This study evaluated commercially available 4D-US in a mouse model of myocardial infarction 1 week and 4 weeks after CAL and compared 4D-US endpoints to 2D-US, M-mode, CMR, and histologic parameters. The mouse CAL model presents a challenge for evaluation using traditional echocardiography, given that wall motion abnormalities and dyssynchronous contraction can cause abnormal LV geometry. Traditional M-mode and 2D-US assume ellipsoid geometry for volumetric measurements6. Following CAL, mice typically have a hyperdynamic base, static apex, and increased sphericity that limit the effectiveness of the geometric assumptions17,19−21. 4D-US allows for real-time image analysis at multiple planes encompassing the LV that is gated by ECG and respiration13,14 to minimize motion artifact during image reconstruction The result is a data-driven volumetric model that eliminates spatial and temporal assumptions about LV morphology. Our data demonstrate that for calculation of EF and ESV, 4D-US provides closer agreement when compared to benchmark CMR than other ECHO modalities 4 weeks after CAL. Additionally, we show that 4D-US has lower inter-user variabilty when measuring ventricular volumes at 4 weeks. 4D-US has the highest correlation when comparing volumetric measurements taken at 1 week to those taken at 4 week when compared to other ECHO modalities. Finally, 4D-US allows for simplified evaluation of WMSI, which correlates well with scar size by histologic analysis.
We find that 4D-US has lower percentage bias by Bland-Altman analysis than 2D-US and M-mode compared to CMR for the evaluation of ESV and EF at 4 weeks (Fig. 2), suggesting more accurate assessment of volumetric parameters. 4D-US is outperformed by 2D-US in the evaluation of EDV at this time point. However, across the three endpoints evaluated (EDV, ESV, and EF), 4D-US is the most reliable modality in terms of bias and 95% level of agreement. 4D-US has much lower bias than M-mode, which demonstrated > 40% bias in the estimation of ESV. 2D-US bias was improved compared to M-mode, but is characterized by the highest inter-user variability as demonstrated by lowest ICC for the evaluation of EDV, ESV, and EF (Fig. 2C). Inter-user variabilty is lowest using 4D-US for evaluation of EDV and ESV, but was mildly outperformed by M-Mode when evaluating EF (Fig. 2C). In total, our data is consistent with other studies that finds that 4D-US provides more accuracy and less inter-user variability when comparing to CMR than 2D-US or M-mode values, making it the optimal modality for surgical animal studies.
4D-US uniformly underestimates EDV and ESV when compared to CMR, though the changes were more pronounced in EDV (Fig. 2B, Table 2). The 4D-US measurements may suffer limitations in surgical mouse models, as the base of the heart is difficult to clearly visualize given the overlying scar following thoracotomy required in the CAL surgery, and may result in lower total volumes. However, a similar underestimation of volumes was previously noted in evaluation of round “phantoms” using identical software15, suggesting that a systemic underestimation of volumes may be present in this modality. Regardless of this underestimation, the modality still represents an improvement over other traditional ECHO modes.
Recent work by Russo et al. has shown that step-wise short-axis ECHO imaging performed on CAL mice strongly correlates with CMR17, providing reproducible volumetric evaluations nearly on par with CMR, but at a fraction of the time and cost required for CMR. Our data support the increased value of this step-wise approach, but incorporate ECG- and respiratory-gating and the higher resolution of the Vevo-3100 for enhanced temporal and spatial resoluation. Soepriatna et al. previously demonstrated similar benefits of gated 4D-US in infarcted mice using manual 3-D reconstruction throughout the cardiac cycle14. While that study excellently demonstrated the value of the hardware and imaging through the cardiac cycle, here we use an easily-accessible commercial system, which utilizes edge-tracing software to simplify volumetric calculation for users, resulting in both rapid analysis and low inter-user variability.
An additional benefit of the 4D-US imaging modality is the ability to quickly and reproducibly quantify scar characteristics using clinical scales of wall-motion. Previous works have quantified scar size using ECHO based on wall-motion abnormalities from 2D-US and 3-d reconstruction, demonstrating good comparison between ECHO estimations and histologic scar size22,23. We evaluated WMSI using a 16-segment model across three short axis views used in the clinical settings as previously described24,25. We found that WMSI correlates strongly with histologic analysis of scar size (r = 0.77, Fig. 5), while gadolinium-enhanced CMR correlated very strongly (r = 0.90). The 4D-US method allows for easy standardization on the WMSI across animals, as the automated 0.2 mm step function utilized in this paper can be utilized to standardize the height of short-axis image. In this study, we used three short-axis images taken 1 mm, 3 mm, and 5 mm from the apex to quantify WMSI. It should be noted that his model utilizes permanent coronary occlusion. Models of transient occlusion may be complicated by myocardial stunning, which may impair the value of WMSI and its correlation to scar size.
A key point of this study is the comparison of 1 week and 4 weeks with 4D-US, 2D-US, and M-mode to assess the predictive value of early imaging vs 4 week imaging. A major challenge following CAL is that cardiac remodeling is a dynamic process, characterized by progressive LV dilation, scar thinning, and compensatory hypertrophy of viable tissue9,19,26. Early imaging after CAL in the mouse model is limited by free air in the chest cavity and inflammation at the thoracotomy site that reduces the image quality of ECHO images. These post-surgical changes compound the limitations of 2D-US and M-Mode, as slight errors in 1-dimensional or 2-dimensional analysis cause major deviations in volumetric calculations. 4D-US does not rely on these geometric expansions. We find that 4D-US has highest correlation across volumetric measurements when comparing 1 week and 4 week values.
Potential caveats
A potential confounder in the study is the use of multiple genotypes on a single background. Specifically, we utilized transgenic mice for the evaluation of volumetric measurements, pooling data from α-MHC-Cre (+) x TFAM-flox and α-MHC-Cre (-) x TFAM-flox mice. Because we found no difference between survival, scar size, baseline and 4 week EDV, ESV, or EF between groups (Supplemental Fig. 1, Supplemental Table 1), we included both models for this data set. We believe that the transgenic background does not limit the correlations drawn in this paper between volumetric studies and agrees with the NIH’s effort to reduce animal suffering by utilizing mice readily available in our lab.