Coherence analysis of invasive blood pressure and its noninvasive indicators for improvement of cuffless measurement accuracy

Cuffless technology for blood pressure (BP) measurement offers great potential for unobtrusive monitoring of hypertension. However, the accuracy is still a challenging issue probably owning to the inadequacy of existing noninvasive BP indicators in tracking BP changes, i.e., either the poor trackability of the indicators or the lack of relationship to map the indicators to BP. To figure out the covariation of BP-related noninvasive indicators and BP, we propose to analyze the coherence of intra-arterial systolic BP (SBP) and a selected group of noninvasive indicators, including the mostly used pulse transit time (PTT), photoplethysmogram intensity ratio (PIR) and RR interval (RRI) of electrocardiogram (ECG), in both time domain, with singular spectrum analysis (SSA), and frequency domain, with cross-spectral analysis. Through SSA the SBP and the indicators are decomposed into fast oscillations superimposed on a slow time-varying component. It is found that the fast variation components oscillate at similar frequencies for the SBP and the indicators, while that their slow variation components are out of phase synchronization. Cross-spectra between SBP and the indicators further illustrate stronger coherence correlation at high frequencies (HF, 0.15-0.40 Hz) (PTT: 0.54±0.30, PIR: 0.73±0.25, RRI: 0.63±0.25) than at low frequencies (LF, 0.04-0.15 Hz) (PTT: 0.24±0.19, PIR: 0.37±0.23, RRI: 0.33±0.23). The results suggest that the widely used indicators are poor to track the LF component of BP. Our study reveals the importance of exploring LF-related BP indicators as well as new models to integrate LF and HF information for improvement of cuffless BP measurement accuracy.