Artificial Intelligence (AI)-enabled mining of big longitudinal datasets collected with wearable sensors.

Medical diagnosis and patient follow-up currently rely on the experience of highly specialised clinicians, who evaluate patient symptoms during in-person visits, augmented with results coming from specialised laboratory tests and subjective patient reports. Such assessments are infrequent, potentially missing out on longitudinal symptom variation and depend on the specialist’s ability to find patterns in a variety of complex and heterogeneous data. However, a wide range of digital measurement technologies (smartwatches and smartphones) offer the opportunity to record and store various aspects of health and pathological symptoms in a more frequent and detailed way throughout the lifespan of a patient. For example, regular testing of tremor and memory, passive continuous assessment of gait and geolocation or extracting physiological signatures from wearable sensors can provide a wealth of information on neurologic (e.g. epilepsy) and neurodegenerative disorders (e.g. Parkinson’s disease (PD), and Alzheimer Disease (AD)), which are infamously heterogeneous in regard to onset age, symptom prevalence, and severity progression rate. The goal of this project is to capitalize on previous experience and recent advances in artificial intelligence (AI) to develop cutting-edge theory and algorithmic approaches for advanced medical decision support systems, diagnostic screening and continuous patient monitoring aimed at providing actionable intelligence to patients, clinicians, medical stakeholders. We will develop analysis techniques that robustly integrate information across different tests and across time with a minimal amount of sensors.  We will investigate which analysis techniques will be able to achieve stellar diagnostic accuracy and disease severity assessment, even in the presence of real life noise.

Keywords:AI, clinical informatics, health monitoring, digital health

Disciplines:Health informatics