Personalised multipoint calibration curves obtained by controlled hydrostatic changes of the upper limb for cuffless estimation of blood pressure from pulse transit time

Background: Blood pressure (BP) monitoring is an essential clinical tool for hypertension diagnosis. Conventional cuff-based methods of BP measurement are limited by their discontinuous nature. The new frontier of cuffless techniques may enable widespread, high frequency, BP estimation in the community. This thesis investigates the relationship between BP and arterial stiffness parameters by examining plausible calibration techniques for use in cuffless devices. Methods: Seated BP measurements were made using the first commercially released cuffless device (Casio BP-100 watch), with a two-point calibration obtained using a BP intervention (exercise) and compared to cuff-based measurements. A five-point calibration technique, was undertaken using hydrostatic changes to BP and corresponding pulse arrival time (PAT) measurements by linear regression. Calibration curves were assessed for repeatability over 60 days and BP estimates were assessed by mixed model analysis. Results: The mean difference in watch- and cuff-based measurements were 1±20 mmHg (p>0.05) for both systolic BP (SBP) and diastolic BP (DBP), with poor correlation for both (SBP: R=0.36, p=0.13; DBP: R=0.044, p=0.37). Calibration curves obtained using hydrostatic pressure changes had a mean slope of -1198±873 mmHg·s^-1 and intercept 288±153 mmHg. Curves had good repeatability, with mean changes of -185±1270 mmHg·s^-1 and 39±228 mmHg for slope and intercept respectively after 60 days. DBP estimates made using PAT and hydrostatic calibration curves had a mean difference of 2.4±18 mmHg from cuff-based measurements. PAT-based estimates accurately tracked directional changes in Cuff-BP. Conclusions: These findings present the utility of multipoint calibration curves for use in cuffless BP estimation. Calibration curves obtained by the hydrostatic effect showed repeatability and acceptable estimation accuracy using PAT. Future research is needed regarding the integration of these calibration methods in wearable devices and the effects of PAT confounders.