About

Monitoring blood pressure (BP) with proper reliability is crucial for early prevention of cardiovascular diseases. Conventional approaches, although accurate enough, lack in real-time measurements and may cause discomfort due to its invasive nature. Predictive models pave the way for accurately predicting BP values but fail to be demographically appropriate. This study presents a novel joint validation framework that incorporates 3D Convolutional Neural Networks (3D-CNNs) and Vision Transformers (ViT) for interpretable prediction of systolic and diastolic BP of the normal BP patients. The model leverages multi-parameter physiological signals including ECG and PPG, which are transformed into image representations to capture temporal, frequency, and contextual patterns. It will integrate a smartphone video-based contactless BP estimation method using remote PPG (rPPG) signals and a demographic-aware prediction module that personalizes outputs based on age, gender, and BMI. The system is benchmarked with AAMI and BHS standards, demonstrating clinical relevance and robust performance. With enhanced interpretability and potential for real-time deployment, this approach marks a significant step toward scalable, personalized, and non-invasive BP monitoring.