Date of Award

2021

Degree Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering

Department

Electrical, Computer, and Biomedical Engineering

First Advisor

Kunal Mankodiya

Abstract

This work is motivated by the impact of chronic diseases such as Parkinson's disease (PD) on the world. In particular, this work aims to expand digital health's ability to understand a disease's progression. The progression of the research walks through the current methods for tracking this progression, to find that this information is captured mainly in the healthcare office, starting with the initial evaluation and treatment plan formation. Afterward, roughly 2-3 times a year, the clinician uses various assessments to observe the patient through writing, speaking, and motor tasks to understand the current state of health. These efforts included 100+ people with PD (PwPD) and PD health care providers to craft the appropriate and straightforward tools needed for the holistic remote assessment of Parkinson's disease. The limit of this work concludes with an exploratory study, where PwPD use a wearable, any electronic device capable of being worn, and complete a questionnaire for 4 weeks to answer a critical question "Can daily worn wearables make the patient reported outcomes (PROs) contextual based on intelligent sensing of activity, experiences, and symptoms?" The work performed includes the initial understanding and mapping of the same assessments conducted in-office -- such as the Movement Disorder Society’s – Unified PD Rating Scale (MDS-UPDRS), PD Sleep Scale (PDSS), and PD Questionnaire (PDQ) -- together with wearable devices. To address this question, we built the PD-Aware digital health framework that would enable integrating the user's perception with wearables measure of performance. The aim is to ensure we understand the appropriateness of incorporating data generated by these new devices in the clinical conversation between doctor and patient.

The design of wearables includes both the form factor of the device as well a distribution of data collection and processing across the local network. In “WearUp: Wearable E-Textiles for Telemedicine Intervention of Movement Disorders” (published 2018) the ergonomics, materials, sizing and placement of the glove and its sensors are shown as import design choices to make early on to ensure the person using the design can provide feedback.

“Wearable IoT based Telemedicine” (published 2019) we step back from the textile design and focus on the hardware. Here we show the use of sensor fusion for complex sensors such as inertial measurement units, as well as ways to filter other one-dimensional sensors like the flex sensors. It is noted throughout the data pipeline considered factors such as available power, storage and transmission bandwidth influenced the systems architecture. We show how this large difference in power and bandwidth for personal area network devices against local area network devices can be considered against the other factors when determining the right layering of computational tasks. It’s concluded that personal area devices are appropriate for the intimate sensing but may require the incorporation of a local area network to maintain a larger context.

Afterwards we take a deeper dive into some of the implemented systems used for in-lab data collection with participants both with and without PD. The system included thing devices (gloves), edge devices (tablet), and a fog node (headless hub). The distribution of computing showed that the thing was capable of both feature extraction and threshold-based detection, while edge or fog devices with more computational power could run more complex models for classification. We perform and report on the timing and accuracy found across the system when used for the MDS-UPDRS section 3 (movement focused). The conclusion shows that we can identify the task performed and report back within 2-3ms at ~93% accuracy after the initial task start-up period of 2 seconds.

We review the exploratory study where 4 people with PD answered daily questionnaires PDQ (night) and PDSS (morning) while they wore a smart ring intended for capturing this information as well. The study protocol started with a 2-week session where participants had to report daily but without the ring such that we could determine their baseline of scoring without the wearable. The second half had them now wear the ring alongside the daily reporting. We found that select ring measurements had mixed correlations with the patient reported outcome measurement, while inconclusive due to too few participants. The early results are promising in that they showed an ability to integrate these two reporting mechanisms in an accessible manner for a person with a chronic movement disorder to record their own perceived experience alongside their performance.

We review the exploratory study where 4 people with PD answered daily questionnaires PDQ (night) and PDSS (morning) while they wore a smart ring intended for capturing this information as well. The study protocol started with a 2-week session where participants had to report daily but without the ring such that we could determine their baseline of scoring without the wearable. The second half had them now wear the ring alongside the daily reporting. We found that select ring measurements had mixed correlations with the patient reported outcome measurement, while inconclusive due to too few participants. The early results are promising in that they showed an ability to integrate these two reporting mechanisms in an accessible manner for a person with a chronic movement disorder to record their own perceived experience alongside their performance.

We conclude this work by resolving how wearables are currently used in health and where this space may head in the near future as healthcare costs rise alongside an aging population. The existing fabrics and computational devices face new guidelines in the manufacturability of e-textiles. The improved standardization is being seen also in the way personal health information can be shared or made available with appropriate access and control mechanisms in place. Ultimately, we see devices becoming more responsible for collecting important health information. These devices are constrained and require other devices for more complex processing and longer-term storage and dissemination of insights. Furthermore, the state of the research stands at interesting cross-roads of technology today where we have distributed ledgers making the tokenization of personal data possible helping to improve not only traceability but ultimately data ownership. We conclude this work by looking forward to a person with Parkinson's disease being able to perform specific assessments remotely and log daily activity in a manner that keeps raw data local and propagates patient generated health insights into the EHR ecosystem.

Available for download on Thursday, August 17, 2023

Share

COinS