BPAlert: A Non-Invasive, AI-Enhanced, Continuous Blood Pressure Monitoring System for Neonates in NICU - Project Summary Continuous blood pressure (BP) monitoring in the neonatal intensive care unit (NICU) is critical for detecting life-threatening conditions such as sepsis and organ failure. However, current BP- monitoring methods have major limitations. The gold-standard umbilical arterial catheter (UAC) provides accurate but invasive measurements, posing risks of infection and complications, while cuff-based oscillometry offers only intermittent readings. Photoplethysmography (PPG), already routinely used in NICUs for oxygen monitoring, provides an ideal signal source for developing continuous non-invasive BP-monitoring solutions. Yet, existing AI-driven PPG-based BP models continue to struggle with generalizability, interpretability, and clinical adoption. This project introduces an innovative hybrid BP-monitoring solution that combines AI-driven ∆BP estimation with periodic cuff-based measurements, to enable continuous, non-invasive and real- time tracking of neonatal BP. The proposed approach reframes the conventional problem of BP estimation: instead of directly predicting absolute BP values, the model focuses on estimating changes in BP (∆BP) over short intervals (e.g., every minute), thus improving generalizability of the model by decoupling intra-subject BP dynamics from inter-subject variability. Its self- explaining AI (S-XAI) architecture can highlight the signal regions driving predictions, enhancing transparency and clinical trust, while an incorporated alarm module detects clinically significant BP fluctuations in real time for early intervention. Aim 1 develops generalizable and interpretable AI-driven PPG-based ∆BP-estimation models with integrated alerts for adverse events. Aim 2 builds and validates a hybrid BP-monitoring method that combines oscillometric baselines with continuous AI-driven ∆BP estimates for real- time BP monitoring. Aim 3 assesses feasibility, usability, and workflow integration in collaboration with NICU clinicians, optimizing interpretability, alarm settings, and user experience. By combining the strengths of cuff-based and AI-driven methods, this project fills a critical gap in neonatal BP monitoring, delivering a safe, continuous, and clinically reliable solution. The expected outcome is a validated, ready-for-trial system with the potential to transform neonatal care by enhancing safety, enabling early intervention, and advancing cardiovascular research.
