CineCT Platform for In Vivo and Ex Vivo Measurement of 3D High Resolution Lagrangian Strains in the Left Ventricle Following Myocardial Infarction and Intramyocardial Delivery of Theranostic Hydrogel

Abstract

Myocardial infarction (MI) produces acute changes in strain and stiffness within the infarct that can affect remote areas of the left ventricle (LV) and drive pathological remodeling. We hypothesized that intramyocardial delivery of a hydrogel within the MI region would lower wall stress and reduce adverse remodeling in Yorkshire pigs (n=5). ^99mTc-Tetrofosmin SPECT imaging defined the location and geometry of induced MI and border regions in pigs, and in vivo and ex vivo contrast cine computed tomography (cineCT) quantified deformations of the LV myocardium. Serial in vivo cineCT imaging provided data in hearts from control pigs (n=3) and data from pigs (n=5) under baseline conditions before MI induction, post-MI day 3, post-MI day 7, and one hour after intramyocardial delivery of a hyaluronic acid (HA)-based hydrogel with shear-thinning and self-healing properties to the central infarct area. Isolated, excised hearts underwent similar cineCT imaging using an ex vivo perfused heart preparation with cyclic LV pressurization. Deformations were evaluated using nonlinear image registration of cineCT volumes between end-diastole (ED) and end-systole (ES), and 3D Lagrangian strains were calculated from the displacement gradients. Post-MI day 3, radial, circumferential, maximum principal, and shear strains were reduced within the MI region (p<0.04) but were unchanged in normal regions (p>0.6), and LV end diastolic volume (LV EDV) increased (p=0.004), while ejection fraction (EF) and stroke volume (SV) decreased (p < 0.02). Post-MI day 7, radial strains in MI border zones increased (p = 0.04) and dilation of LV EDV continued (p=0.052). There was a significant negative linear correlation between regional radial and maximum principal/shear strains and percent infarcted tissue in all hearts (R²>0.47, p<0.004), indicating that cineCT strain measures could predict MI location and degree of injury. Post-hydrogel day 7 post-MI, LV EDV was significantly reduced (p = 0.009), EF increased (p=0.048), and radial (p=0.021), maximum principal (p=0.051), and shear strain (p = 0.047) increased within regions bordering the infarct. A smaller strain improvement within the infarct and normal regions was also noted on average along with an improvement in SV in 4 out of 5 hearts. CineCT provides a reliable method to assess regional changes in strains post-MI and the therapeutic effects of intramyocardial hydrogel delivery.

Graphical Abstract

1. INTRODUCTION

Ischemic heart disease and myocardial infarction (MI) is the leading cause of heart failure (HF) with reduced ejection fraction and remains a major public health problem worldwide [1]. MI causes regional dysfunction and wall thinning, which is associated with increased left ventricle (LV) wall stress. This injury results in infarct expansion, global myocardial remodeling, and LV dilation [2]. Early reperfusion with percutaneous coronary artery intervention has been shown to reduce infarct size, adverse remodeling, and subsequent development of HF, thus reducing long term morbidity and mortality [3]. Hemodynamic factors post-MI also appear to play an important role in development of HF and mortality as higher LV end diastolic pressure following MI associates with higher mortality at 3 months [4].

Recent studies have suggested that intramyocardial delivery of biomaterials, particularly injectable hydrogels, within the MI region may have therapeutic benefits via mechanical bulking [5–7] and delivery of therapeutics [8–9]. Beneficial effects of injectable hydrogels post-MI have been demonstrated in rodent [5–6] and several large animal [7,10–13] models. In one study, intramyocardial delivery of hydrogels significantly reduced LV diastolic and systolic volumes at 30 and 60 days when compared to controls [14]. Similarly, intramyocardial delivery of a decellularized matrix in porcine models of MI demonstrated a significant reduction in LV end diastolic and systolic volumes at 6 weeks post-treatment compared to controls [15]. These findings have recently been translated to clinical studies with promising results. In a randomized, double-blind, saline-controlled trial, Rao and colleagues demonstrated event-free survival following intracoronary injection of alginate in patients presenting with acute coronary syndrome over a 6-month follow-up period [16]. Intramyocardial delivery of alginate in patients with advanced heart failure also demonstrated an improvement in the 6-minute walk test (6MWT) at 6 and 12 months following injection [17,18]. More recently, a small group of patients received an intramyocardial delivery of decellularized porcine matrix, with significant improvement in functional exercise capacity evidenced by increments in the 6MWT and a decrease in symptoms at 3 and 6 months [18].

These experimental findings are consistent with computational modeling studies, which have predicted that hydrogel inclusions within the myocardium can lower the stress state of the infarct and peri-infarct regions by thickening and stiffening the region [19–23], which may lead to improved ejection fraction, stroke volume, and healing [19,23–24]. Beneficial effects appeared to depend strongly on the hydrogel material properties and distribution [24–26], with Wenk et al. [25] finding that a symmetric distribution of small inclusions in a 3 × 6 array might be particularly beneficial.

We hypothesize that intramyocardial delivery of injectable hydrogels within the MI region may provide therapeutic benefit by leading to lower regional strain-induced wall stress as well as reduced infarct LV wall thinning, infarct expansion, and subsequent LV dilation. Such changes should reduce subsequent adverse remodeling. In line with this hypothesis, our group recently developed a set of novel hyaluronic acid-based hydrogels with shear-thinning and self-healing properties that have significant therapeutic benefits when delivered intramyocardially [27]. In this study, we present a novel method for non-invasively tracking high resolution three-dimensional (3D) myocardial strains with ECG-gated contrast cine computed tomography (cineCT) using a well-established chronic porcine model of reperfused MI. We also present a novel apparatus for ex vivo cineCT imaging of a cardioplegia-perfused intact heart during cyclic pressurization to study local tissue deformation following infarction and intramyocardial delivery of these therapeutic hydrogels. We co-localized these high resolution in vivo and ex vivo regional 3D strain measurements with 3D perfusion/viability maps obtained with ^99mTc-Tetrofosmin SPECT imaging. This analysis allows us to investigate how the degree and extent of myocardial injury and the intramyocardial injection of hydrogels within the central MI region influences the active (in vivo) and passive (ex vivo) regional changes in LV strain. This detailed multimodality 3D analysis allows us to evaluate the beneficial mechanical effects of intramyocardial hydrogel injection and to investigate whether or not 3D strain could be used as a diagnostic tool for the non-invasive evaluation of MI location and severity.

2. METHODS

2.1. Overview of Experimental Model and Protocol

Studies were performed on eight male Yorkshire pigs (23–40 kg) with approval of the Yale University Institutional Animal Care and Use Committees (IACUC) in compliance with the National Institutes of Health Guidelines for Care and Use of Laboratory Animals. Control pigs (n = 3) underwent baseline in vivo and ex vivo x-ray CineCT imaging and hemodynamic monitoring without myocardial infarction or hydrogel delivery. A second group of pigs (n = 5) had hybrid SPECT/CT imaging performed at baseline, 3 and 7 days after induction of an anteroseptal MI as well as one hour after intramyocardial delivery of a hydrogel. Upon completion of in vivo imaging, animals were euthanized and 3 hearts were further prepared for ex vivo imaging under controlled, cyclic pressurization.

2.2. Animal Preparation and Hemodynamic Assessment

For all surgical and imaging studies, pigs were sedated with telazol/dexamedotomidine (4.4/0.02 mg/kg IM), intubated, and anesthetized with isoflurane (1.5–2%) delivered with an oxygen/nitrous mixture (3:1 L/min). The femoral artery and vein were cannulated for instrumentation, hemodynamic monitoring and administration of fluids (10 ml/min/kg). Heparin was administered (2000 units, IV) after catheter placement, then hourly (1000 units, IV) to prevent thromboembolic events during all interventional procedures.

2.3. Model of Reperfused Myocardial Infarction

Prior to induction of myocardial infarction, pigs were premedicated daily for 3 days with aspirin (80 mg po) and amiodarone (300 mg po) followed by intravenous infusion of amiodarone (1 mg/min) immediately prior to induction of MI. A transdermal fentanyl patch (50mcg/hr) was placed 12–16 hr prior to the first surgical procedure and maintained for 3 days after. Animals were administered Carpofen (2–4 mg/kg IM and PO) for 3 days post-MI. Under fluoroscopic guidance, a guide catheter was placed in the left main coronary artery and an angioplasty balloon catheter (2–2.5 mm diameter, 10 mm length, Cordis) was placed in the left anterior descending coronary artery distal to the first diagonal and inflated for 90 minutes followed by reperfusion. Coronary occlusion and reperfusion were verified with serial coronary angiography. Animals were monitored continuously for hemodynamic changes and arrhythmias during interventions and advanced cardiac life support was provided in the advent of life-threatening arrhythmias and hemodynamic instability.

2.4. Intramyocardial Hydrogel Delivery

A hyaluronic acid (HA)-based hydrogel with shear-thinning and self-healing properties that has been previously described in detail [27] was injected via a subxiphoid window through the open chest and pericardium into the mid myocardium in a 3×3 array [9] within the central infarct region with the aid of a grid pattern placed on the epicardial surface of the heart. Briefly, HA was modified with hydrazides (HA-HYD) via amidation or with aldehydes (HA-ALD) through oxidation with sodium periodate. HA-HYD and HA-ALD components were dissolved at 4 wt% in 100 mg/mL iohexol solution and formed into hydrogels by mixing the two components prior to injection. A total of 9 hydrogel injections of 0.1 ml [9] each were delivered to the mid-myocardial region within the MI region in all five post-MI pigs at approximately 7 days after MI. This delivery approach distributes the hydrogel evenly in discrete locations across the central infarct region and has been shown to reduce post-MI remodeling [9]. A 7-day window was also chosen to allow for stabilization of regional and global function and initial scar formation post-MI prior to hydrogel injection. This resulted in less variation in function indices post-MI, allowing us to better study the unique acute mechanical effects of hydrogel injection post-MI. This delivery protocol also better reflects optimal timing for clinical translation of this therapy, where an intervention may be delayed a few days until the patient has stabilized and long-term functional prognosis is better predicted.

2.5. Imaging

2.5.1. SPECT and Non-Contrast CineCT:

Pigs were injected with ^99mTc-tetrofosmin (18–20 mCi) 45 min into the 90-min balloon occlusion and underwent SPECT imaging ~2 hours post-injection to assess the area at risk using a stationary cardiac-dedicated hybrid SPECT/64-slice CT scanner (Discovery NM/CT 570c; GE Healthcare) equipped with cadmium zinc telluride (CZT) detectors and 19 tungsten pinhole collimators. At three days post-MI, pigs were reinjected with ^99mTc-tetrofosmin (18–20 mCi) during low dose dobutamine (5 ug/kg/min) stress and SPECT imaging was acquired to assess myocardial salvage. At day 7 post-MI, pigs were reinjected with ^99mTc-tetrofosmin (18–20 mCi) at rest and SPECT imaging was performed before hydrogel delivery to assess final infarct size. All SPECT data was acquired for 15 min with 20% energy window (± 10%) centered at the ^99mTc photopeak (140 kev) using a 32 × 32 matrix and 2.5 × 2.5 mm² pixel size. A non-contrast CT scan was acquired for nonuniform attenuation correction of SPECT images using the following acquisition parameters: 120 kV, 60 mA, pitch of 0.98, slice thickness of 2.5 mm, and rotation speed of 0.4 s. CT images were reconstructed using filtered backprojection with a voxel size of 0.98 mm³ and SPECT images were reconstructed with a matrix size of 70 × 70 × 70 and pixel size of 4 × 4 × 4 mm².

2.5.2. Contrast X-ray CineCT Imaging

In vivo contrast cineCT images were acquired on same 64-slice CT scanner following the intravenous injection of 30 ml of iodinated contrast (iohexal 350, Novaplus) via a power injector (Medrad) at a rate of 3 ml/min followed by 25 ml saline flush. The CT images were acquired at 120 kV, 350 mA, with a 0.625 mm slice thickness at end-expiration achieved by disconnecting the ventilator for a short duration (15–20 secs). Contrast CT images were acquired prior to MI, 3 and 7 days post-MI, and immediately after hydrogel injection. These ECG-gated contrast CT were acquired retrospectively and reconstructed over the entire cardiac cycle in 10% phases using filtered back projection with a voxel size of 0.488×0.488×0.625 mm³.

2.6. Ex vivo Heart Preparation and Imaging for Passive Mechanical Testing

After terminal in vivo imaging, animals were injected with heparin (2000 U i.v.) and euthanized with potassium chloride to assure end-diastolic cardiac arrest. Following euthanasia, hearts were rapidly excised with the aorta intact via a sternotomy and immediately perfused retrograde with chilled (4°C) UW cardioplegic solution to slow metabolic processes. The aorta was trimmed above the aortic valve to a length of 5 cm and a custom aortic valve insert for connection to 10 mm inner diameter (#36) tubing was secured to the aorta above the aortic valve with suture (Figure 1e). The connector was designed with a vent near the coronary artery root so that the coronaries could be continuously perfused as the pump supplied UW solution (Figure 1f–g). A small hole was cut in the left and right atrial appendages to free trapped air and the heart was immersed in a 5 Liter tank of UW solution and flushed of air. The aortic insert was then connected via tubing to a pump and the heart was suspended within the chilled UW chamber for imaging. A 3F Millar catheter was inserted into the LV through a port placed just above the aortic valve, and the output signal was connected to 2 microcontrollers. One was interfaced with a laptop running Labchart for recording the continuous pressure signal (LabChart 8.0, AD Instruments). The other was interfaced with a second laptop to provide real-time pressure feedback to a custom Labview control program that controlled the speed and direction of an external peristaltic pump connected to a reservoir of chilled UW solution. Within the custom Labview control program, the user set the peristaltic pump speed to a starting value of 100 revolutions per minute and selected high and low pressure bounds of 10 and 60 mmHg. Pump speed and high and low pressure targets were gradually adjusted until the heart cycled stably at 35 cycles per minute (cpm) between 10 and 60 mmHg. An artificial ECG signal was produced by the program each time the LV began filling to allow for 3D gated imaging. A heart cycle rate of 35 cpm was chosen because the lowest heart rate that the CT scanner would gate was 30 bpm, and this cycle rate provided a sufficient margin for repeatable 3D-gated imaging with a temporal resolution of approximately 20 frames/cycle for CT imaging. The lower pressure bound of 10 mmHg was chosen as the tank pressure at the depth the heart was immersed was 6–9 mmHg from basal to apex, thus this bound allowed the heart to reach as close to a traction-free state as possible at the lower limit of the pressure cycle without the LV cavity collapsing. The upper pressure bound of 60 mmHg was chosen as this was a high enough pressure to observe significant LV expansion and wall thinning that could be compared between regions without significant microstructural damage, which occurs in biaxial testing of excised myocardium at principal stretches 30% above baseline [28]. Early pilot studies (unpublished) indicated that higher pressures could lead to myocardial damage and LV stretching over time, which we hypothesize was related to the myocardium not being stress-shielded as might occur in vivo when the heart is blood perfused, metabolically active, contracting, and contained within the pericardial sac and closed thorax. Iohexol (20 ml) was stirred into the UW solution within the tank and dilute iohexol was injected into the aortic root and UW perfusate (15 ml, 1 ml/sec) during contrast CineCT imaging to obtain gated contrast images of the cycling heart. The spiral CT images were acquired at 100 kV, 350 mA, a 0.625 mm slice thickness, and ECG-gated contrast CT volumes were retrospectively reconstructed over the pressurization cycle in 5% phases using filtered back projection with a voxel size of 0.32×0.32×0.625 mm³.

Figure 1,

Contrast CineCT and ^99mTc-Tetrofosmin SPECT imaging and masking: a) in vivo contrast CineCT short and long axis imaging of porcine heart on day 7 after MI, b) in vivo SPECT short and long axis imaging of the porcine heart on day 7 after MI, c) SPECT-derived regional masks for strain fields of normal (white, > 60% max intensity), border (light gray, > 50% and < 60% max intensity), and infarct (dark gray, < 50% max intensity) regions, and d) SPECT strain masks overlaid on radial strain field. Ex vivo contrast CineCT imaging of the arrested porcine heart suspended in UW tank on day 7 after hydrogel delivery and euthanasia: e) custom aortic valve insert and suspension fixture, f) MI region and perfusion of right coronary artery (RCA) and left circumflex coronary artery (LCx), and g) MI region and left anterior descending coronary artery (LAD).

2.7. SPECT and CT Image Processing:

In vivo and ex vivo CT DICOM volumes were imported into ImageJ for formatting. Images were cropped to be centered on the LV cavity within a region of 125×125×125 mm. SPECT DICOM volumes were imported into ImageJ and the stack histogram was normalized using a 1% target saturation level. SPECT and CT images were then converted to 8-bit grayscale, saved as TIFF stacks, and imported into Matlab.

Using a custom Matlab program, normalized SPECT maps were resampled to the same resolution as the in vivo CT images they were acquired after and manually registered to those CT images. Ex vivo images were processed using a custom ImageJ macro, which inverted the grayscale images so that the myocardium was the brightest feature, and then used successively larger 3D median filters of size 2×2×2, 3×3×3, and 4×4×4 pixels to smooth the background signal of the tank and reduce radial CT streak artifacts. In vivo and ex vivo CT and SPECT images were then resampled to the LV axis as follows. A custom Matlab program was used to trace the endocardial boundaries of the LV at 2 different regions: 1) below the basal valve plane and just above the papillary muscles and 2) 7.5 mm above the endocardial surface of the apex. These boundaries were then extracted and fit to 2 ellipses to obtain the ellipse centroids. The centers of both ellipses were then used to define the LV axis, and the volume was resampled to be centered on the LV axis in XY, with the LV axis corresponding to the Z axis with a new isotropic resolution of 0.5×0.5×0.5 mm/pixel. Volumes were then manually rotated so that within the midcavity view the right ventricle (RV) was oriented to the left and the X-axis bisected the mid-septum (Figure 1a). SPECT images were transformed using the same transformation as their corresponding CT image (Figure 1b).

The volume of the LV cavity was calculated for the 10 phases in the cardiac cycle using a custom Matlab program. The user selected the long-axis position of the annulus of the basal valve plane and an intensity threshold corresponding to the contrasted LV cavity. Contrasted volumes were then segmented from the rest of the image, and the LV volume was selected as the largest continuous, contrasted region within each image (Matlab: bwconncomp). Stroke volume (SV) was calculated as the volume change between ED (the largest volume) and ES (the smallest volume) and ejection fraction (EF) was calculated as the fractional volume change between ED and ES.

2.8. U-Net AI Masking of In Vivo LV myocardium

To automate the segmentation of the LV from in vivo CT images for strain summary and presentation, we trained a deep neural network U-Net [29] model to segment contrast CineCT images of porcine hearts using LV masking data generated by 2 different experienced cardiac imagers. Training data was constructed by centering and resampling to the cardiac axis 567 CineCT volumes of 20 porcine hearts at various imaging times and contractile states between ED and ES with a resolution of 1×1×1 mm/voxel. Manual masking of the LV myocardium was conducted on short-axis images every 3 mm along the long axis from the basal valve plane to the apex using an FDA-approved clinical software package for 2D image analysis (SegmentCT, Medviso, Sweden). The trained AI model was an adapted version of the U-Net, a convolutional neural network (CNN). After training, the model was used to predict new LV segmentations at ED that were used to mask out displacement calculations for the papillary muscles and non-LV tissues prior to strain calculation to isolate the deformation of the LV myocardium. For further discussion and illustration of the training and accuracy of this tool, see Supplemental Materials Section 1.1.

2.9. High Resolution Displacement Tracking and Strain Analysis

Down sampled NIfTI images (1×1×1 mm) of the resampled CT volumes were exported corresponding to ED, ES, and a phase intermediate between ED and ES in vivo and to 10 and 60 mmHg at peak cavity expansion ex vivo (60–65% through the pressurization cycle). NIfTI volumes were processed with nonlinear registration, using BioImageSuite [30] (CP Spacing:4, Iterations:10, Resolution:1.25, Affine), to calculate displacement fields between these corresponding images. Incremental registration, from ED to the intermediate phase and then from the intermediate phase to ES, was used in vivo because it improved the accuracy of the tracking due to the large deformations and geometric changes between ED and ES. Displacement fields were imported into Matlab for strain analysis.

A custom Matlab program was used to mask displacement calculations to the LV. For in vivo displacements, LV masks generated by the U-Net AI at ED were used to remove displacements corresponding to nearby tissues from the displacement field of the LV. For ex vivo displacements, an intensity threshold was used to remove parts of the displacement field corresponding to the background, and a conical cutout was manually sized to remove displacements corresponding to the RV just beyond the septum. Once displacement fields were properly segmented to the LV, the X, Y, and Z component fields were smoothed by a 3D, symmetric Gaussian filter of size 5×5×5 (σ=3) with a tolerance of 50% for missing values to avoid over smoothing LV volume edges. Displacements were then converted to radial (U_R), circumferential (U_C), and long axis (U_Z) components, and the displacement gradients at each calculation point (1×1×1 mm) were calculated by fitting each component to local trilinear polynomials over the neighborhood of 11×13×11. 3D Lagrangian strains were calculated from the cylindrical deformation gradient at each point as follows:

$$\begin{array}{l}{E}_{RR}=\frac{\partial U_R}{\partial R}+\frac{1}{2}\left[{\left(\frac{\partial U_R}{\partial R}\right)}^2+{\left(\frac{\partial U_C}{\partial R}\right)}^2+{\left(\frac{\partial U_Z}{\partial R}\right)}^2\right],\\ \end{array}$$

$$E_{ZZ}=\frac{\partial U_Z}{\partial Z}+\frac{1}{2}\left[{\left(\frac{\partial U_R}{\partial Z}\right)}^2+{\left(\frac{\partial U_C}{\partial Z}\right)}^2+{\left(\frac{\partial U_Z}{\partial Z}\right)}^2\right],$$

$$E_{CC}=\frac{U_r}{R}+\frac{1}{R}\frac{\partial U_C}{\partial C}+\frac{1}{2}\left[{\left(\frac{1}{R}\frac{\partial U_Z}{\partial C}\right)}^2+{\left(\frac{1}{R}\frac{\partial U_R}{\partial C}-\frac{U_C}{R}\right)}^2+{\left(\frac{1}{R}\frac{\partial U_C}{\partial C}+\frac{U_R}{R}\right)}^2\right],$$

$$E_{RZ}=\frac{1}{2}\left[\frac{\partial U_R}{\partial Z}+\frac{\partial U_Z}{\partial R}+\frac{\partial U_R}{\partial R}\frac{\partial U_R}{\partial Z}+\frac{\partial U_C}{\partial R}\frac{\partial U_C}{\partial Z}+\frac{\partial U_Z}{\partial R}\frac{\partial U_Z}{\partial Z}\right],$$

$$E_{CR}=\frac{1}{2}\left[\frac{\partial U_C}{\partial R}+\frac{1}{R}\left(\frac{\partial U_R}{\partial C}+\frac{\partial U_R}{\partial C}\frac{\partial U_R}{\partial R}+\frac{\partial U_Z}{\partial C}\frac{\partial U_Z}{\partial R}+\frac{\partial U_C}{\partial C}\frac{\partial U_C}{\partial R}+U_R\frac{\partial U_C}{\partial R}-U_C\frac{\partial U_R}{\partial R}-U_C\right)\right],$$

$$E_{CZ}=\frac{1}{2}\left[\frac{\partial U_C}{\partial Z}+\frac{1}{R}\left(\frac{\partial U_Z}{\partial C}+\frac{\partial U_Z}{\partial C}\frac{\partial U_Z}{\partial Z}+\frac{\partial U_R}{\partial C}\frac{\partial U_R}{\partial Z}+\frac{\partial U_C}{\partial C}\frac{\partial U_C}{\partial Z}+U_R\frac{\partial U_C}{\partial Z}-U_C\frac{\partial U_R}{\partial Z}\right)\right].$$

E_RR is radial strain, E_CC is circumferential strain, E_ZZ is LV long-axis strain, E_RC is radial-circumferential shear strain, E_RZ is radial-axial shear strain, and E_CZ is circumferential-axial shear strain, all relative to convenient cardiac coordinates. We also calculated E^N_max (maximum principal strain) and E^S_max (maximum shear strain) to obtain measures independent of the (R,C,Z) cylindrical coordinate system as follows: 1) the eigenvalues of the Lagrangian strain were found at each calculation point, 2) E^N_max was calculated as the largest eigenvalue and E^N_min as the smallest eigenvalue, 3) E^S_max was calculated as (E^N_max − E^N_min)/2.

The chosen BioImageSuite nonlinear registration settings and strain-fitting parameters were optimized by applying a numerical deformation to the in vivo and ex vivo volumes and calculating the displacement and strain errors as shown in our previous work [31]. For this study, several improvements were made to the numerical deformation test including a step to overlay and verify registration of the two volumes in BioImageSuite and a more physiological numerically-applied deformation for in vivo images of −10% compression of the cavity. The results of this error analysis are presented in Supplemental Section 1.2 and Table S1.

2.10. Definition of Myocardial Regions

Cardiac strains were computed for regions based on the American Heart Association (AHA) 17-segment model and also divided into normal, peri-infarct border zone, and infarct zones as determined by threshold masking with registered, normalized SPECT Tc99m-tetrofosmin images. The true apex was excluded from the analyses. Voxels with a SPECT intensity below a threshold of 50% of the maximum in normal myocardium were classified as within the MI region. Voxels with SPECT intensities between 50–60% max were defined as the border region. All other regions were classified as normal regions. The AHA segments were also assigned a % viability score by averaging the voxel score within each AHA segment: voxels were assigned a 0 for infarct region, 0.5 for border region, and 1 for normal region (Figure 1d). Segments with a score < 50% were labeled as infarct segments; segments with scores ≥ 50% as normal segments. Regional strains were expressed as mean ± standard deviation of the mean.

2.11. Statistical analysis

Statistical analyses were performed using the MATLAB statistics and machine learning toolbox (version R2021). We analyzed for: 1) changes in strain between baseline, post-MI day 3, post-MI day 7, and post-Gel day 7 (in vivo time points); 2) regional differences in in vivo and ex vivo strain between the infarct, border, and normal regions; 3) changes in strain within the infarct, border, and normal regions between time points; 4) correlations between percent of normal tissue and average strain in the individual AHA segments; and 5) changes in ejection fraction (EF), heart rate, and LV ED volume between different time points.

Differences between the standard spatial regions and between normal, border, infarct, and area of risk regions were compared using a 2-way repeated measures ANOVA with a Sidak multiple comparison test (Matlab: ranova, fitrm). Repeated measures models were grouped such that measures from the same heart but different locations were considered repeated within-subjects measures, whereas observations from different hearts were considered independent measures. The Huynh-Feldt adjustment for lack of sphericity for was made to p-values for comparisons of 3 or more regions at once such as the comparison of infarct, border, and normal regions. In vivo changes between the time points Baseline Day 1, Post-MI Day 3, Post-MI Day 7, and Post-Gel Day 7 were tested for significance using a student T-test (Matlab: ttest). Changes in adjacent regions were compared using a 2-way repeated measures ANOVA (Matlab: ranova, fitrm). Pairwise comparisons within groups, mean differences, and 95% confidence intervals were obtained afterward from the fitrm model with the Matlab function multcompare, which adjusted p-values for multiple comparisons. P-values were considered significant if the p-value was less than 0.05 (**) and borderline significant if the p-value was less than 0.1 (*). A linear model (Matlab: fitlm) was used in each heart to test for correlation between the percent of each AHA segment that was normal (viability score) and the average strain in that segment.

3. RESULTS

3.1. Comprehensive CT Derived 3-Dimensional Strains: Baseline

3.1.1. In Vivo Strain For Control and Post-MI Hearts Prior to Intervention.

Baseline in vivo regional LV strain from ED to ES for both the control group (n=3) and the post-MI group at baseline (n=5) prior to intervention are summarized in Table 1 and illustrated in Figure 2a. Global radial strain (E_RR: 32.3±4.6%) was largely tensile as it represents the myocardium thickening radially into the cavity during contraction. Radial strains were of similar magnitude in the midcavity and basal regions but were reduced significantly toward the apex. This is primarily due to the change in curvature, as the surface normal changes from the radial direction (R) to the long axis direction (Z) in this region. Global circumferential strain (E_CC : −7.9±2.2%) was primarily compressive and represents the contraction around the circumference of the LV in basal and mid-cavity regions. E_CC did not vary significantly along the long axis but tended to be smaller near the apex. Global long-axis strain (E_ZZ: −1.5±2.5%) was primarily compressive in basal and mid-cavity areas as the LV foreshortens axially in these regions as the myocardium contracts and the basal valve plane descends. However, the strain changes to tensile at the apex since the myocardium thickens in the Z direction in this region. Maximum normal (E^N_max: 44.0±6.0%) and shear (E^S_max: 30.8±3.8%) strains were calculated to obtain a measure of strain independent of (R,C,Z) and changes of curvature. The circumferential shear strains tended to be near-zero on average (E_RC: 1.3±1.9%; and E_CZ: 0.4±0.7%), while radial-axial shear strain (E_RZ : 9.8±1.7%), tended to be slightly negative in the basal region and largely positive in the mid-cavity and apex regions relative to the prescribed coordinate system.

Table 1:

Average strains over the entire LV and basal, midcavity, and apex regions in vivo at baseline between ED and ES (n=8 pigs: 5 Gel, 3 Control) and ex vivo from pressurization of 10–60 mmHg (n=6 pigs: 3 Gel, 3 Control).

Baseline Strain In Vivo (n=8)	ERR Avg %	ECC Avg %	EZZ Avg %	ENmax Avg %	ESmax Avg %	ERC Avg %	ERZ Avg %	ECZ Avg %
Entire LV	32.3±4.6	−7.9±2.2	−1.5±2.5	44.0±6.0	30.8±3.8	1.3±1.9	9.8±1.7	0.4±0.7
Basal	38.2±7.8	−9.5±1.8	−14.2±2.3	43.0±8.5	32.1±4.7	1.9±2.2	−0.9±3.2	−0.7±1.3
Midcavity	38.0±5.5	−7.0±1.9	−4.6±3.7	46.1±6.0	31.6±3.8	2.1±2.2	13.4±2.4	0.9±0.9
Apex	18.6±3.9	−7.3±4.5	17.1±5.7	42.5±6.8	28.3±3.8	−0.5±1.7	18.2±2.9	1.1±1.9
Control Strain Ex Vivo (n=3)	ERR Avg	ECC Avg	EZZ Avg	ENmax Avg	ESmax Avg	ERC Avg	ERZ Avg	ECZ Avg
Entire LV	−2.7±0.2	8.5±1.1	−2.0±0.7	16.9±4.0	15.5±3.6	−2.2±2.2	−3.0±2.4	−2.0±0.8
Basal	−5.0±0.8	5.1±1.6	−0.1±0.6	13.9±4.4	13.7±3.8	−4.9±3.0	−0.2±4.0	−3.9±0.8
Midcavity	−4.0±1.2	7.3±0.8	−1.1±1.5	15.4±3.5	14.9±3.6	−1.8±2.5	−5.7±1.3	−2.4±1.1
Apex	3.1±2.4	15.6±1.5	−6.3±1.5	24.1±6.9	19.3±5.1	1.4±1.0	−3.4±2.8	0.1±0.7
PostGel Strain Ex Vivo (n=3)	ERR Avg	ECC Avg	EZZ Avg	ENmax Avg	ESmax Avg	ERC Avg	ERZ Avg	ECZ Avg
Entire LV	−2.7±0.2	8.5±1.1	−2.0±0.7	16.9±4.0	15.5±3.6	−2.2±2.2	−3.0±2.4	−2.0±0.8
Basal	−5.0±0.8	5.1±1.6	−0.1±0.6	13.9±4.4	13.7±3.8	−4.9±3.0	−0.2±4.0	−3.9±0.8
Midcavity	−4.0±1.2	7.3±0.8	−1.1±1.5	15.4±3.5	14.9±3.6	−1.8±2.5	−5.7±1.3	−2.4±1.1
Apex	3.1±2.4	15.6±1.5	−6.3±1.5	24.1±6.9	19.3±5.1	1.4±1.0	−3.4±2.8	0.1±0.7

Figure 2,

Baseline LV strain fields. Representative color-coded strain maps with superimposed CT image (black & white) are shown for: a) baseline in vivo strain for LV between ED and ES and b) Representative control ex vivo strain in LV when pressurized from 10 to 60 mmHg. Shown from left to right are: radial strain (E_RR), circumferential strain (E_CC), long-axis strain (E_ZZ), maximum principal strain (E^N_max), and maximum shear strain (E^S_max). From top to bottom of (a) and (b) panels: midcavity view (50% of heart length), long axis view at 0° (mid-septum axis), long axis view at 90° (perpendicular to mid-septum axis).

3.1.2. Control Ex Vivo Strain

Control ex vivo strain patterns in the LV for pressurization from 10 to 60 are summarized in Table 1 and illustrated in Figure 2b. Global radial strain (E_RR: −2.7±0.2%) was largely compressive and represents the myocardium thinning radially during pressurization. Radial strains were of similar magnitude in the mid-cavity and basal regions but were reduced significantly toward the apex. This is primarily due to the change in curvature, as the surface normal changes from the radial to the long-axis direction in this region. Global circumferential strain (E_CC: 8.5±1.1%) was primarily tensile as it represents the expansion of the circumference of the LV during pressurization. This strain varied with long axis and tended to be larger near the apex. Global long-axis strain (E_ZZ: −2.0±0.7%) was tensile and small in magnitude in basal and mid-cavity areas as it measures the strain within the heart wall as myocardium lengthens under pressurization; however, E_ZZ was larger in magnitude and compressive in apex regions. Maximum normal (E^N_max: 16.9±4.0%) and shear (E^S_max: 15.5±3.6%) were calculated to obtain a measure of strain independent of (R,C,Z) and changes of curvature. Global shear strains (E_RC :−2.2±2.2%; E_CZ: −2.0±0.8%) tended to be slightly negative on average with the largest shear strains in the basal region, but radial-axial shear (E_RZ : −3.0±2.4%) tended to be near-zero in basal regions and negative in the mid-cavity and apex regions. Thus, some regional variations in ex vivo strains were observed with LV for pressurization in normal hearts.

3.2. Serial Changes in LV Volume Post-MI and Hydrogel Delivery

Post-MI on days 3 and 7, we noted a progressive dilation of the LV cavity at ED (Day 3: +8.8 mL, Day 7: +14.0 mL, p < 0.015) and worsening of LV ejection fraction (Day 3: −11%, Day 7: −14%, p < 0.15) and stroke volume (Day 3: −2.8 mL, Day 7: −2.8 mL, p < 0.09) compared to baseline (Table 2–3, Figure 4). Post-Gel day 7 (Table 2–3, Figure 4), we noted a significantly smaller LV ED volume (−4.8 mL, p = 0.009) and improved ejection fraction (+6%, p = 0.048). Stroke volume also improved in 4 of 5 hearts.

Table 2:

Average strain, LV myocardial volume, heart rate, LV ED cavity volume, LV ejection fraction, LV stroke volume, and LV myocardial thickness at ED for the MI Group (n=5) at Baseline, Post-MI Day 3, Post-MI Day 7, and Post-Gel delivery Day 7 assessed in vivo and ex vivo. Regional strain summaries are defined as follows: Infarct 16S is the average within the infarcted 16 AHA segments, Normal 16S is the average within the normal segments, and Normal 3D, Border 3D, and Infarct 3D are averages within the normal, border, and infarct regions, masked in 3D within the myocardium wall from registered SPECT perfusion images.

Time Points	Regions	LV Myo-Cardial Volume	ERR Avg %	ECC Avg %	EZZ Avg %	ENmax Avg %	ESmax Avg %	Heart Rate (bpm)	LV EDV (mL)	LV EF (%)	LV SV (mL)	LV ED Thickness (mm)
Baseline	Full LV	60±1 cm3	34.2±5.0	−8.9±2.1	−0.9±3.0	46.4±6.2	32.3±3.9	95±13	58.3±1.9	53.6±5.3	31.2±3.2
Day 1 In Vivo N = 5	Normal 16S	-	39.5±6.9	−8.8±1.7	−8.0±4.7	46.6±6.7	33.1±4.0					7.41±0.13
	Infarct 16S	-	25.9±3.8	−8.9±2.8	9.9±4.3	45.9±6.7	31.0±4.1					7.36±0.29
Day 1 In Vivo N = 5	Full LV	60±1 cm3	27.6±5.7	−6.6±1.0	−1.7±1.6	36.7±7.1	26.0±4.2	101±8	67.0±3.5	42.5±5.3	28.5±3.3
	Normal 16S	-	40.6±9.5	−8.9±0.9	−5.4±2.0	47.5±10.4	32.6±6.0					8.08±0.49
	Infarct 16S	-	9.6±5.5	−3.1±2.0	3.0±3.9	21.4±7.9	16.6±5.1					9.03±0.50
	Normal 3D	47±5%	40.3±7.7	−9.3±0.9	−4.6±1.2	47.8±9.2	32.8±5.5
	Border 3D	12±2%	16.6±4.5	−5.8±0.9	−0.1±3.8	27.5±7.9	20.9±4.6
	Infarct 3D	40±3%	8.5±5.8	−1.9±2.8	3.0±5.9	19.5±6.3	15.2±4.5
PostMI	Full LV	58±4 cm3	25.2±3.4	−5.6±2.3	−1.8±2.3	32.7±3.7	22.9±2.4	104±8	72.3±7.6	40.0±7.0	28.5±1.8
Day 7 In Vivo N = 5	Normal 16S	-	35.7±3.0	−8.0±2.3	−6.8±3.7	40.8±2.4	28.6±14.9					8.00±0.49
	Infarct 16S	-	9.8±4.3	−1.7±2.8	4.3±2.4	20.7±5.6	14.9±3.5					8.43±0.68
	Normal 3D	48±4%	39.9±6.8	−9.2±1.6	−6.3±3.0	45.5±7.9	31.4±4.8
	Border 3D	7±1%	19.5±5.7	−6.2±2.6	−0.2±1.1	27.9±7.1	20.5±4.5
	Infarct 3D	45±5%	7.8±2.5	−1.1±3.1	3.9±2.7	18.5±4.1	13.6±2.2
PostGel	Full LV	60±5 cm3	26.0±8.4	−6.7±3.2	−1.8±3.5	36.6±12.5	26.4±8.0	84±3	66.8±8.8	47.2±9.2	31.2±5.7
Day 7 In Vivo N = 5	Normal 16S	-	40.6±9.5	−10.0±3.2	−5.5±3.7	48.2±10.4	33.6±6.4					8.50±0.70
	Infarct 16S	-	12.4±6.5	−3.7±2.4	4.9±5.6	27.8±13.1	20.7±8.6					8.86±0.84
	Normal 3D	48±4%	42.4±9.7	−10.2±3.7	−5.3±3.2	50.4±10.6	34.8±6.5
	Border 3D	6±0%	26.3±9.6	−7.3±2.9	−0.1±7.9	37.8±12.6	27.3±7.8
	Infarct 3D	46±4%	10.0±5.4	−3.1±1.2	4.9±5.2	25.2±11.2	19.1±7.1
PostGel	Full LV	60±5 cm3	−7.9±4.8	8.8±1.6	1.0±1.2	15.5±2.6	15.4±2.9	35 cpm	-	-	-
Day 7 Ex Vivo N = 3+	Normal 16S	-	−10.1±4.2	7.5±1.7	3.7±1.8	14.5±2.4	15.0±2.9
	Infarct 16S	-	−4.4±5.5	10.7±1.6	−2.3±3.5	16.9±2.8	16.9±2.8

⁺Two of the pigs did not have ex vivo imaging post-mortem.

Table 3:

Statistical results for T-tests of the significance of changes in vivo to strain and LV volume measurements between the 4 time points of the experiment (n=5 pigs). Regional strain summaries are defined as follows: Infarct 16s is the average of strain within the infarcted 16 AHA segments, Normal 16S is the average of strain within the normal segments, and Normal 3D, Border 3D, and Infarct 3D are the averages of strain within the normal, border, and infarct regions, masked in 3D within the myocardium wall from registered SPECT perfusion images. Changes of p < 0.1 are in bold, and borderline significant changes of p > 0.05 are also italicized.

T-tests of	Region	Post-MI Day 3 -	Post-MI Day 7 -	Post-Gel Day 7 -
Changes		Baseline Day 1	Baseline Day 1	Post-MI Day 7
LV EDV	Full LV	0.004	0.014	0.009
LV EF	Full LV	0.001	0.013	0.048
LV SV	Full LV	0.017	0.086	0.307
EZZ	Infarct 16S	0.068	0.065	0.796
	Normal 16s	0.210	0.516	0.367
	Normal 3D	-	-	0.517
	Border 3D	-	-	0.977
	Infarct 3D	-	-	0.599
ERR	Infarct 16S	0.002	0.001	0.112
	Normal 16s	0.619	0.148	0.192
	Normal 3D	-	-	0.486
	Border 3D	-	-	0.021
	Infarct 3D	-	-	0.186
ECC	Infarct 16S	0.039	0.003	0.183
	Normal 16s	0.927	0.541	0.191
	Normal 3D	-	-	0.579
	Border 3D	-	-	0.608
	Infarct 3D	-	-	0.139
ENmax	Infarct 16S	0.003	0.003	0.129
	Normal 16s	0.722	0.050	0.113
	Normal 3D	-	-	0.322
	Border 3D	-	-	0.051
	Infarct 3D	-	-	0.123
ESmax	Infarct 16S	0.005	0.003	0.086
	Normal 16s	0.618	0.044	0.097
	Normal 3D	-	-	0.283
	Border 3D	-	-	0.047
	Infarct 3D	-	-	0.081

Figure 4,

Changes in a) LV end-diastolic volume (EDV), b) stroke volume (SV), and c) ejection fraction (EF) between Baseline day 1, Post-MI day 3, Post-MI day 7, and Post-Gel day 7 (n=5 pigs). From left to right: (1) averages over time, (2) changes Post-MI on Day 7 compared to Baseline (worsening function), and changes Post-Gel on Day 7 compared to Pre-Gel (recovering function). Statistical tests: student’s t-test of changes from baseline (*) and sequential changes (+) (Table 3). Significance levels: ** or ++ = p < 0.05, * or + = p < 0.1. Whiskers on the line plots show the standard error of the mean. Boxplots indicate median changes and 25% and 75% quantiles, with individual changes overlaid as points.

3.3. Serial Changes in Strain Post-MI and after Hydrogel Delivery

Within AHA segments that had less than 50% of viable tissue as determined by SPECT imaging (Table 2–3, Figure 5) we observed a significant decline in the radial, circumferential, maximum principal, and maximum shear strain at days 3 and 7 post-MI (p < 0.04). Strains were relatively unchanged in the normal segments of the heart on day 3, but maximum principal/shear strain declined from baseline on average by day 7 (p ≤ 0.05). After hydrogel delivery on day 7, there was an average increase in radial, circumferential, and maximum principal/shear strain in both the infarct and normal segments in 4 of 5 hearts.

Figure 5,

Changes in average strain within the AHA 16 segments labeled as either infarcted segments (<50% viable tissue, Infarct) or normal segments (> 50% viable tissue, Normal) (n=5 pigs): a) changes in radial strain, b) changes in circumferential strain, c) changes in maximum principal strain, and d) changes in maximum shear strain. From left to right: (1) average strain over time, (2) changes Post-MI on Day 7 compared to Baseline (worsening function in infarct segments), (3) changes Post-Gel on Day 7 compared to Pre-Gel (improving function in infarct and normal areas in 4/5 hearts). Statistical tests: student’s t-test of changes from baseline (*) and sequential changes (+) (Table 3). Repeated measures ANOVA (*) comparing normal and infarcted segments. Significance levels: ** or ++ = p < 0.05, * or + = p < 0.1. Whiskers on the line plots show the standard error of the mean. Boxplots indicate median changes and 25% and 75% quantiles, with individual changes overlaid as points.

To further investigate functional differences between the normal, border, and infarct regions, we constrained the strain fields in 3D to within the myocardium wall by masking the CT images with registered SPECT perfusion images as described in Section 2.10. In Figure 6, we define the regional differences in myocardial strain post-MI on day 7. All strain components were shown to decline in the border and infarct regions over time, with the fully infarcted region having the lowest strain (p < 0.02, Figure 7). Post-MI between days 3 and 7 (Figure 7) there was an increase in E_RR strain within the border zone (p = 0.037), and a reduction on average in E_RR, E^N_max, and E^S_max in normal regions. After hydrogel delivery on day 7 (Figure 7), there was a significant increase in E_RR (p = 0.021), E^N_max (p = 0.051), and E^S_max (p = 0.047) within the border regions, and an average increase in E^N_max (p = 0.123) and E^S_max (p = 0.081) within the infarct region in 4 of 5 hearts.

Figure 6,

Strains for normal, border, and infarct regions of LV Post-MI Day 7 (n=5): a) radial strain (p=0.00002), b) circumferential strain (p=0.0002), c) maximum principal strain (p=0.0004), and d) maximum shear strain (p=0.0001). Statistical tests: repeated-measures two-way ANOVA with Huynh-Feldt sphericity adjustment comparing all 3 regions at once. P-values for pairwise comparisons of regions were adjusted for multiple comparisons and are shown in Table 4. Significance levels: ** = p < 0.05, * = p < 0.1. Boxplots indicate median strains and 25% and 75% quantiles, with individual regional strain averages overlaid as points.

Figure 7,

Changes in strain within the normal, infarct, and border zones (n=5 pigs) masked from 3D registered SPECT perfusion imaging: a) changes in radial strain, b) changes in circumferential strain, c) changes in maximum principal strain, and d) changes in maximum shear strain. From left to right: (1) average strain over time and (2) changes Post-Gel on Day 7 compared to Pre-Gel (improving function, most significant in border zones). Statistical tests: Student’s t-test of changes (*) (Table 3). Repeated-measures two-way ANOVA with Huynh-Feldt sphericity adjustment comparing changes in all 3 regions (*). Significance levels: ** = p < 0.05, * = p < 0.1. Whiskers on the line plots show the standard error of the mean. Boxplots indicate median changes and 25% and 75% quantiles, with individual changes overlaid as points.

3.4. Assessment of Myocardial Viability Using 3D CT Strain Analysis

To investigate the feasibility of using 3D cineCT regional strain analysis to reliably delineate the infarct region with higher spatial resolution, we correlated the percent of normal tissue in each of the AHA segments (the % viability score) with each of segmental strain averages at 7 days post-MI (Figure 8). For all strain components in all hearts, strain appeared to decline linearly with the percentage of infarcted tissue in each segment. Post-MI day 7 we observed strong correlations with similar slopes for all 5 hearts between the percent viability score and percent strain (Table 5) for E_RR (R²>0.69, Slope=0.38, p<0.0006), E^N_max (R²>0.47, Slope=0.31, p<0.004), and E^S_max (R²>0.63, Slope=0.20, p<0.0003), which may reflect the resolution of stunned myocardium. Weaker, negative correlations were seen for E_CC (R²>0.21, Slope=−0.10, p<0.08) and E_ZZ (R²>0.18, Slope=−0.13, p<0.1) as these components are compressive in vivo. Similar correlations were seen on day 3 post-MI, but with more variability in slope (data not shown).

Figure 8,

Variation of AHA segment (1–16) in vivo strains Post-MI on day 7 compared to degree of myocardial viability based on the percent of the segment that was normal as assessed by SPECT (n=5 pigs): a) radial strain, b) circumferential strain, c) maximum principal strain, and d) maximum shear strain. Statistical tests: linear, single-variable model fit to the regional viability scores and strains in each heart and one model fit to all regions across all hearts (the trendline shown). R² values and slopes shown are reported for the trendline along with the range of p-values for each of the 5 individual linear models for each heart (Table 5).

Table 5:

Statistical data (R², slope, p-value) for linear models fit to investigate the relationship between average segment viability and strain (n=5) Post-MI on day 7.

Linear Models of Segment Viability/Strain Post-MI
Strain	Pig #	R2	P-Value	Slope
ERR	1	0.717	0.00004	0.356
	2	0.789	<0.00001	0.367
	3	0.740	0.00002	0.350
	4	0.764	0.00001	0.440
	5	0.695	0.00006	0.368
	All	0.736	<0.00001	0.379
ECC	1	0.547	0.00106	−0.073
	2	0.212	0.07293	−0.075
	3	0.643	0.00019	−0.149
	4	0.371	0.01231	−0.087
	5	0.375	0.01167	−0.102
	All	0.390	<0.00001	−0.101
EZZ	1	0.316	0.02353	−0.149
	2	0.296	0.02943	−0.124
	3	0.260	0.04351	−0.092
	4	0.182	0.09890	−0.089
	5	0.402	0.00833	−0.248
	All	0.247	<0.00001	−0.127
ENmax	1	0.470	0.00336	0.245
	2	0.762	0.00001	0.322
	3	0.549	0.00102	0.272
	4	0.751	0.00001	0.387
	5	0.463	0.00374	0.283
	All	0.605	<0.00001	0.308
ESmax	1	0.633	0.00023	0.159
	2	0.863	<0.00001	0.204
	3	0.701	0.00005	0.179
	4	0.792	<0.00001	0.239
	5	0.693	0.00006	0.214
	All	0.728	<0.00001	0.202

3.5. Evaluation of Ex Vivo Regional Strains Post-MI and Intramyocardial Gel Delivery

The ex vivo preparation allowed us to assess regional variations in myocardial strain in response to cyclic pressurization independent of contraction (Figure 9). The infarcted segments were defined as those with less than 50% normal tissue, while normal segments were defined by those segments with greater than 50% viable tissue based on in vivo SPECT imaging post-MI day 7. Based on our ex vivo pressure deformation analysis, the infarct region post hydrogel delivery appeared significantly stiffer radially than the normal regions (E_RR, p=0.028), although more compliant on average in the circumferential and long-axis directions (E_CC: p=0.028, E_ZZ: p=0.057, E^N_max: p=0.085). To determine if this variation was due to myocardial injury or potential regional variations due to differences in wall thickness related the location of papillary muscles and RV, we compared the average ex vivo strain in the same segments classified as either infarcted (7–8: midcavity and 13–16: apex) or normal (1–6: basal and 9–12: midcavity) post-MI as those in the control hearts (Figure 9). We did not observe any significant differences in the strain response of the infarct regions between the post-MI hydrogel and control groups, but did note that strain tended to be larger on average in the post-MI hydrogel group, particularly radial strain in the apex (p=0.15). Strain within normal regions, however, did appear to be significantly larger within the post-MI hydrogel group, with E_RR more compressive (p=0.085) and E_CC (p = 0.020) and E_ZZ (p = 0.021) more tensile. This suggests an alteration of myocardial deformation in the remote regions of the heart 7 days post-MI following acute intramyocardial delivery of gel in the infarct region, possibly due to structural remodeling in remote normal regions 7 days post-MI. While intramyocardial gel delivery normalized pressure-induced deformations in the infarct region.

Figure 9,

Regional analysis of ex vivo strain within Post-MI/Gel hearts (n=3) and normal hearts (n=3) averaged within the AHA segments that were consistently infarcted (<50% viable, AHA: 7–8,13–16) and normal (>= 50% viable, AHA: 1–6,9–12): a) radial strain, b) circumferential strain, c) long-axis strain, and d) maximum principal strain. From left to right: (1) strain compared within normal and infarct segments of the Post-Gel hearts (n=3), (2) strain within the consistently infarcted segments compared between Gel (n=3) and Control (n=3) groups (no significant difference), (3) strain within the consistently normal segments compared between Gel (n=3) and Control (n=3) groups (significantly different). Statistical tests: repeated-measures two-way ANOVA for regional comparison (1) and two-ANOVA for comparisons between groups (2–3). Significance levels: ** = p < 0.05, * = p < 0.1. Boxplots indicate median strains and 25% and 75% quantiles, with individual regional strain averages overlaid as points.

3.6. Numerical Deformation Test and Displacement and Strain Errors

The accuracy of BioImageSuite’s nonlinear registration and custom strain calculation methods for a numerically-applied deformation within in vivo and ex vivo images at ED for Pig 1 are presented in Supplemental Section 1.2 and Table S1. Overall, global average tracking errors were less than 0.21 mm and absolute average tracking errors (a measure of local accuracy) were less than 0.15 mm for both ex vivo and in vivo images. Global average strain errors were less than 0.5% in our analysis of ex vivo images, but larger at 0.21–2.84% for our in vivo strain analyses. Absolute average strain errors (local accuracy) were E_ZZ = 1.17%, E_RR = 0.85%, E_CC = 1.23% ex vivo and E_ZZ = 1.08%, E_RR = 2.84%, E_CC = 1.02% for in vivo image analyses.

4. DISCUSSION

We presented a novel experimental platform to assess changes in active (in vivo) and passive (ex vivo) deformations of the LV myocardium within infarct, border, and remote normal regions using semi-automated 3D analysis of contrast CineCT images before and 3 and 7 days after inducing a reperfused anteroseptal MI in porcine hearts as well as after intramyocardial delivery of a therapeutic hydrogel 7 days post-MI. ^99mTc-tetrofosmin SPECT imaging was used to assess the ischemic risk area and myocardial viability post-MI and regional perfusion was related to regional strains derived from serial high resolution contrast cineCT images. Our 3D cineCT analysis was able to evaluate serial regional changes in the cumulative 3D displacement field with an average error of 0.00–0.21 mm, along with cardiac-specific strains, maximal principal strain, and maximal shear strain early post-MI and following intramyocardial hydrogel delivery, with an average error of 0.9–2.8%. Our novel ex vivo perfused heart preparation with controlled LV pressurization from 10 to 60 mmHg similarly allowed assessment of the passive material properties following chronic MI and acute intramyocardial hydrogel delivery within the infarct region.

As expected, control and baseline in vivo Lagrangian strains between ED and ES were largely tensile in the radial direction (E_RR) at +32.3%, while circumferential strain E_CC was primarily compressive at −7.9%. Using our cylindrical model, long-axis strain (E_ZZ) was primarily compressive in basal (−14.2%) and mid-cavity (−4.6%) regions reflecting descent of the basal valve plane during contraction, and tensile in apical regions (+17.1%) due to myocardium thickening in the Z direction in this location. Maximum normal E^N_max (+44.0%) and shear E^S_max (+30.8%) were both positive and reflect strain independent of (R,C,Z) coordinate directions and changes of curvature. The radial-circumferential (E_RC) and circumferential-long axis shear strains (E_RC) tended to be near-zero on average, although radial-long-axis (E_RZ) shear strain tended to be slightly negative in the basal region and largely positive in the mid-cavity and apex regions. The sign change of E_RZ is consistent with the observed change in the orientation and curvature of endocardial surface of the LV wall in these locations. While the heart is also known to twist during contraction [32], we did not observe a consistently large E_CZ shear strain. This may relate to the limited temporal resolution available from in vivo cineCT imaging. Overall, the Lagrangian strain pattern that we observed at baseline and in control pigs is consistent with well-established normal cardiac physiology [33–36]. The magnitude of the strains measured from ED to ES are similar to those reported in other image-based strain studies employing large animal models such as Sinusas et al. [37] with magnetic resonance imaging (MRI) of dogs (radial: +17–26%; circumferential: −7–14%; and longitudinal: −3% to +2% basal to apex gradient) and Pislaru et al. [38] with echocardiographic imaging of pigs (radial: +31–36%). Stathogiannis et al. [40] tracked LV strain using feature tracking with another commercially available software package (SuiteHEART, Neosoft, Pewaukee, WI) with MRI in human subjects and also reported similar strains (circumferential: −11.3%, longitudinal: −12.2%). We note that these prior strain studies either tracked the epicardial surface of the heart [37] or used clinical programs such as Echoinsight, SegmentCT, or SuiteHeart, which provide only a 2D strain formulation, with at best a coarse approximation of true 3-dimensional strain due to out of plane displacements. To our knowledge, this has been the case for most clinical strain studies. Lagrangian strains, as described in Section 2.9, account for the effect of the out of plane deformation on in-plane strains and provide a more accurate description of 3D strain, particularly in situations where 3-dimensional strain patterns are complex and deformations can be large as in the heart, and may also be complicated by superimposed respiratory motion. Thus, 2D strain methods will tend to underestimate true strain by significant margin. Gupta et al. [39] also evaluated Lagrangian strain in the LV using cardiac cineCT and a commercially available 2D image analysis package to segment the myocardium (SegmentCT, Medviso AB, Lund Sweden) and a 3D image registration approach to obtain Lagrangian strain in human subjects, however, strains were then averaged and only reported for the standard 17 segments. They also reported global strains (radial: +31.1%, circumferential: −23.5%, longitudinal: −17.2%). Our analysis differs significantly from that of Gupta et al, since they masked the volume at ED and tracked these masks directly to subsequent phases such as ES and calculated their strains in the original XYZ coordinate system of the images before converting the strains to cylindrical. In our study, we used 3D nonlinear image registration to track the entire volume of the heart and surrounding tissues from ED to an intermediate phase and then from that intermediate phase to ES, using a myocardial mask to segment out the LV from surrounding tissues post-tracking. We found this multi-step approach yielded a much more accurate strain calculations than a direct mapping from ED to ES due to the large deformations and poor contrast across the wall in cineCT. We also converted our cartesian displacements to the cylindrical coordinate system and smoothed them prior to strain calculation, which we found to markedly improve strain smoothness and accuracy. In summary, our semi-automated Lagrangian 3D strain analysis provided in vivo regional and global strains comparable to the global strains reported in these prior studies, but offers an advantage by providing a high-resolution index of true 3D deformation of the heart that can be colocalized with masks of the infarct region to define changes to both the infarct and bordering peri-infarct regions.

Post-MI day 3, we observed a significant decline in radial, circumferential, maximum principal, and maximum shear strain within each of the 16 AHA segments that had less than 50% viable tissue, as determined by ^99mTc-Tetrofsomin SPECT imaging, for which myocardial retention depends on an intact mitochondrial membrane potential and myocellular viability. In vivo strains were mostly unchanged in remote normal regions of the heart. Our ^99mTc-Tetrofsomin SPECT maps were used to define fully infarcted, border, and normal regions in 3D within the myocardium wall post-MI on days 3 and 7. Between days 3 and 7, we noted an average increase in infarct size reflecting early infarct expansion, along with a progressive dilation of the LV cavity at ED and worsening ejection fraction. There was a significant increase in E_RR within the border zone, and an average reduction in E_RR, E^N_max, and E^S_max (reduced compensation) in normal regions reflecting early LV remodeling.

After acute hydrogel delivery on day 7, we noted a consistently smaller ED volume (less cavity dilation) and improved ejection fraction and stroke volume in 4 of 5 hearts. There was also a significant increase in E_RR, E^N_max, and E^S_max in the border regions, and an average increase in E^N_max and E^S_max within the infarct region in 4 of 5 hearts. These changes indicate some degree of recovery of function immediately after hydrogel injection, particularly in the peri-infarct border regions. Acute increases in wall thickness due to the nine 1 mL hydrogel injections within the central infarct region may have resulted in less desynchrony [41], which may have also improved contractile function at the border zone. We previously demonstrated that acute hydrogel delivery post-MI reduces LV end diastolic pressure [42], which has been shown to be associated with reduced mortality post-MI [4].

To further investigate the relationship between degree and extent of myocardial injury and cardiac strain, we correlated the percent of normal tissue in each AHA segment with strain measures in each segment post-MI on day 7. For all strain components in all hearts, strain appeared to decline with the percentage of infarcted tissue in each segment, with near-zero radial strain in fully infarcted segments. Slopes for the relationship between strain and percent viable tissue were similar at days 3 and 7 for E_ZZ and E_CC, but smaller on day 7 for E_RR, E^N_max, and E^S_max, indicating reduced compensatory hypercontractility in remote regions by day 7. Correlations between percent of normal tissue and strain were strongest for E_RR, E^N_max and E^S_max, indicating that these Lagrangian strain measures could be used to delineate the MI region in vivo with potentially higher resolution than would be provided by ^99mTc-Tetrofsomin SPECT imaging. Pislaru et al. [38] similarly demonstrated near-zero radial strain in the infarct region of pigs following coronary occlusion and reperfusion with 2D echocardiographic imaging (3% infarct vs. 26% remote). Stathogiannis et al. [40] showed a similar decrease in 2D circumferential (−10% ischemic vs. −19% normal) and longitudinal strains (−10% ischemic vs. −21% normal) in the scar area among patients with ischemic heart disease assessed from MRI using SuiteHEART. Our study adds to mounting evidence that high resolution strain measurements from in vivo images can delineate dysfunctional regions of the heart with potentially higher resolution than radiotracer imaging.

Using our ex vivo preparation we demonstrated that radial E_RR Lagrangian strains were largely compressive (−2.7%) and represented thinning of the myocardium during passive pressurization, while circumferential strain E_CC was primarily tensile (+8.5%) representing expansion along the circumference of the LV during pressurization. We also computed maximum normal (E^N_max: 16.9%) and shear (E^S_max: 15.5%) strains to evaluate strain independent of (R,C,Z) and changes of curvature of the left ventricle. These strains were positive and much larger in the apex than in other regions. The largest E_RC shear strains occur near the base of the heart, while E_RZ tended to be larger in magnitude in mid-cavity and apical regions. As would be expected, the normal strain pattern with pressurization is functionally the opposite of in vivo contraction. Measuring the strain response of the LV with cyclic pressurization allows examination of differences related to changes in geometry and stiffness given an appropriate constitutive model. Herein, we focused on simply analyzing the regional variations in strain, which will be biased by regional differences in both wall thickness and stiffness. Thus, we expect higher strains in thinner regions of the wall such as the apex and lower strains in thicker areas such as near papillary muscles or the attachment points of the RV in the septum. These variations were observed clearly in the circumferential strain fields, as seen in Figure 2b. The goal of the ex vivo regional strain analysis was to define changes in distensibility due to MI and acute intramyocardial delivery of hydrogels. For this analysis, segments with less than 50% normal tissue were designated as the infarct region, while segments with more 50% normal tissue were designated as the normal region. This method of classification was used rather than a transformation of existing SPECT maps because of significant baseline shape changes between in vivo and ex vivo imaging. During ex vivo testing, the infarcted regions with hydrogel had significantly less radial strain than the normal region, but more circumferential and long-axis strain. When compared to the control group, we noted greater radial strain only within the infarcted apex of the hydrogel group. Overall, intramyocardial gel delivery in the infarct region seemed to have partially normalized pressure-induced deformations in the infarct region. We also identified significantly larger tensile circumferential and long-axis strains within normal segments of the post-MI hydrogel group compared with control hearts. This suggests an alteration of pressure-induced changes in myocardial deformation in the remote regions of the ex vivo heart 7 days post-MI following acute intramyocardial delivery of gel in infarct region. Prior pre-clinical studies using 3D MR diffusion tensor imaging and tractography of excised ovine hearts post-MI demonstrated alterations in the myocardial fiber architecture with changes in the helix angle in the remote areas of the heart post-MI due to structural remodeling in these remote normal regions [43]. Therefore, the regional differences in strain and biomechanics that we observed in the remote normal myocardium post-MI may have been due to altered myofiber architecture associated with acute LV remodeling. The hypothesized alteration of remote fiber architecture in our study would need to be confirmed by future histological analyses. One additional metric that we measured that also indicated changes within the remote normal regions of the heart was LV wall thickness at ED. Post-MI day 3, LV wall thickness at ED was measured to increase by +1.7 mm in the infarcted region and +0.7 mm in normal regions. This can likely be attributed to inflammation and remodeling after injury as the dilation of the LV cavity should otherwise have led to a thinner wall at ED due to volume conservation. We observed small changes in wall thickness post-MI between 3 and 7 days. Thickness at ED decreased −0.6 mm in the infarcted region and −0.1 mm in normal regions, likely due to a reduction in inflammation, remodeling of the scar region, and continued dilation of the LV cavity. On day 7 post hydrogel delivery wall thickness at ED increased by +0.4 mm in the infarcted regions and +0.5 mm in normal regions. These homogeneous increases in ED wall thickness may simply reflect changes in loading conditions and the acute reduction in LV cavity volume at ED after hydrogel delivery, which due to volume conservation should lead to a thicker wall at ED.

While we have developed a unique approach to define changes in strain associated with MI and hydrogel delivery, several limitations should be acknowledged. First, our sample size was small at 5 animals, though we were able with this sample size to demonstrate significant strain changes due to MI and acute intramyocardial hydrogel delivery. Second, the current study did not include a control MI only group, therefore it is not possible to separate the effects of open-chest surgery from that of hydrogel delivery in the in vivo study or to separate the effects of post-MI remodeling from hydrogel delivery in the ex vivo study. We could address this issue in the future via additional evaluations at 7-day MI and open-chest surgery but without hydrogel injection. These experiments are currently being pursued in an ongoing parallel study [41]. Third, we employed a model of reperfused infarction that does not reflect changes associated with permanent coronary artery occlusion or the presence of reperfused infarction with residual ischemia. Nevertheless, this model reflects the more common clinical scenario. Fourth, hydrogel injection and analysis was in an acute setting to evaluate the initial biomechanical effect both in vivo and ex vivo after acute delivery of hydrogels. It would be important to study the effects of hydrogel delivery over a longer period, as local tissue remodeling and function likely change in response to hydrogel delivery. Fifth, we chose to use a cylindrical model for evaluating strain. We chose this system rather than an ellipsoid or prolate spheroidal coordinate system (which would be closer to LV geometry) because of noted differences in LV cavity length and apex geometry, as we did not want fitting differences to influence the magnitude of our strain measures. Since the purpose of this study was to measure strain changes due to MI and hydrogel injection, we considered it more important to have a simple, fixed coordinate system for comparison between hearts and over time. For this reason, the radial strains reported in this study are not the same as radial strain calculated along an endocardial normal and should not be taken as a measure of myocardial contraction near the apex. We have presented maximum normal E^N_max and shear E^S_max strains as a measure of strain independent of (R,C,Z) and changes in curvature, and these strains can be compared to other studies that may have chosen different coordinate systems. Lastly, use of CT imaging for in vivo assessment of regional strain is limited to approximately 10 frames per cardiac cycle due to rotations speeds of standard CT scanners. Our analysis was limited to evaluation of peak strain at an estimated end-systolic time point. Analysis of dynamic changes in strain over the cardiac cycle may offer additional advantage in the evaluation of post-MI remodeling and the effects of hydrogel delivery.

Several clinical studies have evaluated long-term functional benefits of intramyocardial delivery of alginates or hydrogels following myocardial infarction [12–14]. There are many versions of hydrogels that can be investigated, including those with altered mechanical properties and degradation or through the encapsulation and release of various drugs and cells.

5. CONCLUSIONS

We demonstrated the ability of contrast cineCT to track infarct expansion and early post-MI remodeling and were able to correlate these changes with regional variations in active and passive strain using in vivo and ex vivo imaging. Radial, circumferential, maximum principal, and maximum shear strains were reduced from baseline in the MI region, with very little initial change in most components of active strain in the remote normal regions. The observed pattern of strain within the infarct area was relatively stable over seven days post-MI but was associated with a progressive dilation of the LV and reduction in ejection fraction and stroke volume. We observed near-zero radial strains within the core MI regions, with gradual improvement in radial strain within peri-infarct border region by day 7. We found a significant negative linear correlation between regional radial, circumferential, and maximum principal/shear strains and percent of infarcted tissue, suggesting this approach might be used for infarct sizing. Hydrogel delivery resulted in an immediate improvement in ejection fraction and acute reduction in LV ED volume (less dilation). This was associated with significant increases to radial and maximum principal/shear strain within the peri-infarct border regions. A smaller strain improvement within the infarct and normal regions was also noted on average along with an improvement in stroke volume in 4 out of 5 hearts. During ex vivo testing, the infarct region appeared significantly stiffer radially compared with normal regions, although more compliant on average in the circumferential and long-axis directions. Normal remote regions in the post-MI hydrogel hearts also appeared more compliant when compared to the same regions of control hearts and might reflect changes in fiber architecture associated with early global post-MI remodeling. Further studies using this methodology will be needed to define the immediate and long-term mechanical effects of intramyocardial hydrogel delivery post-MI and for optimization of gel delivery. This approach may be optimized by use of iodinated hydrogels that could provide an exact localization of the gels within the infarct and peri-infarct area and define the local mechanical effects of intramyocardial gel delivery.

Figure 3,

Serial changes in color-coded in vivo and ex vivo strains. Shown are changes in LV radial and circumferential strain fields in short and long axis views for infarcted pig. Shown from left to right: Baseline, Post-MI Day 3, Post-MI Day 7, Post-Gel Day 7, Post-Gel Ex Vivo Day 7. From top to bottom: radial strain midcavity view, radial strain long axis view at 90°, circumferential strain midcavity view, circumferential strain long axis view at 90°.

The infarct area is defined by dashed circle.

Table 4:

Statistical results for repeated measures ANOVA test of the difference between in vivo strain in the normal, border, and infarct regions, masked in 3D within the myocardium wall from registered SPECT perfusion images (n=5 pigs). Changes of p < 0.1 are in bold, and borderline significant changes of p > 0.05 are also italicized.

ANOVA Strains	Region 1 Compare	Region 2 Compare	Mean Difference	Lower 95% CI	Upper 95% CI	P-Value Pairwise	F-Value Group	P-Value Group
EZZ	Border	Infarct	−0.041	−0.078	−0.004	0.036	37.867	0.00008
	Border	Normal	0.061	0.028	0.095	0.006
	Normal	Infarct	−0.102	−0.156	−0.049	0.005
ERR	Border	Infarct	0.117	0.033	0.202	0.017	56.086	0.00002
	Border	Normal	−0.204	−0.320	−0.088	0.007
	Normal	Infarct	0.321	0.198	0.445	0.002
ECC	Border	Infarct	−0.051	−0.094	−0.008	0.029	30.293	0.00019
	Border	Normal	0.030	−0.004	0.064	0.070
	Normal	Infarct	−0.081	−0.116	−0.046	0.003
ENmax	Border	Infarct	0.094	−0.008	0.196	0.065	32.508	0.00014
	Border	Normal	−0.176	−0.318	−0.035	0.025
	Normal	Infarct	0.270	0.107	0.433	0.009
ESmax	Border	Infarct	0.068	0.004	0.132	0.041	27.303	0.00466
	Border	Normal	−0.110	−0.191	−0.028	0.019
	Normal	Infarct	0.178	0.088	0.268	0.005

Highlights:

• Developed novel platform to assess left ventricle deformation in vivo and ex vivo

• Local cardiac strains decreased progressively over 7 days in infarcted regions

• Measured linear correlation between local degree of infarction and strain

• Hydrogel injection improved strain in regions bordering the infarct

• Hydrogel injection resulted in less ventricle dilation and larger ejection fraction

6. GLOSSARY

3D

Three Dimensional

AHA

American Heart Association

ANOVA

Analysis of Variance

bpm

beats per minute

CineCT

Cine Computed Tomography

cpm

cycles per minute

ED

End-Diastolic

ES

End-Systolic

E_CC

3D Lagrangian circumferential strain

E_RR

3D Lagrangian radial strain

E_ZZ

3D Lagrangian LV long-axis strain

E^N_max

3D maximum principal strain

E^S_max

3D maximum shear strain

HF

Heart Failure

LV

Left Ventricle

MI

Myocardial Infarction

RV

Right Ventricle

SPECT

Single Photon Emission Computed Tomography