inference on mobile hardware. We evaluated Gait-Guard on a dataset consisting of 26 PD patients and a total of 1591 FoG events. Gait-Guard obtained a FoG event accuracy of 96.7% with a reduction in the false positive treatment rate of 112% compared to prior real-time FoG detection works including FoG-Finder. Gait-Guard also achieved an average FoG detection latency of just 378.5ms in a LOSO setting. Third, we created an automatic clinical FoG labeling system capable of accurately segmenting FoG and non-FoG time sequences. Our system utilizes real-time FoG detection models, a gait cycle metric estimation CNN, and clinical domain knowledge to estimate start and stop times for FoG. We evaluated our clinical FoG labeling system against our Gait-Guard PD dataset in a LOSO setting, and achieved a precision, recall, F1-score, and accuracy of 91.0%, 93.1%, 91.0%, and 91.6%, respectively. We achieved a start MAE of 0.75s and duration MAE of 1.18s. These results indicated that our system is potentially suitable for long-term FoG symptom monitoring in the real-world.
DOI
https://dx.doi.org/10.21220/s2-5n83-cj42
Rights
© The Author
Recommended Citation
Koltermann, Kenneth Howard, "Machine Learning And Wearable Computing For The Detection And Treatment Of Neurological Diseases" (2024). Dissertations, Theses, and Masters Projects. William & Mary. Paper 1727787878.
https://dx.doi.org/10.21220/s2-5n83-cj42
