The PRISMA project aims to develop an advanced technological platform for the remote monitoring of elderly people with multimorbidity and frailty. The platform integrates biometric sensors and artificial intelligence algorithms to collect and analyse clinical data in real time, anticipating signs of deteriorating health and providing personalised recommendations. The software part of the platform consists of three modules: a real-time data analysis system based on Recurrent Neural Networks, which involves an initial training phase and a subsequent active operation phase; an Anomaly Detection and Risk Classification system based on machine learning algorithms such as Isolation Forest and Gradient Boosting, in order to identify any anomalous patterns in the indicators monitored by the biometric sensors and classify their level of risk; a personalised recommendation system based on MedGemma, an open source LLM developed by Google and specialised in medical applications.
During the first half of the PRISMA project, the initial activities of WP1 were completed, aimed at developing an integrated remote monitoring platform for frail elderly people. Clinical and technical requirements were defined, low-invasive biometric sensors (heart rate, blood sugar, blood pressure, SpO₂) were selected, and hardware-software integration was completed via API for continuous data collection. At the same time, the modular architecture of the platform was designed and the training of artificial intelligence algorithms was started: RNN for real-time analysis, anomaly detection models (Isolation Forest, Gradient Boosting) for risk classification and MedGemma LLM for the generation of personalised recommendations. Preliminary tests were carried out in simulated environments with synthetic and experimental data, verifying transmission quality and basic predictive performance. The clinical data collection phase has slowed down due to difficulties in recruiting patients, resulting in a two-month extension of WP1 (new deadline M8). A contingency plan has been activated with an extended healthcare network and alternative suppliers to ensure the necessary sample volumes and the completion of deliverables (D1.2 and MS1) by the new deadline.