A flexible optoacoustic blood ‘stethoscope’ for noninvasive multiparametric cardiovascular monitoring

Continuous, noninvasive monitoring of clinically significant physiological signatures from deeply embedded blood vessels remains challenging due to the skin barrier and limitations of existing flexible sensors. Blood carries abundant clinical information for early detection and management of cardiovascular diseases, including oxygen saturation, pharmacokinetics of exogenous agents, and hemodynamics. However, current optical, thermal, ultrasonic vibration, and electrochemical approaches either lack depth specificity, only access superficial vasculature, or infer blood state indirectly. This study proposes a flexible optoacoustic blood ‘stethoscope’ (OBS) to noninvasively locate deep vessels and continuously monitor multiparametric blood biomarkers (e.g., hypoxia, exogenous agent decay) and hemodynamic functions (e.g., venous distensibility, arterial endothelial function), with 3D visualization, to address unmet needs in cardiovascular diagnosis and prediction.

Prior work on flexible sensors has enabled on-skin monitoring of temperature, electrophysiology, and biofluids but cannot resolve deep vessels or provide direct, distributed blood properties. Optical sensors (e.g., PPG, spectroscopy) provide oxygen saturation, glucose, blood pressure, and flow but are limited by photon diffusion and cannot resolve specific deep vessels. Thermal transport-based flow sensing accesses only superficial vasculature. Flexible ultrasound/vibration sensors can measure deep hemodynamics at single-vessel resolution but are insensitive to biochemical biomarkers. Electrochemical methods infer blood parameters from sweat/tears/saliva, which only partially reflect blood composition. Conventional optoacoustic imaging offers deep penetration and spectral contrast but relies on bulky, rigid probes and complex setups. These limitations motivate a flexible OA approach that combines depth specificity, biochemical sensitivity, and wearability.

Device design: The OBS integrates 25 optical nodes and 36 acoustic nodes in a multilayer flexible stack. Optical delivery uses a PDMS micro-lens array (5×5 hemispherical lenses) with nested pre-drilled holes to focus and deliver light into tissue. Acoustic reception uses a 110 µm PVDF piezoelectric film with patterned 7 µm Ag electrodes and insulating layers; PVDF is chosen for softness, low acoustic impedance, and high piezo-stress constant. The cross-distributed design (each optical node surrounded by four acoustic nodes) improves light-acoustic coupling. Bespoke side-firing fiber bundles deliver pulsed laser light without compromising flexibility. Materials: PDMS provides high transparency (400–1100 nm, >90%) and easy molding; Ag electrodes minimize resistance versus ITO and reflect stray light to reduce heating. Electrical interconnects use low-resistance silver paste. Optical element characterization: Monte Carlo simulations compared micro-lens delivery with side illumination and center window schemes in dermis-like media (µs′≈357 cm⁻¹, µa≈0.46 cm⁻¹ at 532 nm). The micro-lens method achieved higher superficial uniformity and a larger effective illumination area at 1 mm depth across 532–1064 nm (effective area >90% at 1 mm), obviating a pre-diffusion layer and preserving thinness. Experiments with 633 nm light and thin diffusion layers validated uniform transmission under flat and bent conditions; multi-spectral capacity verified via phantom absorption tests. Acoustic element characterization: PVDF elements showed mean resonance/antiresonance at 7.64/8.01 MHz (s.d. ~39 kHz). Center frequency ~7.8 MHz; ~78% of sensors had >100% bandwidth. Estimated axial resolution ~200 µm (c=1500 m/s); lateral ~1200 µm post-reconstruction. Simulations and phantom tests showed uniform AFOV at shallow depths due to small inter-element gaps (0.15 mm) and minimized grating lobes; AFOV uniformity improves with depth; bending tests (radii 40, 30, 20 mm) showed stable spectral response. Durability confirmed after 24,000 laser pulses over 8 days. 3D imaging algorithm (PS-NUFFT): A phase-shift non-uniform FFT reconstruction decomposes wave components and extrapolates wavefields through layered, heterogeneous media, accommodating refraction and curved sensor geometry with pre-migration. Compared to time-domain Delay-and-Sum (DAS), PS-NUFFT improves resolution, CNR, and GCNR at substantially lower computational cost (O(L n^3 log n^3) vs O(n^7)). Numerical phantoms showed 0.4 mm resolution (vs 1.5 mm DAS), CNR 14.1 dB (vs 9.8 dB), GCNR 0.73 (vs 0.57), and full 3D reconstruction (~256×256×300) in ~2 s on a laptop vs ~600 s for DAS. Experimental platform: Front-end OBS with flexible two-stage preamplifier adhered to skin; pulsed laser (532 nm; fluence ≤20 mJ/cm²) via side-firing fiber bundle; signals routed through multiplexers, LNA, oscilloscope/DAQ, and reconstructed at 10 Hz frame rate (limited by PRF). FPGA control coordinated acquisition; magnetic tracking registered skin surface profile for accurate node positioning. Animal studies: Nude mice under anesthesia underwent alternating normoxia (21% O2) and hypoxia (5.25% O2) cycles for oxygenation tracking; OBS attached to abdomen; pulse oximeter on paw for arterial SO2 reference. For exogenous agent monitoring, mice received IV Rhodamine B (200 µL, 2 mg/mL); after ~1 h stabilization, OBS continuously monitored abdomen to track decay; concurrent blood sampling every 5 min enabled in vitro fluorescence for comparison. Human trials: Three healthy volunteers participated. Venous tests: OBS on dorsum of hand; upper-arm cuff occlusion and step releases (110→0 mmHg) to quantify venous volume responses and compliance; ultrasound imaging cross-validated vessel area. Arterial tests: OBS on wrist over radial artery; occlusion (160 mmHg, 2 min) then release for perfusion and FMD; OA trends compared to ultrasound-derived diameter changes.

• Device performance: Optical micro-lens array achieved >90% effective illumination area at 1 mm depth across 532–1064 nm; no pre-diffusion layer required. Acoustic sensors: resonance/antiresonance 7.64/8.01 MHz; center frequency ~7.8 MHz; ~78% bandwidth >100%; axial resolution ~200 µm; lateral ~1.2 mm; robust under bending and prolonged laser exposure.
• Reconstruction algorithm: PS-NUFFT outperformed DAS with 0.4 mm vs 1.5 mm resolution (−6 dB FWHM), CNR 14.1 dB vs 9.8 dB, GCNR 0.73 vs 0.57, and ~2 s for full 3D reconstruction vs ~600 s for DAS on comparable data volume.
• Hypoxia monitoring (mice): OA amplitudes strongly correlated with arterial oxygen saturation (R²=0.99); every 50% decrease in SO2 produced ~38 mV OA amplitude reduction. OBS distinguished arterial vs venous dynamics, with venous response delays relative to arterial during hypoxia cycles; 3D maps visualized spatiotemporal oxygenation changes.
• Exogenous agent (Rhodamine B) decay (mice): OBS measured blood OA signal decay of 28% at 2000 s; tissue signal decreased ~10% over same period. Blood decay followed exponential with coefficient −3.6×10⁻4 s⁻¹ (R²=94.3%), comparable to in vitro fluorescence decay −3.2×10⁻4 s⁻¹ (R²=88.4%); OBS showed higher R² and RMS performance; 3D snapshots visualized agent redistribution between blood and tissue.
• Venous compliance (humans): During occlusion, venous OA amplitudes increased by ~23% (blood pooling), dropping upon release; OA amplitude ratios at cuff pressures: 102% (50 mmHg), 108% (80 mmHg), 114% (110 mmHg). Estimated venous compliance ΔR/ΔP ≈ −0.196%/mmHg (50–80 mmHg) and 0.185%/mmHg (80–110 mmHg). OA-derived trends matched ultrasound-derived volume ratios with reduced errors.
• Arterial perfusion and FMD (humans): Occlusion reduced arterial OA amplitude by ~37%; post-release hyperemia produced transient overshoot. FMD within 60 s post-release showed relative OA signal peaks of +16.3%, +15.5%, +9.3%, corresponding to estimated diameter increases of ~7.8%, 7.2%, 4.6%, consistent with ultrasound (8.0%, 7.1%, 5.6%). Typical clinical FMD ≈6.7% in healthy individuals, underscoring clinical relevance.

The OBS directly addresses the need for noninvasive, continuous, and multiparametric monitoring of deep-vessel blood properties. By combining spectral optical specificity with acoustic depth resolution in a flexible, skin-conformal form, it overcomes limitations of purely optical, thermal, ultrasonic vibration, and electrochemical sensors that either lack depth specificity or biomarker sensitivity. The tailored PS-NUFFT algorithm enables accurate 3D reconstructions in heterogeneous, curved skin environments with superior resolution and speed, facilitating real-time visualization and quantification. Measured relationships between OA amplitude and oxygen saturation (R²=0.99) validate the system’s capacity to monitor hypoxia and differentiate arterial vs venous dynamics. The ability to track intravascular exogenous agent decay noninvasively and continuously, with decay constants matching in vitro assays, demonstrates utility for pharmacokinetics. Venous compliance and arterial endothelial function (via FMD) measurements align closely with ultrasound references while using a simpler, wearable setup, showing promise for longitudinal cardiovascular risk assessment and disease prediction. Using relative percentage changes rather than absolute amplitudes mitigates inter-individual variability (e.g., skin tone, coupling). Overall, OBS expands flexible electronics from surface 1D sensing to volumetric 3D imaging of blood physiology.

This work introduces a lightweight, flexible optoacoustic blood ‘stethoscope’ capable of noninvasive, continuous, 3D, multiparametric cardiovascular monitoring, including hypoxia, pharmacokinetics of exogenous agents, venous compliance, and arterial FMD. A micro-lens optical delivery with PVDF acoustic reception provides efficient light-acoustic coupling and high sensitivity, while a PS-NUFFT reconstruction yields fast, high-resolution 3D imaging in heterogeneous media. Animal and human studies validated accuracy against gold standards (oximetry, ultrasound). Future efforts will pursue higher-resolution arrays with denser optical/acoustic elements, integration of LED sources, miniaturized and wireless acquisition, and broader clinical validation. Multi-wavelength operation could enable additional biomarkers (e.g., pH, glucose, temperature) and therapy monitoring (e.g., tumors), facilitating distributed, real-time cardiovascular healthcare.

Current limitations include: (1) imaging resolution constrained by array density; higher-density optical/acoustic elements are needed for finer voxels; (2) reliance on a pulsed laser with a non-negligible driver size; front-end remains tethered to wired acquisition for post-processing; (3) absolute OA amplitudes are susceptible to in vivo factors (e.g., skin tone, coupling), necessitating self-referenced relative changes; (4) residual bending-induced distortions require accurate surface registration and algorithmic compensation; (5) broader clinical validation across diverse populations and conditions is needed before widespread deployment.