🏫 Academia Work (UM)
# Academic Research
At the University of Michigan in Ann Arbor, I researched specific applications of wearable technology for health monitoring. In particular, I researched the intersection of data science, signal processing, and biomechanics to enhance secondary prevention of motion sickness and dehydration. In these excerpts, I’ll quickly summarize the tangible, experimental work I completed.
For context, an abstract is copied below. For an in-depth look at the whole body of work, please refer to my dissertation.
# Motion Sickness
Motion sickness is typically understood as a condition whereby the afflicted feel nauseous, dizzy, and warm (though symptom profiles differ between people). With greater access to automated vehicles and alternate modes of transportation, incidence of motion sickness is likely to increase. Therefore, early detection of motion sickness is critical to the successful use of these mobility solutions. In my research, I designed the experiment and carried out human-subject testing to collect data and built a machine learning model to predict the likelihood of motion sickness onset.
# Dehydration
Exercise-induced dehydration is associated with a variety of symptoms including elevated heart rates, increased core body temperatures, and fatigue. On the field, excessive levels of dehydration can lead to decrements to performance; at worst, death in rare cases. However, current methods for detection tend to be either invasive or inaccurate. In my research, I explored using wearable sensors to develop and test a non-invasive method for assessing a significant level of dehydration (conventionally 2%). An experimental protocol was designed and executed to collect data across volunteering participants, from which machine learning models were developed.
# Dissertation Abstract
Wearable devices have enhanced health monitoring in clinical settings by effectively measuring physiological signals to inform prevention strategies. With the rapid development of sensors and data-driven decision-making, wearables can be applied in non-clinical settings to monitor various health conditions. Oftentimes, the most direct, accurate measurements are inaccessible or impractical during real-life, unscripted daily activities (e.g., equipment access). In this dissertation, signal-based models were developed to evaluate common wearables for health monitoring, with specific applications on motion sickness and dehydration.
Motion sickness can range from stomach discomfort to severe nausea and affects passengers more frequently than drivers. As automated vehicles and mobility solutions become normalized, motion sickness incidence is anticipated to increase among on-road passengers. As such, there is a greater need for early detection of vehicular motion sickness. Previous studies have shown postural instability to be associated with motion sickness. Therefore, assessments of standing balance may be useful for estimating levels of motion sickness. However, there are limited studies of post-drive standing balance that have been conducted in passenger vehicles or under ecologically-relevant conditions. In this dissertation, three studies quantified motion sickness and standing balance of vehicle passengers following continuous driving exposures deployed on a closed test track and on-road environments using a wearable inertial measurement unit. In the closed test track study, trunk postural sway increased significantly during the more challenging balance exercises. Post-drive changes to postural sway metrics (e.g., sway velocity and path length) were larger for drives during which participants performed a visual-based task on a handheld tablet-based device, as compared to drives without a task. In the on-road study, changes in post-drive postural sway were consistent with the findings from the closed test track study. However, there was no meaningful effect of performing a task on changes in postural sway metrics. In the third study, significant changes in post-drive postural sway were associated with the severest motion sickness responses, suggesting that sway metrics could characterize motion sickness. While preliminary, these findings could inform monitoring approaches of vehicular motion sickness using postural sway data from wearable sensors. Additional work would further explore wearables as a potential screening tool for motion sickness susceptibility prior to the drive.
In the fourth study of this dissertation, wearables were used to develop a noninvasive method for continuously measuring dehydration; untreated, dehydration can lead to performance detriments and in severe cases, death due to heat-related complications. Participants performed a series of orthostatic postural movements before and after a cycling session while donning common wearable that measured heart rate and trunk kinematic data. A machine learning model was trained and accurately classified a level of fluid loss equivalent to 2% of bodyweight. Using data from wearable devices, this method can support preemptive fluid replenishment and subsequently minimize potential decreases in performance; reduce the risk of serious heat injuries; and inform users to take additional hydration assessments.
These findings demonstrated the feasibility of wearable technologies for monitoring health conditions that are difficult to assess in non-clinical settings. Specifically, this dissertation developed models that could relate motion sickness and post-drive postural sway measured from wearable devices, and could reliably leverage common sensor-based signals to minimize dehydration. Future applications with wearable devices could especially support secondary prevention strategies, which are approaches aimed at minimizing the impacts of health conditions once they have occurred.