A novel art of continuous noninvasive blood pressure measurement

The study addresses the need for accurate, continuous, and wearable blood pressure (BP) monitoring suitable for ambulatory and general ward use. Continuous hemodynamic monitoring is a mainstay in perioperative and intensive care but typically requires invasive or bulky stationary systems. Existing cuff-less wearable approaches based on time metrics (e.g., pulse transit/arrival time) lack sufficient clinical accuracy and require frequent recalibration, while noninvasive direct BP technologies based on vascular unloading are too bulky for wearables. The purpose is to introduce and validate CNAP2GO, a direct continuous noninvasive BP measurement method using a volume control technique (VCT) designed for miniaturization (e.g., finger-ring), capable of self-calibration, robust tracking of mean BP, and derivation of advanced hemodynamics. The significance lies in enabling medical-grade continuous BP and flow monitoring beyond critical care settings, potentially improving patient surveillance and outcomes, including during crises such as COVID-19.

The paper reviews cuff-less time-based BP estimation methods (pulse transit time, pulse arrival time, PPG/piezo amplitude and timing, machine learning-enhanced models). While hardware is simple, these methods are confounded by autonomic and vasomotor changes, require frequent cuff calibration, and currently lack clinical-grade accuracy. Intermittent oscillometric wearables (e.g., Omron HeartGuide) and smartphone finger-press methods can provide BP values but require user interaction and are not continuous. Stationary finger-cuff devices using the vascular unloading technique (VUT, volume clamp) such as Finapres and CNAP provide continuous BP waveforms. They address vasomotor influences via transfer functions and/or calibration (CNAP uses upper arm cuff for heart-level correction and calibration), but existing VUT systems are bulky due to pumps/valves/hoses for fast pulsatile control. It was assumed full vascular unloading with constant finger volume was essential for direct BP measurement, a premise the current work re-examines by proposing a slower actuator approach (VCT) targeting mean BP while reconstructing the pulsatile waveform from PPG.

Concept and sensing: CNAP2GO uses a photoplethysmography (PPG) sensor (LED at ~890 nm and photodiode) integrated with an actuator that varies the contact pressure p(t) of the light elements on the finger. The PPG signal v(t) inversely reflects arterial volume changes. CNAP2GO employs a closed-loop control based on filtered PPG components (e.g., v_rhythm, v_VCT, v_pulse) to adjust p(t) and track mean arterial pressure (mBP) via a volume control technique (VCT). The actuator moves slowly, at rates sufficient to follow mBP changes but not beat-to-beat pulsations. Initial calibration (setpoint finding): An open-loop oscillometric phase varies contact pressure while measuring PPG pulse amplitudes to generate an oscillometric curve. A Gaussian-like envelope fitted to the PPG pulse amplitudes identifies mBP at the maximum oscillation (maximum oscillation rule). Systolic (sBP) and diastolic (dBP) can be derived from the envelope; these initial values can be obtained from finger oscillometry (with heart-level correction and transfer function) or external brachial values. After this, the controller sets contact pressure to the mBP setpoint and closes the loop. Continuous tracking and control: CNAP2GO maintains balanced arterial volume over each beat (integral of pulsatile volume ~0). Deviations due to mBP shifts or vasomotor changes (vasoconstriction/vasodilation alter the p–v transfer function and pulse shape) are corrected by a proportional–integral (and rhythm PID) control that updates the setpoint beat-to-beat to re-balance volume. Anti-resonance elements mitigate control loop oscillations in frequency bands ~0.2–1 Hz. The reconstructed continuous BP waveform is computed as P_CG(t) = p_C(t) + k·v_pulse(t), where p_C(t) tracks mBP and v_pulse(t) captures pulsatility; calibration scales k using initial sBP/dBP. Prototype implementation: The team implemented CNAP2GO as C++ firmware on a CE-marked CNAP Monitor HD platform, repurposing its hardware while limiting pressure-change speed to emulate slow VCT. The CNAP sphygmomanometer provided calibration. Supporting functions (ambient light removal, beat detection, data transmission) remained as in CNAP Monitor HD. Validation experiments: 1) Laboratory tests in 20 healthy subjects compared mBP from CNAP2GO vs the standard CNAP Monitor HD (VUT) during maneuvers provoking BP and vasomotor changes (deep/fast breathing, cold pressor, Stroop, PLR, Valsalva). 2) Clinical prospective method comparison in 46 neurosurgical patients under general anesthesia: CNAP2GO mBP was compared to invasive radial arterial catheter mBP over ~45 minutes (synchronized and averaged over 1 s). Bland–Altman and concordance analyses assessed agreement and tracking of changes. Wearable actuator prototype: The authors developed a wearable concept using a fluid-filled bladder in a finger-ring housing the PPG elements; a small motor/plunger compresses the fluid to set contact pressure. Tests in 7 subjects showed the prototype produces correct oscillometric curves comparable to standard CNAP. Engineering analyses determined actuator requirements (max speed ~25–30 mmHg/s; average ~1.4 mmHg/beat), heart-level correction using inertial sensors (accelerometer/gyroscope), power budget (~58.7 mW average enabling 24 h with common batteries), and operational modes: measurement mode (initial oscillometry + VCT) and interpolation mode (reduced pressure, model-based estimation between measurement phases).

• Lab comparison (20 healthy subjects): Mean difference between CNAP2GO and CNAP Monitor HD mBP was approximately 0.3 ± 4.1 mmHg in text; Table 1 reports mean of differences 0.3 ± 4.4 mmHg with 95% limits of agreement −8.3 to 8.9 mmHg (n = 400 data points).
• Clinical study (46 surgical patients, ~45 min recordings): mBP ranged 45–113 mmHg, covering hypo- and hypertensive phases. Bland–Altman analysis showed bias −1.0 ± 7.0 mmHg with 95% limits of agreement −14.8 to 12.7 mmHg (n = 11,803 data points). Patient-specific offsets remained stable over time; intra-subject SDs ranged 1.1–6.3 mmHg.
• Concordance of changes: High concordance between spontaneous mBP changes detected by CNAP2GO and invasive reference within 5-minute windows (n = 10,390 changes; qualitative description provided).
• Actuator dynamics: Required median contact pressure change ~1.4 mmHg per beat; maximum up to ~24.4 mmHg per beat. Spectral analysis showed CNAP2GO’s contact pressure band-limited well below full pulsatile BP content, consistent with slow VCT control.
• Wearable feasibility: Fluid-filled bladder finger-ring prototype produced correct oscillometric envelopes across 7 subjects, comparable to standard CNAP. Engineering estimates indicate a small piezo stepper motor with >20:1 gearing can achieve up to ~30 mmHg/s pressure change, low power hold (high stall torque), and 24 h operation at ~58.7 mW total system power (actuator ~45.2 mW, PPG ~5.6 mW, BLE MCU ~1.7 mW, motion sensor ~6.2 mW) with common batteries.
• Additional outputs: The reconstructed pulsatile waveform enables derivation of cardiac output and other hemodynamic variables; sample traces demonstrate CO tracking comparable to CNAP Monitor HD.

The findings demonstrate that CNAP2GO, using a slow actuator volume control technique, can directly measure mBP continuously and reconstruct the pulsatile BP waveform with performance comparable to clinically established systems and close agreement to invasive arterial measurements. By controlling mean contact pressure to balance arterial volume over each beat and compensating for vasomotor-induced transfer function shifts, CNAP2GO mitigates a key limitation of time-based cuff-less methods that require frequent recalibration. The method self-calibrates via an initial oscillometric envelope, can remain stable over time (patient-specific offsets), and captures dynamic BP changes with high concordance. The feasibility of a ring-form factor with a fluid-filled bladder and low-power actuator suggests practical wearable implementation. The available waveform enables advanced hemodynamic monitoring (e.g., cardiac output) beyond intermittent BP values, potentially expanding continuous monitoring to general wards and ambulatory settings with minimal user interaction.

CNAP2GO introduces a novel continuous noninvasive BP sensing paradigm suitable for miniaturization: a slow-actuated volume control technique that directly tracks mean arterial pressure and reconstructs the pulsatile waveform from PPG, robust to vasomotor changes and capable of self-calibration. The software prototype showed strong agreement with both a standard CNAP system in lab tests and with invasive arterial pressure in surgical patients. A wearable finger-ring prototype using a fluid-filled bladder validated the ability to generate correct oscillometric signals, and engineering analyses indicate that actuator speed and power requirements are compatible with 24-hour wearable use. Future work should focus on full miniaturization into a ring device, robust heart-level (hydrostatic) correction using inertial sensors, enhanced motion robustness during vigorous activity, further validation across diverse clinical populations and environments, and refinement of measurement–interpolation modes for user comfort while maintaining accuracy.

• Validation was not performed on the final miniaturized wearable hardware; the clinical study used a software prototype on CNAP Monitor HD hardware and post-hoc invasive calibration.
• Heart-level (orthostatic) correction using motion sensors is proposed but not yet fully implemented and validated during movement.
• Motion robustness was demonstrated in lab settings with brief artifacts; performance during vigorous activities (e.g., sports) remains untested.
• Continuous external finger pressure is required during measurement phases (VCT), which may affect comfort; the interpolation mode is proposed but needs further validation for accuracy and user acceptance.
• Clinical validation was limited to 46 neurosurgical patients under general anesthesia; broader populations and settings require evaluation.
• Finger oscillometric calibration algorithms and transfer functions for finger-to-brachial level require development and validation for the final wearable implementation.