An Adaptive Algorithm for Detection of Arterial Blood Pressure Pulses

Clinically relevant cardiovascular system metrics can either be directly measured (e.g., systolic and diastolic pressure) or estimated (e.g., arterial compliance and resistance) from arterial blood pressure waveforms. Accurate characterization of such information requires identification of each individual arterial pressure pulse. This task is difficult due to variability in pulse morphology over time and throughout the arterial tree. While various pulse detection methods have been developed, many rely on arbitrary rules and thresholds or are computationally complex. We propose a simple real-time adaptive matched filter algorithm for pulse identification robust to changes in pulse morphology. Our algorithm uses recently identified pulses as a template for identification of future pulses. Our algorithm was validated on 186,030 annotated ABP pulses and yielded an average sensitivity of 99.82% and positive predictability of 99.94% across various cardiovascular conditions such as hypertension and hypotension. Moreover, our algorithm performed consistently regardless of the recording location on the arterial tree (i.e., central, femoral, and brachial). Overall, our method provides a reliable and computationally efficient framework for detecting the onset of ABP pulses in real time.