Synchronous wearable ultrasound for early detection of coronary and carotid artery comorbidity

Figures

• 

  Fig. 1.. Concept of synchronous wearable ultrasound for evaluation of concomitant coronary and carotid artery disease. (A) Schematics showing concomitant coronary and carotid artery disease (left), synchronous wearable ultrasound imaging (middle), capture of frame-by-frame ultrasound images under different physiological conditions (right), and extraction of clinically important dynamic metrics with AI algorithms for supporting diagnosis (right bottom). In this work, the developed system typically achieves an imaging frame rate of 15 to 20 Hz, with a temporal offset of ~2.5 ± 0.5 ms between acquired cardiac and carotid images within a single frame. (B) Exploded view of the developed wearable ultrasound patch, with the key components labeled on the right. (C) Photos of the patches secured to the skin by adhesive tapes and optical images of the cardiac and carotid patches. Panel (A) was created using C4D and Photoshop software.

• 

  Fig. 2.. Design and characterization of the wearable ultrasound patches. (A) Pulse-echo characteristics (left) showing the center frequency (~3 MHz) and −6-dB bandwidth (~1.8 MHz, ~60%) of the cardiac patch; phased-array beam steering with focused transmissions for cardiac imaging (right). (B) Pulse-echo characteristics of the carotid patch (left), showing the center frequency of ~4.5 MHz and bandwidth of ~63% (~2.84 MHz); phased-array beam steering with wide-beam transmissions for carotid imaging (right). (C) Cardiac ultrasound images captured in response to various body postures, demonstrating stable image acquisition. RV, right ventricle; IVS, interventricular septum; AO, aorta; LA, left atrium; LV, left ventricle; MV, mitral valve; LVPW, left ventricular anterior wall. (D) Ultrasound images of the carotid artery under different postures. The plane-wave compounding imaging method (top images) fails to visualize the artery during neck twisting, whereas phased-array imaging with wide-beam transmissions (bottom images) maintains full artery visibility. (E) Frame rate for dual-patch imaging. (F) Temporal offsets for collecting cardiac-carotid image data under various combinations of…

• 

  Fig. 3.. Transient heart motion and pulsation. (A) Pedaling power setup for collecting cardiac-carotid images under the physiological states from rest to exercise. The illustration of the pedaling figure at the left panel was created using C4D and Photoshop software. (B) M-mode cardiac and carotid ultrasound images under graded exercise intensity. (C) Extracted PR, HR, and RR by the STFT method. PRs measured from carotid M-mode ultrasound showed near-perfect agreement with HRs across all exercise intensities. (D) Pearson correlation coefficient and mean absolute difference between HR and PR curves for the five healthy participants (details are shown in fig. S28). The RR versus RR comparison illustrates the difference in RR when extracted via cardiac versus carotid images. The PR versus HR comparison illustrates the difference between PR and HR when extracted via carotid versus cardiac images. (E) Transient pulsation and heart motion waveforms show beat-to-beat motion.

• 

  Fig. 4.. Dynamic cardiac metrics. (A) Workflow for quantifying cardiac metrics through automatic image segmentation and reconstruction algorithms. (B) The schematic of the cardiac cycle illustrates coordinated periods of relaxation (diastole) and contraction (systole) in both the atria and ventricles, working together to fill the chambers with blood and then eject it. (C) Detailed volumetric waveforms showing EDV and ESV. (D) Volumetric change of the heart. The black curve depicts cyclic volumetric variations, while the red curve shows the overall volumetric trend derived from moving average smoothing. (E) Exercise intensity–dependent variations in LV internal diameter at diastole (LVIDd) and LV internal diameter at systole (LVIDs). (F) Cardiac metrics including EDV, ESV, SV, and EF. (G) Transient CO and HR. (H) Performance validation of the cardiac ultrasound patch against commercial ultrasound equipment (Philips, EPIQ 7C); see fig. S32 for details. Ultrasound images captured by the cardiac patch (middle left panel) and commercial equipment (middle right panel) and quantitative comparison of the measured EDV/ESV/EF/CO/SV metrics.

• 

  Fig. 5.. Dynamic carotid metrics. (A) Workflow for quantifying carotid metrics through automatic image segmentation and tracking algorithms. (B) Schematics of the pulsation cycle, showing vasoconstriction and vasodilation of carotid artery. (C) Quantitative evaluation of the arterial circularity, with the inset depicting the segmented carotid artery boundary. A circularity index of 1.0 indicates perfect geometric circularity, with decreasing values reflecting increasing deviation from an ideal circle. (D) Diameter variation of the carotid artery. (E) Exercise-induced dynamics of carotid artery geometric center displacement. (F) Blood pressure waveforms under graded exercise intensity. (G) Calibration and validation of the carotid blood pressure (BP) measurements.

• 

  Fig. 6.. Evaluation of the system for coronary and carotid artery comorbidity analysis. (A) Schematic of the human circulatory model. The model uses transient cardiac volume and HR as inputs to predict carotid and cardiac metrics. (B) Comparison of the blood pressure waveforms, SVs, and CO between modeling and experiment. In vivo measurements obtained from patients with confirmed (C) CAD, (D) CHD, and (E) CAD-CHD comorbidity (see table S4 for detailed symptoms). On the basis of their clinical records, CAD was modeled as 70% arterial stenosis, and CHD was modeled by reducing myocardial compliance to 70% of the normal value (Materials and Methods). Two modeling results are shown for each patient: a healthy model [left panels of (C) to (E)] and a model incorporating the pathological condition [right panels of (C) to (E)]. Results are evaluated using the NRMSE (normalized root mean square error), Pearson correlation, mean, and standard deviation. The anatomical illustrations in (A) and top panels of (C) to (E) were referenced from Servier Medical Art, licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/deed.en).