Open‐chest cardiac ultrasound‐mediated imaging with a vacuum coupler

Abstract

Background

A fundamental obstacle for the preclinical development of ultrasound‐(US) mediated cardiac imaging remains cardiac motion, which limits interframe correlation during extended acquisition periods.

Purpose

To address this need, we present the design and implementation of a 3D‐printed vacuum coupler that stabilizes a US transducer on the epicardial surface of the heart for feasibility assessment and development of advanced, cardiac, US‐mediated imaging approaches.

Methods

The vacuum coupler was 3D printed with biocompatible resins and secured with a standard intraoperative suction aspirator. US‐mediated imaging (i.e., B‐mode and photoacoustic [PA] imaging) was performed in an open‐chest porcine model with and without the vacuum coupler. Based on inter‐frame displacement tracking and cross‐correlation (CC) coefficients, changes in frame motion and stability were compared for each imaging mode/configuration through a prolonged (∼1 min) acquisition, while the impact on PA‐based SO₂ accuracy was assessed.

Results

When compared to conventional handheld imaging, stand‐off imaging, and coupler without suction, epicardial imaging with the vacuum coupler and suction applied led to a significantly reduced mean axial displacement of 0.15 mm versus 0.89, 0.49, & 0.49 mm, respectively (p‐values ≤ 8.65e‐7). Comparing the coupler without suction to that with suction applied, physiologically unrealistic SO₂ estimates reduced from 1.72 to 0.81%, respectively, and lateral interframe displacement reduced from 4.58 to 2.01 mm, respectively (p‐value = 5.07e‐23). Overall, reduced cardiac tissue motion and increased interframe CC coefficient (baseline = 0.43 vs. coupler with suction = 0.80) allow for more accurate PA unmixing.

Conclusions

Epicardial US‐mediated imaging with a vacuum coupler reduces cardiac motion artifact, providing a consistent sampling of an intended region of interest (ROI) over multiple cardiac cycles. This could help facilitate the development of advanced US‐mediated imaging, which is often hindered by cardiac motion. Stable implementation of these imaging techniques could allow for intra‐operative assessments of local cardiac perfusion as well as tissue characterization.

1. INTRODUCTION

Ultrasound (US) and US‐mediated imaging are valuable tools for the clinical diagnosis of cardiovascular diseases because they afford real‐time, point‐of‐care, nonionizing, cost‐effective imaging. ¹ , ² The ability to use US‐mediated imaging to diagnose and monitor cardiac diseases, such as ischemic heart disease, ³ early graft failure, ⁴ valvular disease, ⁵ and myocardial infarction, ⁶ is crucial for patient care. Although noninvasive transthoracic echocardiography is standard, invasive echocardiography is also used clinically as an intracardiac catheter‐based platform for guiding interventional electrophysiologic cardiac procedures, ⁷ a transesophageal approach, ⁸ or for direct epicardial imaging through an “open‐chest” preparation for coronary vessel graft assessment ⁹ and valve‐replacement. ¹⁰

Advanced US‐mediated imaging approaches, such as strain elastography, ¹¹ shear wave elasticity imaging (SWEI), ¹² photoacoustic‐ultrasound (PA‐US) imaging, ¹³ and super‐resolution ultrasound (SRUS) imaging ¹⁴ have shown significant potential for cardiac diagnostics and therapy monitoring. However, such imaging techniques tend to have relatively long acquisition times because of the added dimensionality (e.g., spectral for PA‐US imaging or temporal dynamics for SRUS or SWEI) of their required data sets. For instance, because of the relatively low repetition rate (e.g., tens of Hz) of many PA laser sources and the need for frame averaging to improve the signal‐to‐noise ratio (SNR) of imaging data, the acquisition time to obtain one spectrally unmixed PA imaging frame commonly exceeds 1 s, ¹⁵ which is often longer than a full cardiac cycle. Thus, preclinical development in translationally relevant large‐animal models of such approaches can be hampered by the significant cardiac motion experienced during conventional transthoracic or open‐chest imaging.

Regardless of the type of cardiac US imaging method used (e.g., transthoracic or open‐chest), cardiac motion can generate artifacts that are difficult to eliminate and that can cause significant inaccuracies in derived imaging metrics. In healthy subjects, early‐diastolic myocardial wall velocities can exceed 100 mm/s, ¹⁶ which can result in significant decorrelation of successive US imaging frames. ¹⁷ , ¹⁸ Specifically for US‐mediated imaging, cardiac motion has been shown to affect the accuracy of Doppler ¹⁹ and SWEI ¹² imaging, resulting in “stitching” artifacts with 3D acquisitions ⁸ or causing inaccuracies in spectral unmixing (e.g., local SO₂ estimation) of multiwavelength PA‐US. ²⁰ , ²¹ While advancements in high‐frame‐rate imaging, ²² acquisition gating (relative to the cardiac cycle), and postprocessing‐based motion correction ²³ have helped mitigate such motion artifacts, further reduction of relative cardiac motion during open‐chest procedures could be critical in the preclinical (i.e., large‐animal model) development of advanced US‐mediated imaging approaches (particularly for those requiring prolonged acquisition times) and in the optimal implementation of clinical epicardial imaging procedures.

Herein, we show that applying an external vacuum adapter that forms stable coupling between the transducer and myocardial surface significantly reduces motion artifact, particularly for modes requiring increased acquisition times (e.g., multiwavelength PA‐US imaging). The concept of a vacuum coupler to aid US‐mediated epicardial imaging was motivated by previous work that implemented a wholly machine‐milled coupler. ¹² In this work we demonstrate a 3D‐printed coupling platform that allows for PA‐US imaging and permits easy adaptation to accommodate a variety of transducers. In providing vacuum‐mediated adhesion, we mitigate transducer shift without disrupting tissue integrity or causing damage to proximal myocardial vessels, which could result in local ischemia or intractable bleeding that can lead to hemodynamic compromise. We apply PA‐US imaging in an in vivo porcine model with and without the vacuum coupler to assess differences in intraframe motion and decorrelation and spectral unmixing accuracy.

2. MATERIALS AND METHODS

2.1. Vacuum‐coupler design

The coupler was designed such that the housing would fit the shape of a PA‐US transducer (LZ‐250, FUJIFILM VisualSonics Inc., Toronto, Canada). The dimensions of each coupler component can be found in the blueprints presented in Figure S1. To create a stable connection on the epicardial surface of the heart, vacuum suction was incorporated using a standard intraoperative suction aspirator, accessible in most operating room environments. A perforated, flexible ovular ring was designed to provide suction adhesion between the probe and myocardium (Figure 1a) while affording an unobstructed central region for US transmission/reception and PA light transmission. The flexible mating ring has two rows of differently shaped holes (oval and circular) that provide access to four independent compression chambers that can be connected via Luer lock to an external vacuum (Figure 1b). This design ensures a stable connection even if occlusion occurs in one or more of the independent suction networks or if a region of the mating ring fails to maintain full contact with the epicardial surface. The flexible coupling surface inserts into a rigid mounting face that is attached to two translational shafts, one of which is threaded, to allow for transducer height adjustment (relative to the coupling ring) through a hand‐controlled gear wheel (black arrow in Figure 1c,d). This feature allows for the use of varied thickness stand‐offs to control epicardial surface depth for a particular imaging application (e.g., optimized for PA light delivery). The entire vacuum coupler is composed of biocompatible resin (Figure 1d).

FIGURE 1.

Photo of 3D‐printed coupling ring with flanking suction holes and vacuum‐attachment ports (a). CAD schematic of the perforated, flexible coupling ring (b). 3D rendering of the assembled vacuum coupler (c). Photo of the assembled vacuum coupler with inserted PA‐US imaging probe (d). CAD, computer‐aided design; PA‐US, photoacoustic‐ultrasound.

2.2. Vacuum‐coupler fabrication

Three‐dimensional modeling was performed with SolidWorks 2022 SP 4 computer‐aided design (CAD) software (Dassault Systèmes, Vélizy‐Villacoublay, France). These STL files were then imported into PreForm software (V.3.28.0, Formlabs Inc.) to render each component for 3D printing, which was done with a Formlabs 3B+ 3D printer (Formlabs Inc., Somerville, MA). Components were printed using BioMed resin (Ref# RS‐F2‐BMAM‐01, Formlabs Inc.) for the rigid parts (0.050‐mm vertical resolution) or Elastic 50A resin (Ref# RS‐F2‐ELCL‐01, Formlabs Inc.) for the flexible sections (0.100‐mm vertical resolution). Both resins are certified to be biocompatible. The rigid components can withstand autoclaving (120°C), while the flexible parts are sterilized using isopropyl alcohol (IPA, 99%). Connections were made using 3D‐printed M6 threaded screws and gears to minimize metal components and ensure easy replacement if needed. Once printed, parts were removed using clamps, and print supports were detached. Parts were washed manually in 99% IPA (Cat. #LC157504, Labchem Inc., Zelienople, Pennsylvania) and then sonicated for 15 min in a microsonicator bath containing IPA to remove any uncured resin from the component surfaces. Parts were then dried using compressed air. Lastly, each component was cured in an ultraviolet Form Cure oven (Formlabs Inc., Somerville, MA) at 70°C for 60 min, according to the manufacturer's protocol for the BioMed resin.

2.3. In vivo coupler testing

Imaging was performed on a combined PA‐US imaging system that consisted of a Vevo 2100 US imaging system (FUJIFILM VisualSonics Inc., Toronto, Canada) paired with an LZ‐250 transducer (21‐MHz center frequency; 256‐element linear array; 13‐24‐MHz bandwidth; 15‐mm elevational focus) with integrated fiber optics connected to a pulsed, tunable Nd:YAG laser (680–970 nm). PA‐US imaging data were acquired at the following wavelengths: 750, 800, and 850 nm. Wavelengths were alternated sequentially without frame averaging; PA frame acquisitions for a single wavelength took ∼200 ms. All experiments were approved for study (protocol no. IS00006601 approved January 25, 2022) by the Institutional Animal Care and Use Committee (IACUC) at the Houston Methodist Research Institute and were performed according to the principles of the NIH Guide for the Care and Use of Laboratory Animals, the provisions of the Animal Welfare Act, PHS Animal Welfare Policy, and the policies of the Houston Methodist Research Institute.

Imaging was performed on female Domestic Swine (Sus domesticus; 81–100 lbs; Oak Hill Genetics, Ewing, IL); heart rates of 90–100 bpm were maintained during imaging. Animal sedation occurred via intramuscular injection of a cocktail solution of 20 mg/kg ketamine, 0.5 mg/kg midazolam, 0.1 mg/kg hydromorphone, and intravenous administration of 0.04 mg/kg atropine. Animals were intubated and maintained with isoflurane via mechanical ventilation and monitored throughout the experiment. A lidocaine infusion (0.5–1 mg/min) was administered throughout the procedure for arrhythmia suppression, while norepinephrine (1–4 mcg/min) was given as warranted for blood pressure support. To prepare the animal for open‐chest imaging, a 7–9″ midline incision was made on the chest wall at the level of the xiphoid process extending cranially along the sternum. The mediastinum was then exposed, and a pericardial sling was created, exposing the ventricles. Once the heart was exposed, a thin layer of US gel was spread on the epicardial surface to help maintain acoustic coupling, and the US probe was placed on the epicardial surface of the mid‐to‐apical left ventricle. For baseline imaging (Figure 2a), a bottomless plastic beaker filled with degassed water and plastic wrap secured at the base (i.e., to provide an acoustic window) was utilized to maintain acoustic coupling. For this acquisition type, the transducer was placed inside the beaker and held fixed throughout imaging, while the water‐coupling beaker was allowed to move with the epicardial surface (i.e., resulting in variable imaging depth through the cardiac cycle). This acquisition was followed by imaging with a rigid Aquaflex stand‐off (Parker Labs Inc., Fairfield, NJ; Figure 2b), which maintained a relatively fixed imaging depth of ∼12 mm. Lastly, the PA‐US transducer was placed into the vacuum coupler, and imaging was performed with and without active suction. To maintain acoustic coupling, an Aquaflex cutout was prepared to fit in the distal end of the vacuum coupler (i.e., in the flexible mating ring). For the latter three acquisition types, the transducer was held directly in the operator's hand to allow the transducer to move with the cardiac surface. Pressure from the intraoperative suction aspirator was set to the minimum needed to achieve durable adhesion to the epicardial surface.

FIGURE 2.

Comparison in a porcine model of baseline acquisition using a probe submerged in contained warm saline (a), probe with Aquaflex stand‐off (b), and probe inside vacuum coupler (c).

2.4. Image processing and displacement estimation

After acquiring in vivo imaging data, the epicardial surface in B‐mode images was segmented using Otsu's method ²⁴ to establish a threshold cut‐off based on a region within the nearfield stand‐off, which consistently presented with high contrast relative to the tissue surface. A rectilinear region of interest (ROI) centered laterally with a width of 17.6 mm and extending in‐depth approximately 4 mm from the segmented surface was defined for each B‐mode imaging frame to establish a kernel for cross‐correlation (CC) analysis. This kernel size was selected to ensure adequate PA signal through depth, while providing the widest possible kernel to allow for lateral motion within the 23‐mm‐wide field of view. Normalized, two‐dimensional, CC was calculated using each of these templates as a search kernel within each of the other frames in the acquisition, producing an N‐by‐N matrix of CC coefficients. Inter‐kernel overlap was maximized by allowing for single‐pixel shifts (nominal pixel size: 0.02 [axial] × 0.09 mm [lateral]) in each dimension. The reference frame yielding the highest mean CC across all frames in an acquisition was then selected as the optimal reference frame and utilized for the CC coefficients reported for that acquisition. Images possessing a surface segmentation with a net slant greater than 3° (i.e., resulting from significant cardiac motion) were excluded as reference frames. To characterize axial displacement between approaches, the mean surface segmentation axial position was calculated for each frame. Lateral displacements for the two coupler acquisitions (i.e., with and without suction) were calculated based on the lateral ROI (as defined previously) shift needed to achieve the maximal CC coefficient for each frame (relative to the reference frame). Although this approach could be used to estimate axial displacement as well, the low CC coefficients presented by some of the non‐coupler acquisition frames (i.e., due to significant out‐of‐plane motion) made such tracking unreliable. Thus, the surface‐segmentation approach, which proved to be robust to decorrelation, was used for all acquisitions. To assess statistical significance between acquisitions of reported imaging metrics, two‐tailed t‐tests with a Bonferroni‐adjusted significance level of 7.14e‐3 were used to establish a 5% probability of a Type‐I error.

To compare and assess the effect of residual cardiac motion on multiwavelength PA data, SO₂ estimations were compared between the baseline and coupler with suction acquisitions. Frames for each acquisition during the entire capture period were resorted based on wavelength. Each acquisition then underwent pixel‐wise linear unmixing to estimate relative levels of oxygenated and deoxygenated hemoglobin based on their unique absorption spectra. An SO₂ image within a depth region 6 mm from the segmented surface of the reference B‐mode frame (i.e., to help ensure adequate SNR) was then calculated for each as a percentage ² , ²⁵ of the ratio of the unmixed oxyhemoglobin to total hemoglobin. For each acquisition, the percentage of pixels within the analyzed ROI yielding physiologically unrealistic SO₂ estimates, which was defined as > 99% or < 1% SO₂, was then calculated. ²⁶

3. RESULTS

Interframe CC coefficients were compared between baseline, stand‐off, coupler without suction, and coupler with suction acquisitions (Figure 3). Average CC coefficients among the acquisition groups were 0.43, 0.63, 0.75, and 0.80, respectively, while minimum CC coefficients were 0.31, 0.52, 0.68, and 0.76, respectively. There was a statistically significant improvement (p‐values ≤ 2.62e‐10) in CC coefficients when comparing coupler with suction acquisition to all other acquisitions. Figure 3 provides CC coefficients for all frames and for all acquisition types compared to the coupler with suction average.

FIGURE 3.

Plots of CC coefficients versus imaging time for the baseline (a—black), stand‐off (b—green), coupler without suction (c—blue), and coupler with suction (d—red). Dashed lines matched to an acquisition color indicate the average CC coefficient for respective acquisitions, while dashed red lines provide the average CC coefficient for the coupler with suction acquisition for comparison. CC, cross‐correlation.

Interframe axial displacements were compared between baseline, stand‐off, coupler without suction, and coupler with suction acquisitions (Figure 4a–d). Mean axial displacement was 0.89, 0.49, 0.49, and 0.15 mm, respectively, while maximum axial displacement was 5.23, 1.78, 1.00, and 0.26 mm, respectively. Additionally, interframe lateral displacement of the CC ROI was compared between the coupler without suction and with suction acquisitions (Figure 4e–f), with mean lateral displacements of 4.58 and 2.01 mm, respectively, and maximum lateral displacements of 8.22 and 3.52 mm, respectively. There was a statistically significant reduction in both axial (p‐values ≤ 8.65e‐7) and lateral (p‐value = 5.07e‐23) tissue displacement for the coupler with suction acquisitions compared to all other acquisition methods tested. During the use of the coupler, no adverse tissue reactions were observed or negative effects on image quality.

FIGURE 4.

Plots of axial displacement versus imaging time of cardiac surface for the baseline (a—black), stand‐off (b—green), coupler without suction (c—blue), and coupler with suction (d—red). Plots of lateral displacement versus imaging time within the CC ROI for coupler without suction (e—blue) and with suction (f—red) acquisitions. Dashed lines matched to an acquisition color indicate average displacement for respective acquisitions, while dashed red lines provide average displacement for the coupler with suction acquisition for comparison. CC, cross‐correlation; ROI, region of interest.

Figure 5 shows a comparison between B‐mode frames used as the CC reference and the frame with the lowest CC coefficient for the baseline and coupler with suction acquisitions. In the baseline acquisition, the two frames (Figure 5a,b) provide clearly different orientations of the epicardial surface due to the significant axial motion occurring due to the open water‐bath coupling. This results in significant interframe decorrelation and yields an unrealistic SO₂ image (i.e., with many pixels presenting at mathematical SO₂ limits, 0% and 100%) when these frames are spectrally unmixed (Figure 5c). When the coupler with suction is used, however, the lowest‐CC frame (Figure 5e) presents similarly to the reference frame (Figure 5d), yielding a spatially consistent and physiologically reasonable SO₂ when unmixed (Figure 5f). The percentage of pixels with physiologically unrealistic SO₂ estimates (i.e., > 99% or < 1% SO₂) were 4.12, 3.07, 1.72, and 0.81%, for the baseline, stand‐off, coupler without suction, and coupler with suction acquisitions, respectively.

FIGURE 5.

B‐mode images for reference (a,d) and lowest‐CC‐coefficient (b,e) frames and resulting unmixed SO₂ image (c,f) for baseline (top row) and coupler with suction (bottom row) acquisitions, respectively. CC, cross‐correlation.

4. DISCUSSION

This work showed that a vacuum coupler in an open‐chest heart model results in significantly less cardiac motion during US‐mediated imaging, as evidenced by significantly higher CC coefficients during prolonged acquisition in addition to reduced axial and lateral frame displacements. Additionally, the use of the vacuum coupler resulted in a drastic reduction in physiologically unrealistic SO₂ estimates–4.12% of pixels down to 0.81%—obtained from PA‐US imaging. Such a coupling device allows for longer acquisition times, which are necessary for many US‐mediated approaches, during open‐chest cardiac imaging.

There are multiple aspects of the suction coupler design that make it well‐suited for in vivo application, including the use of biocompatible resin and incorporation of a parallel network of suction channels/ports to allow for steady, circumferential adhesion even if some suction ports are occluded or do not maintain full surface contact. Additionally, the 3D‐printed fabrication of the coupler allows for use with multiple different models/types of imaging probes. For most linear arrays, adapting the holder design for use with other probes can be done simply by scaling the 3D CAD model. Furthermore, since the entire device is completely 3D printed and the resins used are commercially available, it can be fabricated in any laboratory with access to a suitable 3D printer.

Although the coupler did not cause any noticeable imaging artifacts or physiologic changes in the animal, it was possible to create a transient imprint (i.e., darker red color corresponding to the suction port holes) on the epicardial surface if the aspirator pressure was not set appropriately. Such a change in color suggests that tissue damage could occur if the suction pressure is too high. However, this effect could also be leveraged at the conclusion of a study whereby the pressure would be increased to then cause a durable coupler imprint on the tissue surface that could then be used to determine the imaging plane during histopathological assessment. Additionally, blood residing in the open‐chest cavity can diffuse into the Aquaflex stand‐off over time, which would introduce variable and wavelength‐dependent attenuation through the laser irradiation path, confounding spectral unmixing accuracy. As such, it is advisable to replace the stand‐off periodically during longer experiments to mitigate this effect.

Several limitations should be noted in the presented study. First, this device was designed to be utilized for direct epicardial imaging, which requires a complete thoracotomy as well as the generation of a pericardial sling, all of which can be considered highly invasive. However, future iterations of this concept could include integration such as coupling integration into an endoscopic transducer (e.g., for transesophageal or intracardiac imaging) for a less invasive implementation. Second, as these data were only acquired from one specific epicardial window (i.e., on the mid‐to‐apical left ventricle), windows on other aspects of the heart (e.g., on the septum) could present with different wall‐motion profiles, affecting the cyclic US decorrelation observed. Third, although the transducer was kept approximately normal to the epicardial surface, such positioning could be difficult to maintain manually for prolonged acquisitions, and thus some out‐of‐plane motion could be a result of inadvertent tilting of the transducer during the acquisition. Fourth, while the coupler can be scaled for other US probes, the design requires a flat array face (e.g., linear or phased) and does not currently allow for array curvature (e.g., curvilinear). Finally, the presented displacement analysis was performed using CC or surface segmentation and did not consider other methods, such as those including motion regularization based on energy‐function optimization, ²⁷ which will be pursued in future work.

In the future, the vacuum coupler presented in this work could help reduce cardiac motion during the preclinical development (e.g., open‐chest preparation in large‐animal models) of US‐mediated techniques used to characterize myocardial tissue by facilitating more reliable SWEI, which is sensitive to diastolic dysfunction, ²⁸ or more accurate multiwavelength PA‐US imaging, which is sensitive to myocardial ischemia ²⁹ or could be used to assess acute ablation lesion characteristics. ²⁵ , ³⁰ Further, open‐chest US imaging is currently done clinically to provide Doppler characterization of coronary artery bypass grafts, ⁹ , ¹⁰ and thus the reduction in motion and improvement in probe‐placement stability afforded by a suction coupler may provide more accurate and precise measurements for such clinical cardiac procedures.

5. CONCLUSIONS

Open‐chest cardiac US imaging using a vacuum coupler allowed for reduced tissue displacement and an increased number of frames that maintained a high correlation throughout multiple cardiac cycles. Such a coupling device facilitates the development of cardiac US‐mediated imaging strategies with prolonged acquisition times by improving interframe stability during open‐chest imaging procedures. Although the most imminent use of this platform is for improved preclinical (i.e., open‐chest, large‐animal models) development of advanced US‐mediated imaging approaches, there could also be limited application for certain clinical procedures (e.g., coronary bypass graft assessment), which are performed with an open‐chest preparation.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts to disclose.

Supporting information

Supporting Information

Mathuria N, Vishwanath K, Brero G, et al. Open‐chest cardiac ultrasound‐mediated imaging with a vacuum coupler. Med Phys. 2025;52:880–888. 10.1002/mp.17511

DATA AVAILABILITY STATEMENT

Data will be made available on request.