More about HKUST
Sensing and Analysis Systems for Parkinson's Disease Management: A Survey
PhD Qualifying Examination
Title: "Sensing and Analysis Systems for Parkinson's Disease Management: A
Survey"
by
Miss Yizhen ZHANG
Abstract:
Parkinson's Disease (PD) is the world's second most prevalent
neurodegenerative disorder, traditionally managed through the
semi-quantitative MDS-UPDRS scale. However, this approach is limited by
subjective clinician bias and a "snapshot" nature that fails to capture daily
motor fluctuations. Consequently, PD management is undergoing a paradigm
shift toward continuous, objective, and home-based digital monitoring, driven
by the need to capture "real-world performance" rather than just "clinical
capacity."
This survey systematically reviews the landscape of PD sensing and analysis
systems through a three-tier hierarchical architecture: Perception,
Processing, and Output. We first analyze the Perception layer, categorizing
multi-modal data sources into contact-based wearables, non-contact ambient
sensing (e.g., Vision and RF), and behavioral digital interactions.
Furthermore, we discuss the evolution of the Processing layer, where raw
heterogeneous data is distilled into refined digital biomarkers using
paradigms ranging from traditional kinematic modeling to AI-driven deep
learning. The Output layer investigates the mapping of these parameters to
actionable clinical insights and quantified MDS-UPDRS scores. Specifically,
we explore the transition of application scenarios from task-driven clinical
assessments to passive longitudinal symptom tracking.
To address the critical research gaps in self-administered home testing and
high-fidelity motion reconstruction, we highlight two of our research
projects: PIGDAssess, which utilizes dual-task interference for autonomous
postural stability evaluation, and TulipTender, a physics-aware 3D motion
analysis framework using monocular commodity cameras. Finally, we identify
future directions, including closed-loop management and the integration of
Large Language Models (LLMs) for clinical decision support. This review
underscores the transformative impact of deep learning-driven sensing in
shifting PD care toward a proactive, precise, and intelligent management
paradigm.
Date: Monday, 20 April 2026
Time: 4:00pm - 6:00pm
Venue: Room 2126A
Lift 19
Committee Members: Prof. Qian Zhang (Supervisor)
Prof. Mo Li (Chairperson)
Dr. Xiaomin Ouyang