Telemonitoring is the use of electronic devices or smartphones to remotely monitor patients. Taking Parkinson’s Disease (PD) as an example, smartphone apps and wearables enable remote, internet-based measurement of PD symptoms. Translating these remote sensing signals into assessments of the disease severity and progression through predictive analytics and machine learning provides a great opportunity to allow for frequent, cost-effective, and timely tracking of patient status, leading to effective treatment adjustment and intervention. Significant data science challenges exist, such as individual differences, big data, low signal-to-noise ratio in patient-self-collected data due to lack of compliance, etc. This project develops machine learning algorithms to address these challenges.
Methodologies:
- Positive transfer learning
- Semi-supervised learning & weakly-supervised learning
- Feature and instance selection