| Popis: |
Wearable sensors are being used in clinical settings to monitor the condition of patients as well as in recreational environments for routine health monitoring. Some of the most advanced clinical applications include monitoring patients with Parkinson's disease through wearable inertial measuring units (IMUs) and patients with diabetes by means of wearable glucose sensors. Prominent examples of wearable sensors in routine use are fitness trackers, step-and calorie counters. Recently, wearables have evolved to being capable of running artificial intelligence algorithms in real-time at the point of sensing which allows to gain analytical insights directly from measurement data. We call such intelligent wearables with AI-at-the-edge functionality THINKables. First use cases for THINKables have emerged in both clinical and nonclinical applications: real-time seizure prediction or detection systems for epilepsy patients, or digital coaches providing real-time feedback to athletes on performance and injury risks. Technological and regulatory challenges of developing and deploying THINKables are multifold: data privacy and security of monitoring data needs to be ensured at all times, analytical AI models need to be transparent, explainable and fair, and all these features need to be implemented taking the limited computing power of point-of-sensing processors into account. In order for THINKables to become integrated into clinical workflows, all stakeholders in the Health AI ecosystem (regulators, clinicians, biomedical device technologists, pharma and biotech sectors, data scientists, and patients) need to work together to create frameworks for responsible and meaningful use. |
