Healthcare represents one of the most promising areas for innovative applications exploiting the computing continuum, and the concept of personalised healthcare is becoming a reality through AI-enabled functionality. The specific focus of this use case was addressing the risk of strokes (one of the most critical health-related risks). Patients wearing both commonly available and specially designed wearable devices could be monitored over extended periods of time, with their personal data directly analysed by AI enabled software to detect potential signs of stroke risks - thereby permitting preventive measures to be taken.
Testing AI-SPRINT technology on personalised stroke risk assessment and prevention with enhanced AI models. Using wearable and mobile devices to capture new insights on patient care powered by AI, with the smart allocation of the workload between cloud and edge. This use case implements the COMPSs programming models and machine learning developments. It also uses the cloud-edge environment as an effective framework for impactful clinical applications for stroke prevention. AI-SPRINT will deliver the benefits of incorporating wearable technology into healthcare, such as continuous data acquisition and low patient burden while ensuring the protection of sensitive data and healthcare risk forecasting models.
The drivers for AI-SPRINT are enhancing BSC’s high-performance data analytics frameworks in edge-to-cloud platforms to manage distribution and parallelism across resources while ensuring the protection of sensitive data and healthcare risk forecasting models.
The underlying architecture joins specific design and runtime tools that encompass all the required components and services. The deployable infrastructure consists of Kubernetes clusters and groups of edge devices as 'simulated hospitals', specifically patients' personal mobile phones connected via Bluetooth to individual smartbands. Secured with SCONE, the deployable infrastructure is used for model training (before and during the pilot study), federated learning, and inference on-demand using OSCAR. At runtime level, COMPSs orchestrates the distribution of the computation, MinIO provides local storage and cloud synchronisation, InfluxDB takes care of monitoring services, IM and EC3 manage container deployment and elasticity, respectively. At design level, SPACE4AI-D selects the resources based on defined performance constraints.
The final model will provide personalised notifications, alerts, and recommendations for stroke prevention, which is a reduced number of strokes, and a faster detection of risks of strokes.
This project, based on Artificial Intelligence in the healthcare industry, represents a unique opportunity for implementing effective healthcare applications working in real-time monitoring patients’ vital signs. Using Artificial Intelligence to help reinvigorate modern healthcare, such as:
- Improved patient monitoring.
- Enhanced patient care.
- Reduced human error.
- Better managed patient flow.
- Continuous and non-invasive monitoring over long periods of time
- Identification of early indicators of health issues
- Management of at-risk and vulnerable people
- Personalisation of technological solutions matching individual needs
- Patient empowerment and information about their own health
- Patient engagement in healthier lifestyle and behaviour change
- Reduced strain on the healthcare system
- Accelerated digitisation of healthcare
- Personalised care treatment plans
- Innovation and growth of the wearable tech industry
- Improve confidence in digital health through privacy and security