Twenty-one male Sprague–Dawley rats, weighing 213.7 ± 14.9 g, were purchased from the Laboratory Animal Center, Southern Medical University, China, and were kept in metal cages under a 12-h light–dark cycle with a standard rat diet and water ad libitum. These rats were randomly divided into three groups (seven rats per group): a sham group, an MI group, and an ischemia–reperfusion (IR) group. Before conducting the surgery, ethical approval was obtained from the Institutional Animal Care and Use Committee (IACUC) of Southern Medical University. Experiments on rats were performed according to the Guidelines for the Care of Laboratory Animals of the National Institutes of Health.

Rats in all groups were anesthetized with pentobarbital sodium injection (35 mg/kg intraperitoneally) (Sigma-Aldrich Corp., St. Louis, MO, United States). Hair at the rats’ chest and neck was removed using a depilatory cream. The animal’s respiration was controlled by tracheal intubation using a rodent ventilator (HX-101E, Chengdu Techman Software Co., Ltd., Sichuan, Chengdu, China). After left-sided thoracotomy, the LAD coronary artery of each rat in the MI group was permanently ligated. In the IR group, a plastic tube with the diameter of 2 mm was simultaneously ligated with the LAD coronary artery of each rat. Furthermore, after a 30-min ligation, the tube and ligation were removed for ischemia–reperfusion (Xu et al., 2014; Lindsey et al., 2021). Rats in the sham group underwent a sham operation of the LAD coronary artery without ligation. Limb-lead ECG was obtained throughout each surgery using a biological function experiment system (H-420E, Chengdu Techman Software Co., Ltd., Sichuan, Chengdu, China). ST-segment elevation was regarded as the indicator of successful ischemia. After the animal model establishment, the thoracic cavity was sewed layer by layer from ribs to skin, and the ventilator was removed. Rats were placed on a heating blanket until recovery from anesthesia. Finally, the rats were returned to metal cages with a standard rat diet and water ad libitum for later examination. Ultrasound scanning was scheduled for all rats at 1 day (acute myocardial infarction phase) and 28 days (remodeling phase) after the surgery (Holmes et al., 2005); a total of six groups of data are represented as Sham1, Sham28, MI1, MI28, IR1, and IR28.

The Vevo2100 ultrasound imaging system (VisualSonics Inc., Canada) and the MS-250 probe (VisualSonics Inc., Canada) were used in this study for RF data acquisition. The probe has a center frequency of 21 MHz, a geometric focus of 15 mm, and a bandwidth of 13–24 MHz. ECG signals can be simultaneously monitored during the scanning.

All rats were first anesthetized with pentobarbital sodium injection (35 mg/kg intraperitoneally) (Solarbio Inc., Beijing, China). The limbs of the animals were fixed to the physiological monitoring platform, which is part of the imaging system, in a supine position. The parasternal long-axis (PLAX) view of the heart was scanned, and the 32-bit in-phase and quadrature (IQ) data of a total number of 100 frames of each scanning were stored and exported (Figure 1). RF data were then reconstructed from IQ data using custom software (VisualSonics Inc.).

The displacement of the myocardium was calculated using a VNCC-based global motion tracking algorithm proposed in our previous study (Wang et al., 2019). Assume that F P r e and F P o s t are two successive RF frames from the sequence of the myocardial systole period which are used for motion tracking; F P r e x , y is an RF sample value located at the x t h row and y t h column in F P r e . By using the traditional NCC-based motion tracking method, the displacement of each F P r e x , y was calculated as follows: first, a matching block (MB) centered by F P r e x , y was defined in F P r e , and several candidate blocks (CBs) with the same size as MB were defined in a larger region centered by current MB in F P o s t . Then, the NCC values were calculated between the current MB and each CB. Finally, the CB with the maximum NCC value was regarded as having the highest similarity to MB, and its location shifts regarding MB represent the calculated displacement results. Compared with the traditional motion tracking method using NCC, the VNCC method used here is able to obtain the NCC values for all RF samples simultaneously rather than sample by sample. Thus, this method could significantly reduce the computational time.

Due to the lower resolution in the lateral direction of RF data, we emphasized axial displacement in this study. Incremental axial displacements between each pair of successive RF frames from the systole period were calculated and then cumulated to obtain the cumulative displacement in the axial direction ( C D A ) which was used for segmental and transmural myocardial motion analysis.

The matching block size and overlap are key algorithmic parameters that determine the performance of VNCC in motion tracking. A matching block size of 4 λ × 7 pitches that was illustrated to be more appropriate in a simulated hybrid deformation scenario (Li et al., 2016) was used in this study. Here, λ is the wavelength calculated using the center frequency of the ultrasonic probe, which is 21 MHz in this study.

The values of the overlap between successive windows could be set from 10% to 99.9%. The 99.9% overlap represents that the window shift of the successive matching window is 1 RF sample. In this study, the optical value of the overlap was determined according to the quality and accuracy evaluation of the displacement calculation with different test overlap values. The quality of calculated displacement maps was evaluated using SNR. Furthermore, a displacement-compensated cross-correlation coefficient (DCCCC) was proposed for evaluating the accuracy of displacement calculation. SNR is defined as S N R = D R O I σ R O I , (1) where D R O I and σ R O I are the average and standard deviation of displacement values in the region of interest (ROI), which is the myocardium region in this study, as shown in Figure 2F.

The axial ( D A ) and lateral ( D L ) displacement maps between the randomly selected successive F P r e and F P o s t from each group were calculated using VNCC with a test overlap value. So, the compensated F P o s t is obtained as each RF sample F P o s t x , y moving − D A x , y and − D L x , y in the axial and lateral directions, respectively. F P o s t with the displacement compensation can be described as F P o s t C : F P o s t C x − D A x , y , y − D L x , y = F P o s t x , y . (2)

Then, the cross-correlation between F P r e and F P o s t C is calculated and defined as DCCCC: D C C C C = ∑ i , j ∈ R O I F p r e i , j − f i j ¯ F p o s t C i , j − g i j ¯ ∑ i , j ∈ R O I F p r e i , j − f i j ¯ 2 ∑ i , j ∈ R O I F p o s t C i , j − g i j ¯ 2 , (3) where f i j ¯ and g i j ¯ are the mean values of RF data of ROI in F p r e and F p o s t C , respectively. The higher the DCCCC value, the more accurate the displacement estimation. In addition, the DCCCC is a ground-truth-free and angle-independent quality metric.

After the SNR and DCCCC calculations with all the overlap values for all groups were completed, the optimal value of the overlap could be selected with the compromise between the SNR and DCCCC values.

Before the segmental and transmural myocardial motion analysis, the semiautomatic regional and layer-specific segmentation of the myocardium was performed. Figure 2 shows the flowchart of the regional and layer-specific segmentation of the myocardium. Five steps of regional segmentation were performed in this study:

Step 1: Myocardial border selection and interpolation. The key points on the endocardial border (Endo-B) and epicardial border (Epi-B) of LV were manually selected on the B-mode image (Figure 2B) and then interpolated to track Endo-B and Epi-B (Figure 2C). The manual identification of myocardial borders was performed by a professional sonographer.

Step 2: Apical point definition. The basal middle points (BMps) of the endocardial border (Endo-BMp) were first defined as the middle point between the endocardial anterior basal point and posterior basal point (Figure 2D). Then, the distances between the Endo-BMp and the points of the Endo-B were calculated, and the farthest point from the Endo-BMp was defined as the apical point of the endocardial border (Endo-Ap) (Figure 2D). Similarly, the basal middle point of the epicardial border (Epi-BMp) and epicardial apical point (Epi-Ap) were determined.

Step 3: Myocardial border division. Endo-B and Epi-B were divided into two sub-borders by Endo-Ap and Epi-Ap, respectively. Then, each sub-border was further divided into three equal-length sections (Figure 2E).

Step 4: Segmental contour connection. The contours of the apical septal wall (A-SEP), middle anteroseptal wall (M-ANT-SEP), basal anteroseptal wall (B-ANT-SEP), apical lateral wall (A-LAT), middle inferolateral wall (M-INF-LAT), and basal inferolateral wall (B-INF-LAT) were obtained (Figure 2F).

Step 5: Cross-wall line division. In order to analyze the transmural systolic function of LV, the equal stratification of the myocardium was performed (Voigt et al., 2015). Each segment was further divided into three equal-width layers representing the endocardial (EndoL), middle (MidL) and epicardial (EpiL) layers of the myocardium, as shown in Figure 2G. Furthermore, each segment was divided into 10 layers for a more comprehensive analysis of transmural motion of LV (Figure 2H).

The average C D A in each layer (Figure 2H) was calculated as C D A L n , n = 1,2,3 ⋯ 10 , and a novel angle-independent quantitative biomarker called the transmural motion index (TMI) was proposed for the estimation of the cross-wall transmural motion gradient and analysis of systolic function. The TMI value of each LV segment was derived as the first derivative of the first-order polynomial interpolation of C D A L n , n = 1,2,3 ⋯ 10 .

MATLAB (R2021b, MathWorks Inc., Natick, MA) was used to program all the codes of displacement estimation and myocardial segmentation on a PC workstation (Intel Core i5-6500 CPU, 3.20 GHz, 12 GB RAM).

After the ultrasound examination, the rats in all groups were euthanized with pentobarbital sodium injection (40 mg/kg intraperitoneally) (Solarbio Inc., Beijing, China) following abdominal aorta exsanguination. The hearts were excised, harvested, and fixed in 10% formaldehyde solution and then sent to the Guangzhou University of Chinese Medicine for histological measurements. Paraffin blocks were prepared for each dehydrated heart specimen. Three pieces of myocardium were cut from the apex to the base of the heart and were stained using Masson. The stained myocardium slices were observed using an optical microscope (BX63, Olympus Inc., Tokyo, Japan).

The Mann–Whitney U-test was used to investigate the statistical differences in C D A and TMI among different groups and ultrasound scanning time points. Inter-group comparisons were performed among the sham, MI, and IR groups from 1 day to 28 days after the surgery, respectively. Inter-time comparisons were performed among sham, MI, and IR groups from 1 day to 28 days of acquisitions. All statistical analyses were performed using SPSS (version 26, IBM, Armonk, NY, United States), and a p-value < 0.05 was considered a significant difference.

The optimal value of the overlap between successive windows was determined as the compromise between the SNR and DCCCC values. Figure 3A shows that the stable SNR values reach above 0.92 for different overlap values and are not able to provide a comparable optimal selection of the overlap. However, Figure 3B shows an obvious increase in the DCCCC value with the overlap value ranging from 30% to 50% and then a slightly stable DCCCC increase with the overlap value increasing from 50% to 80%, but a sudden increase appears at 90% and then drops immediately. So, this sudden increase might be an “outlier.” In view of a high computational cost by a higher overlap value, the optimal overlap value was 80% in this study.

In this study, we calculated the average C D A of each myocardial segment. Boxplots of the six segmental values from the sham, MI, and IR groups obtained between 1 day and 28 days after the surgery are shown in Figure 4, where the positive and negative C D A values, respectively, represent a downward and an upward movement.

As shown in Figure 10, the cardiomyocytes of the specimens from the sham group showed red staining and were arranged in order, indicating a healthy myocardium. Infarcted regions were found in the anterior wall and septum of the MI group, especially the apical and middle myocardium, where large regions of fibrous tissue appeared. Additionally, epicardial hyperplasia was observed in the basal myocardium of the MI group. Less severe infarction was noticed in the apex of the IR group with a relatively small region of fibrous tissue. Some epicardial hyperplasia was also observed in the middle and basal myocardia of the IR group.