Parkinson’s disease (PD) is a progressive neurodegenerative brain disorder that signifi- cantly impacts quality of life for those who are affected. It is a rapidly growing condition affecting millions of people worldwide, where treatments focus on managing symptoms and slowing the degenerative process, as there are no validated treatments that can stop its progression or preemptively prevent it. Effective management of the disease relies on accurate and timely assessment of symptoms based on clinical ratings, traditionally performed through clinical examinations using the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). However, in-clinic assessments are infrequent and may not capture the full spectrum of symptom fluctuations in daily life. While existing literature has focused on diagnosing PD, the current understanding falls short in terms of objectively quantifying its symptoms in daily-living conditions.

Following a design science research methodology, this thesis responds to this research gap by exploring the feasibility of using smartphones to quantify PD symptoms in a real- world, at-home setting. The research presents a cross-platform mobile application de- veloped for data collection from PD patients with the aim to identify promising system components and data types for capturing PD symptoms. Using data mining and machine learning techniques, the research explores if it is feasible to estimate the MDS-UPDRS scale based on objective measurements from smartphone-collected data. Additionally, it investigates the usability of the proposed mobile application for PD patients. By de- veloping and validating a cross-platform mobile application for symptom capturing, this thesis contributes both in terms of research results communicated in the associated peer- reviewed papers, and by providing an open source based app which makes PD symptom assessments more accessible, objective, and patient-centric.

Parkinson’s disease (PD) is a progressive neurodegenerative brain disorder that significantly impacts quality of life for those who are affected. It is a rapidly growing condition affecting millions of people worldwide. Its treatments focus on managing symptoms and slowing the degenerative process, as there are no validated treatments that can stop its progression or prevent it. Effective management of the disease relies on accurate and timely assessment of symptoms based on clinical ratings, traditionally performed through clinical examinations using the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). However, in-clinic assessments are infrequent and may not capture the full spectrum of symptom fluctuations in daily life. Modern consumer technologies, such as smartphones and wearable devices allow one to measure common symptoms thanks to the sensors embedded in these devices. Existing literature has clearly shown that these technologies can be used to diagnose PD; however, the current understanding falls short in terms of objectively quantifying its symptoms in daily living conditions.

Following a design science research methodology, this thesis addresses this research gap by exploring the feasibility of using smartphones and wearable technology to quantify PD symptoms in a real-world, at-home setting. The research presents a cross-platform mobile application developed for data collection from PD patients with the aim to identify promising system components and data types for capturing PD symptoms. Using data mining, statistical, and machine learning techniques, the research investigates whether it is feasible to estimate the MDS-UPDRS scale from objective measurements collected via smartphone data and wrist-worn actigraphy. Additionally, it investigates the usability of the proposed mobile application for PD patients and explores how data are interpreted by both patients and clinicians. The results show that smartphone-derived features can correlate strongly with reference wrist-worn actigraphy features (up to 0.91) for tremor-related assessments, while regression models estimate MDS-UPDRS Part III scores with mean absolute errors of approximately 6 points. The system also captures medication-related symptom fluctuations. Usability and user-centered evaluations further indicate that the proposed system is feasible for at-home use by patients, while also providing insights into how patients and clinicians may co-interpret symptom-tracking data and what they need from meaningful data representations. These contributions are presented through the associated peer-reviewed papers and supported by an open-source application that makes PD symptom assessment more accessible, objective, and patient-centric.