Post-Procedure Micro-CT Analyses of Coronary Artery Stenting in Left Main Vessels of Human Hearts

Background: Percutaneous coronary interventions (PCIs) within left main coronary arteries are high-risk procedures that require optimization of interactions between stent(s) and diseased vessels. Optical Coherence Tomography (OCT) is a widely accepted tool that enhances physicians’ ability to assess proper stent appositions during procedures. The primary aim of this study was to develop complementary post-procedure imaging methodologies to better assess and interpret outcomes of left main PCI procedures, utilizing reanimated and perfusion-xed human hearts. Methods: PCIs were performed while obtaining OCT scans within the left main anatomies of six human hearts. Following PCI, each heart was scanned with a micro-CT scanner with optimized parameters to achieve resolutions up to 20µm. Scans were reconstructed and imported into DICOM software to segment and generate models of implanted stents and coronary vessels. 2D images that were obtained during PCI from OCT were used to assess the accuracy of 3D models generated from micro-CT reconstructions. Additionally, the 3D models were utilized to create virtual reality scenes and enlarged 3D prints for “mixed reality” physician training and education relative to bifurcation stenting in human left main coronary arteries. Results: We developed a reproducible methodology for post-implant analyses of coronary artery stenting procedures, and created 3D models down to 20 microns of resolution for reanimated and perfusion-xed heart specimens. Conclusions: These models can be used to obtain detailed measurements for further clinical insight on procedural outcomes. 3D models resulting from these scans are useful for generating virtual reality scenes and 3D prints for physician training and education.


Background
Coronary artery disease is de ned as the narrowing of coronary vessels caused by buildup of sclerotic fatty deposits, otherwise known as plaque. In cases of right dominant coronary circulation, it has been estimated that the left main (LM) coronary artery can supply >75% of the left ventricular myocardium 1 . It has also been reported that patients with LM stenoses ranging between 50-70% had a projected 3-year survival of 66%, whereas patients presenting with stenoses >70% only had a ~40% 3-year survival rate 2 .
Thus, in patients with signi cant LM disease, percutaneous coronary interventions (PCIs) with drugeluting stents have been increasingly recognized as valid clinical revascularization procedures 3 . Because of such clinical severity, proper deployment of stents in the LM and resulting strut appositions play a critical role in patient outcomes.
Recently, intra-procedural coronary imaging technologies have been developed and optimized to allow interventional cardiologists the ability to view relative plaque depositions, lumen dissections, and/or stent appositions during various PCI procedures. A recent advancement has been the application of Optical Coherence Tomography (OCT) for such interventions. OCT uses a rotating glass ber-optic system that tracks over a 0.014" guidewire until it is positioned in the desired location within the patient's coronary anatomy. Once placed, a contrast ush is introduced into the OCT catheter, removing blood from the artery to allow for appropriate optical scanning; coherent infrared light is then directed and re ected within the tissue to create detailed images 4 . While OCT, in combination with uoroscopy, continues to be a valuable tool in left main PCI cases, it is not universally available and is often reserved for the most complex cases.
For more than two decades, the Visible Heart ® Laboratories have been dedicated to translational research and advancements in cardiac research, devices, and education. We routinely perform PCI procedures, in bifurcations speci cally, on reanimated large mammalian hearts 5 . On rare occasions, we have the opportunity to conduct PCIs within reanimated human hearts 6 . The Visible Heart ® Laboratories provide a unique research platform where PCIs can be performed using experimental devices, novel or approved procedural techniques, and/or multimodal imaging modalities, all with no risk to living patients. The primary aim of this study was to further develop complementary post-procedure imaging methodologies to better assess and interpret the outcomes of left main PCI procedures, utilizing both reanimated and perfusion-xed human hearts.

Results
We successfully performed PCIs in six human LM coronary arteries employing Visible Heart ® methodologies, in both reanimated and perfusion-xed hearts. Both experimental approaches were deemed feasible and valuable for conducting preclinical research. We acquired OCT images immediately after each stenting procedure, then micro-CT scanned each perfusion-xed heart using the same scanning parameters. The 2D OCT images were compared side by side with the 3D micro-CT reconstructions, and each micro-CT reconstruction accurately resembled its OCT counterpart as shown in The three reanimated specimens were relatively healthy, and no signi cant calci cation was observed through endoscopic visualizations, OCT imaging, or micro-CT-generated computational models. However, varying degrees of vascular calci cations were found in the three perfusion-xed specimens, as we intentionally selected hearts with histories of coronary artery disease. This was most notable in HH115 which elicited large amounts of plaque buildup; it was not possible to fully expand the deployed stent into a desired cylindrical shape via multiple balloon expansions employing >20 atmosphere of balloon pressures. Despite the presence of vascular disease and previous perfusion xation in formalin, these hearts still elicited signi cant increases, >60%, in LM cross-sectional areas after stent implantations (Table 1). These values were calculated by measuring the cross-sectional areas of the given vessel lumens at 5mm distances from the determined coronary ostia border, determined before and after PCI ( Fig. 2).
In this study, we also sought to assess the relative effects of the formalin xation process on a heart's vessel geometries. This was achieved by taking OCT measurements post-PCI, while the heart specimen was <12 hours post-recovery (viable -fresh), to examine the LM lumen areas 2mm distal to the deployed stents. We obtained additional OCT scans of the same specimens after formalin xation to evaluate lumen diameter changes induced by the xation process. Additionally, the same lumen areas postxation were measured via micro-CT to enable comparisons between measuring modalities. Reviewing the lumen areas in Table 2, we observed larger tissue desecration after formalin xation in the healthy heart specimens, compared to those with previously noted heart disease. In HH202, an increase in lumen area-from 5.04mm 2 to 5.22mm 2 (OCT) and 6.78mm 2 to 6.94mm 2 (micro-CT)-was observed; this may have resulted from residual expansion from the implanted stent. Some discrepancies in measurements between OCT and micro-CT were observed (Table 2), perhaps caused by methodological differences in how measurements were obtained. Ongoing investigations are currently underway in our laboratory to further assess the relationship between micro-CT and OCT measurements.
Finally, we used generated computational 3D models from micro-CT datasets to create " y through" animations and develop virtual reality scenes (Fig. 3b). These scenes can enhance critical analyses of procedural outcomes and educate clinicians on methodological steps for bifurcation stenting.
Furthermore, our laboratory has developed an approach to transform virtual reality scenes into anaglyph visualizations ( Fig. 3c) in real-time, allowing multiple individuals to view PCI methodological scenes simultaneously (Fig. 3a). Free downloads of such scenes can be found in the Atlas of Human Cardiac Anatomy (http://www.vhlab.umn.edu/atlas/device-tutorial/stents/index.shtml). In addition, these models can be readily adapted for 3D printing as a unique tool that allows individuals to physically hold the same stent that they are ying through in virtual reality (Fig. 3d, 3e)-a "mixed reality" learning experience.

Discussion
Since 1997, the Visible Heart ® Laboratories have performed novel preclinical research on reanimated hearts and have developed an expansive Human Heart Library with >600 perfusion-xed specimens that can be accessed for anatomical research and device investigations 7,8 . We utilize state-of-the-art imaging tools, such as OCT and micro-CT imaging, to expand our ability to develop novel educational experiences and materials. We rmly believe that both reanimated human hearts and perfusion-xed hearts (at enddiastolic states) can be utilized to expand the current knowledge of methodologies in the bifurcation stenting of LM disease, where applicable.

Parameter Finalization
As previously stated, the Visible Heart ® Laboratories routinely use reanimated swine hearts for a large variety of cardiovascular technological studies. We previously utilized these implanted animal specimens to optimize and nalize the scanning parameters for this study 9 . Hence, we knew the imaging, scanning, and modeling parameters needed to achieve approximate reconstructed stent resolutions of 20µm, to accurately depict the stent struts.
Through our initial attempts of scanning coronary stents in perfused tissue, it became apparent that we were not using enough voltage/power to penetrate through the struts, allowing it to be picked up by the scanner's detector and causing a "halo-ing" effect 9 . By increasing the power in these perfusion-xed human heart specimens, we reduced halo artifacts but were still unable to penetrate appropriately through the struts, thus resulting in a hollowed core.
Other issues surfaced from employing too much radiation energy/power during these scans. In the xed human hearts, radiation easily penetrated through the stent struts and through the tissue. This in turn led to signi cant streaking and shadowing artifacts of the struts, which yielded less than desirable 3D models. Once the optimized voltage and current values were determined, we were able to nalize additional parameters such as number of projections and scanning duration to further reduce imaging artifacts. We continue to improve this methodology by investigating how we can detect different cardiac anatomies with similar densities (i.e., cardiac valves, endocardium, myocardium) and we recently installed a micro-CT X3000 North Star Imaging scanner system in our laboratory.

Visible Heart ® Methodology Applications in Coronary Artery Stenting
We continue to perform additional PCI studies to further validate the accuracy of our novel micro-CT imaging. This study focused primarily on achieving reproducible scans of implanted coronary stents, however our investigations continue to evaluate the nal stent apposition via these micro-CT methodologies. We understand that having near perfect apposition long before the end of PCIs is vital, especially when addressing issues such as wiring across a lesion to validate procedures. Such studies are also ongoing within perfusion-xed human specimens in our laboratory, and we have started to perform step-by-step bifurcation procedures with imaging employed for each recommended procedural step.
Finally, each perfusion-xed specimen was repeatedly imaged with micro-CT until we achieved ~20micron resolution scans of implanted devices, then high-resolution computational models were generated, analyzed, and further segmented to create various y through videos, 3D prints, and rendered virtual reality scenes.

Extension of Application
We are conducting ongoing studies to further determine the effects of formalin xation on diseased aortic and coronary tissues (i.e., uniaxial and biaxial testing). We observed here that diseased vessels lose their compliance, therefore the xation process had little to no effect on the mechanical properties of the tissue that we studied. We believe it would be valuable to utilize the 600+ human heart specimens from our Human Heart Library to perform similar bench-top experiments in real human anatomies. Furthermore, the methodologies described in this study are not limited to coronary stenting procedures. Rather they can be utilized for procedural testing of other cardiac devices to study device-tissue interactions with extremely high (~20 micron) resolutions. For example, we conducted a study in which various post TAVR-PCI procedures were performed within reanimated hearts on the Visible Heart ® apparatus; hearts were subsequently perfusion xed and later micro-CT scanned using the same parameters 10 .

Creation of Educational Stenting Modules
We envision that these described methodologies can generate high-resolution 3D models offering unique educational modules. For example, we have utilized these reconstructed models of bifurcation stenting procedures to create a variety of enlarged 3D prints, designed to better understand differences in various bifurcation techniques and their interactions with calci cations and/or complex anatomies. Further, mixed reality educational modules can couple 3D prints and virtual reality scenes 11 that allow individuals or multiple users to " y through" the same scenes 12 , while simultaneously performing physical inspections of the detailed models 13 . In other words, these mixed reality models can be used for unique educational endeavors such as anatomical analyses, identifying clinical plaque deposition trends, assessing coronary stent design, and/or developing computational uid simulations (see free-access website "Atlas of Human Cardiac Anatomy" to visualize and download models and associated procedural videos; http://www.vhlab.umn.edu/atlas/index.shtml). It is critical to develop such educational materials as means to better engage and stimulate learning. Such approaches provide unique insights related to how interventional procedures may be applied in appropriate clinical cases, and are valuable to students, clinicians, and medical device designers.

Conclusions
In conclusion, we developed reproducible methodologies for post-implant analyses of coronary artery stenting procedures performed within reanimated and perfusion-xed human hearts. We generated highresolution 3D models to investigate device-tissue interfaces. These novel approaches for preclinical cardiac device testing can produce models that can be analyzed to gain insights on procedures and device designs. Further, the 3D models have been useful for generating virtual reality scenes and 3D prints that can be utilized for a variety of educational mixed reality training. Our methodologies are not limited to studying coronary stenting technologies, but could be applied to all types of cardiac device technologies.

Specimen Procurement and Selection
The Visible Heart ® Laboratories received viable human hearts and heart-lung bloc specimens for research via LifeSource, a nonpro t organ procurement organization (Minneapolis, MN, USA). Donors gave consent for their organs to be used for scienti c research purposes via LifeSource. The hearts used in this study were deemed non-viable for transplant due to advanced patient age, cardiac downtimes, identi ed poor cardiac function, and/or other reasons. All specimens were received as fresh viable tissues, along with donors' relative cardiac-pulmonary clinical histories. We dissected lungs from the hearts while fresh, and each specimen was carefully cannulated for immediate reanimation, if speci c criteria were met.
Otherwise, hearts were placed within a formalin xation apparatus 14,15 for a 24-hour xation period. The perfusion-xation apparatus preserved hearts in their end diastolic shape, importantly keeping the aortas and coronaries dilated. After being perfusion xed (including hearts initially reanimated), each specimen was placed in its own container and stored for future studies.
We performed PCIs and micro-CT scanning utilizing six human heart specimens (see Table 3 for detailed patient information). Three of the specimens exhibited adequate cardiac function prior to donation and were considered viable candidates for reanimation and eventual perfusion xation, as described earlier.
The remaining three hearts were received from LifeSource two to eleven years prior to this study. We selected these specimens due to noted histories of coronary artery disease and prior imaging indicating the LM coronaries were patent enough to perform PCIs.

Coronary Intervention in Isolated Hearts
We performed PCIs in three reanimated hearts (HH534, HH541, HH556) using Visible Heart ® methodologies 16 ; subsequently each heart was carefully removed from the apparatus and perfusion xed so as not to damage the newly implanted stent(s). Previously xed hearts (HH115, HH202, HH479) were rinsed for 24 hours, re-cannulated, and placed in an acrylic box where they were attached to a Langendorff static perfusion apparatus (Fig. 4a) which continuously perfused the aortas and coronaries with water. The advantages of utilizing Visible Heart ® methodologies while performing PCIs were as follows: 1) endoscopic cameras enabled direct visualization of each procedural step; 2) there were no risks to any living patient, thus unlimited uoroscopy could be used; and 3) OCT could be utilized as many times as desired as there was no need to expedite the procedures.
Each PCI procedure was guided and recorded simultaneously by 2.4mm and 4mm endoscopic cameras (Olympus, Tokyo, Japan), OEC Elite Fluoroscopy (GE, Boston, MA, USA), and episodic OCT imaging, as shown in Figure 4b. The combination of these imaging modalities nearly simultaneously would not be possible without the use of a clear perfusion solution (Krebs-Henseleit buffer for reanimated hearts and water for xed hearts) that continuously circulated through the apparatus. All PCIs were conducted using Resolute Onyx drug-eluting stents, compliant and non-compliant Euphora balloons, and Cougar XT guidewires (Medtronic, Santa Rosa, CA, USA).

OCT Imaging during PCI
After stent implantations, we captured OCT scans using OPTIS Intravascular Imaging System and Dragon y TM Imaging Catheters (Abbott Vascular, Abbott Park, IL, USA). The automatic pullback system captured 540 frames over a scanning trajectory of 54mm with a 5mm penetration distance, to capture the highest resolutions possible (~100µm). Since specimens were continuously perfused using clear solutions in both setups, no contrast injections were needed or administered during OCT scanning.
Following ex vivo stenting, the three reanimated hearts were perfusion xed and then placed in formalin containers for long-term preservation. All OCT image datasets were post-processed to identify both distal and proximal portions of the implanted stents and then stored as 2D images. Since intracoronary OCT is a widely accepted method of imaging implanted stents, the images collected were later used to compare to the micro-CT reconstructions.

Micro-CT Scanning
Before micro-CT scanning each human heart, the specimens were rinsed in water for a minimum of 24 hours to remove traces of formalin before handling and/or transportation. Once thoroughly rinsed, each heart was placed within the specially constructed plexiglass container and then scanned using an X5000 micro-CT scanner (North Star Imaging, Rogers, MN, USA). All heart specimens were carefully perfusion xed to elicit an end diastolic shape for all four chambers (maximal lled) prior to scanning because the internal space of the scanner did not readily allow for our perfusion system to be used during scanning.
Each heart was placed in the scanner as shown in Figure 5, and imaging was performed utilizing the following parameters to achieve approximate isotropic voxel sizes of 20 x 20 x 20 µm: 170kV tube voltage, 144µA tube current, 24.5 isowatts, and 1,500 radiograph images captured throughout a ~15-min scanning duration. We selected these parameters after numerous iterations and scanning trials, all utilizing Resolute Onyx stents implanted in swine heart coronaries, to optimize scanning resolutions while minimizing streaking, shadowing, and/or scanning artifacts. Figure 6 shows the progression of scanning parameters trialed, until we nalized the optimal parameters used in this study. Once scans were completed, each heart specimen was returned to the laboratory and placed in its respective formalin container. Imaging datasets were then reconstructed using North Star Imaging's reconstruction software, into 8-bit 2D images (.tiff), for future analyses as described below.

3D Reconstructions
We imported .tiff les from micro-CT scans into the DICOM analysis software Mimics (Materialise, Leven, Belgium), where they were computationally "stacked" to form 3D volumes from 2D images 17 , followed by further post-processing. Using Mimics, for each heart's image dataset, we generated a high-frequency "mask" to segment out the higher density portions of the scan, i.e., the cobalt alloy shell and platinum iridium core of the Resolute Onyx. We manually created additional masks to segment out the vessel blood volumes and tissues. For subsequent analyses, each generated model consisted of a portion of the aortic wall, left coronary ostia, LM vessels, coronary stent(s), and proximal portions of left circum ex artery (LCX) and left anterior descending (LAD) coronaries. Once these models were generated, we used assessment tools in Mimics to measure the relative lumen areas. The 3D models were then exported from Mimics to be rendered as virtual reality scenes using a video game design software (Unity, Unity Technologies, San Francisco, CA, USA) which allowed for further visual inspections of LM stenting outcomes. Each heart exhibited varying areas prior to percutaneous coronary intervention (PCI) due to variability in heart sizes studied as well as extent of vessel disease, thus measurements were normalized by calculating overall percent increases in vessel lumen areas. OCT and micro-CT measurements were obtained of the lumen at distance 2mm distal to the implanted stents. Comparisons were made in specimens with and without disease (reanimated heart that was subsequently formalin xed) to evaluate lumenal changes associated with xation process.   Additional changes in vessel dimension post-intervention (before and after PCI), determined by importing computational models into 3-Matic. Speci c measurements were obtained at distances 5mm into the left main (LM) of each heart specimen, measured from the de ned border of each coronary ostia. Once vessel distance was determined, a slice (normal to the centerline) was created to expose the face of the blood volume. 3-Matic was then used to measure area of the digitally determined vessel lumen.

Figure 3
Virtual reality and 3D printing have been used as effective educational tools to enhance knowledge of bifurcation stenting interventions. In one application, the instructor can " y around" the anatomy of a bifurcation stent (a,b) while trainees follow along within anaglyph visualization scenes (c).
Simultaneously, mentors and trainees can hold 3D prints of the same model as viewed in the visualizations (d,e).
Page 16/17 Figure 4 (a) Each perfusion-xed isolated human heart was placed in a custom-made acrylic box for perfusion, utilizing uoroscopy. (b) Quad-split of bifurcation showing endoscopic imaging, uoroscopy, Optical Coherence Tomography images, and operator hand manipulations during percutaneous coronary intervention procedure.

Figure 5
The same plexiglass case used for performing percutaneous coronary interventions was then used for micro-CT scanning. Tubes were disconnected from the perfusion apparatus and clamped so uid was not spilled during scanning.

Figure 6
Initial attempts began with lower voltages and power that was unable to penetrate through the stent struts, resulting in outlining of the struts with black center cores. Too much voltage and power caused signi cant streaking in the stent struts, and further re nement of the parameters was needed to distinguish between the stent, tissue, lumen, and calci cation.