Blood Pressure Prediction Analysis from Doppler Ultrasound Signals Using Machine Learning

Blood Pressure is an important diagnostic in field of medical science and sometimes BP measurement which rely on external pressure, can be unsuitable for patients with burns, mus-cle sensitivity. This study presents a low complexity (CNN based) contactless method for estimating systolic and diastolic blood pressure using Doppler ultrasound US samples from patients' blood flow. Each participant provided three Ultrasound (US) files, which were analyzed to capture blood flow patterns associated with varying blood pressures. Employing signal processing and convolutional neural networks, features from these US signals were extracted to predict blood pressure values. Initial results demonstrate the potential of this technique to offer accurate, non-pressure continuous blood pressure monitoring, enhancing diagnostics and real-time health management.