Can Pulse Arrival Time be Used for Cuffless Blood Pressure Estimation? A Clinical Study in ICU

Clinical studies have shown that Pulse Arrival Time (PAT) is potentially related with blood pressure (BP), but the subcomponent of PAT, i.e. Pre-Ejection Period (PEP), affects the reliability of BP estimation, so it is questionable whether PAT can be used to estimate BP. In this paper, we collect clinical data from 12 patients in the Intensive Care Unit (ICU) and validate the question based on different feature sets and regression models. We first verified whether it is feasible to use PAT alone for BP estimation. Subsequently, we explore the feasibility of using auxiliary features, including the statistical features of PAT (maximum, minimum and standard deviation) and three vital signs (heart rate, respiratory rate, and perfusion index), to facilitate the BP regression. The hybrid feature set led to a significant reduction of errors in BP regression. As compared to PAT alone, the MAE obtained by the hybrid feature set is reduced 11.68 mmHg for SBP, 7.90 mmHg for DBP, and 7.58 mmHg for MBP. Our conclusion is that PAT itself alone is not appropriate for BP estimation, due to the negative impact of PEP. But the fusion with auxiliary features can somehow weaken the effect of PEP on BP calibration.