Systolic myocardial volume gain in dilated, hypertrophied and normal heart. CMR study

WarpPINN: Cine-MR image registration with physics-informed neural networks

The diagnosis of heart failure usually includes a global functional assessment, such as ejection fraction measured by magnetic resonance imaging. However, these metrics have low discriminate power to distinguish different cardiomyopathies, which may not affect the global function of the heart. Quantifying local deformations in the form of cardiac strain can provide helpful information, but it remains a challenge. In this work, we introduce WarpPINN, a physics-informed neural network to perform image registration to obtain local metrics of heart deformation. We apply this method to cine magnetic resonance images to estimate the motion during the cardiac cycle. We inform our neural network of the near-incompressibility of cardiac tissue by penalizing the Jacobian of the deformation field. The loss function has two components: an intensity-based similarity term between the reference and the warped template images, and a regularizer that represents the hyperelastic behavior of the tissue. The architecture of the neural network allows us to easily compute the strain via automatic differentiation to assess cardiac activity. We use Fourier feature mappings to overcome the spectral bias of neural networks, allowing us to capture discontinuities in the strain field. The algorithm is tested on synthetic examples and on a cine SSFP MRI benchmark of 15 healthy volunteers, where it is trained to learn the deformation mapping of each case. We outperform current methodologies in landmark tracking and provide physiological strain estimations in the radial and circumferential directions. WarpPINN provides precise measurements of local cardiac deformations that can be used for a better diagnosis of heart failure and can be used for general image registration tasks. Source code is available at https://github.com/fsahli/WarpPINN.

Comparison of mitral regurgitation severity assessments based on magnetic resonance imaging and echocardiography in patients with hypertrophic cardiomyopathy

Mitral regurgitation (MR), which is one of the factors responsible for heart failure symptoms and the development of atrial fibrillation, is an important feature of hypertrophic cardiomyopathy (HCM), and its presence affects which treatment options are chosen. Although cardiac magnetic resonance imaging (MRI) is considered the reference standard for assessing the regurgitant volume (RV) and fraction (RF), echocardiography is the most common method for assessing MR severity. Accordingly, the aim of this study was to compare the results of echocardiography and cardiac MRI for assessing MR severity in a cohort of patients with HCM. MR severity was assessed in 53 patients using cardiac MRI by determining the mitral RV (MRV) and mitral RF (MRF). The results were graded according to thresholds recommended in current guidelines. MR severity assessed by echocardiography was graded by integrating indices of severity. Greater than mild MR, as assessed using echocardiography, was present in 22 patients (41.5%) with HCM and in none of the control patients (p = 0.001). In all, 31 patients (58.5%) had no more than mild MR. When MR severity was assessed using different methods, either moderate (kappa = 0.44, 95% confidence interval = 0.21–0.67), poor or no agreement was found between MRI-derived and echocardiography-derived grades. HCM patients with echocardiography-derived moderate and severe MR had similar median MRVs and MRFs (p = 0.59 and p = 0.11, respectively). In HCM patients, cardiac MRI and echocardiography were at most in modest agreement in assessing MR severity. Importantly, echocardiography-derived moderate and severe MR were not distinguishable by either MRV or MRF.

Predictors of right ventricular function and size in patients with hypertrophic cardiomyopathy

We investigated factors associated with right ventricular (RV) function and size in hypertrophic cardiomyopathy (HCM) patients. Two hundred fifty-three consecutive HCM patients and 20 healthy volunteers underwent cardiac magnetic resonance examination. In addition to measuring RV function (ejection fraction—RVEF) and size (end-diastolic volume—RVEDV), each image was inspected for the presence of RV and left ventricular (LV) hypertrophy, and the maximal wall thickness of the left and right ventricles was recorded. HCM patients had higher RVEF and lower RVEDV than healthy volunteers and similar RV mass. The mean RV wall thickness was higher in HCM patients than in controls. LV late gadolinium enhancement (LGE) was present in 89.7% of patients, and RV LGE was present in 3.1% of patients (p < 0.0001). Univariate and multivariable analyses revealed that LVEF, peak LV outflow tract gradient, LV LGE, maximal LV wall thickness, and tricuspid regurgitation (TR) volume by magnetic resonance imaging were positive predictors of RVEF. In addition to TR volume, the only independent predictor of RVEF < 45% was LVEF (odds ratio = 0.80, 95% confidence interval 0.67–0.95). Multivariable analysis revealed that LVEDV and TR volume were positive predictors of RVEDV, whereas negative predictors were RVEF, maximal RV wall thickness, LV LGE, and age. Neither estimated systolic pulmonary artery pressure nor TR grade by echocardiography proved to be predictors of RVEF. There were no differences in either the maximal RV wall thickness or the maximal left ventricular (LV) wall thickness in patients stratified according to NYHA functional class (p = 0.93 and p = 0.15, respectively). There were no differences in mean RV wall thickness in patients categorised based on the number of clinical risk factors for sudden cardiac death (SCD), i.e., non-sustained ventricular tachycardia, family history of SCD, or unexplained syncope (p = 0.79). On the other hand, there was a weak positive association between RV hypertrophy and the estimated probability of SCD at 5 years (rho = 0.16, p = 0.01). RV systolic dysfunction measured as decreased RVEF was uncommon in HCM and was associated with poor LV systolic function. LV also had a significant impact on RV size.

Cardiac MRI demonstrates compressibility in healthy myocardium but not in myocardium with reduced ejection fraction

BACKGROUND

The professional guidelines assume that the myocardial volume in systole (MVs) is equal to that in diastole (MVd), despite some limited evidence that points to the contrary. The aim of this manuscript is to determine whether this is true in healthy myocardium using gold standard cardiac MRI, as well as transthoracic echocardiography (TTE). The secondary aim is to determine whether there are similar MV changes in patients with heart failure with reduced ejection fraction (HFrEF).

METHOD

A prospectively derived cohort at Mayo Clinic of 115 adult subjects (mean age 42.8 years, 58% female) with no cardiac risk factors was identified. Cardiac MRI was obtained on all 115 patients, 51 of whom also consented to a TTE. MRI from a retrospectively derived cohort of 50 HFrEF patients was also collected. MVs and MVd was calculated using standard approaches with inclusion of the papillary muscles.

RESULTS

In the healthy population, MRI demonstrated MVs/MVd = 0.87 (SD 0.04) and TTE demonstrated MVs/MVd = 0.79 (SD 0.07), suggesting compressibility (p < 0.0001). In the 51 healthy patients who received both imaging modalities, MVs/MVd was 8.0% higher in MRI than TTE (p < 0.0001), but both modalities had MVs/MVd < 1 (p < 0.0001). A Bland-Altman plot demonstrated that as the mean MVs/MVd increases, the difference in MVs/MVd MRI-TTE declines (r = -0.53, p < 0.0001). However, in HFrEF populations, MVs/MVd = 1.01 (0.03), suggesting myocardial incompressibility.

CONCLUSION

Contrary to currently accepted standards, healthy myocardium is compressible but HFrEF myocardium is incompressible. The ratio MVs/MVd merits further study in an expanded normal cohort and in disease states.

Systolic-to-diastolic myocardial volume ratio as a novel imaging marker of cardiomyopathy

BACKGROUND

In patients with normal left ventricular ejection fraction, it may be difficult to distinguish between the normal and diseased heart. Novel assessments of ventricular function, such as extracellular volume imaging, myocardial perfusion imaging and myocardial contraction fraction are emerging to better assess disease burden in these cases. This study endeavored to determine whether the ratio of myocardial volume in systole to myocardial volume in diastole (MVs/MVd), differs between normal hearts and those with disease states characterized by normal ejection fraction.

METHOD

Consecutive patients from 2008 to 2018 with hypertrophic cardiomyopathy (HCM), cardiac amyloidosis, and heart failure with preserved ejection fraction (HFpEF) who underwent cardiac magnetic resonance imaging (MRI) were selected for inclusion, along with a sex- and age-matched cohort of normal volunteers who also underwent cardiac MRI. Manual tracings were performed on each MRI to calculate MVs/MVd, which was then compared across subgroups.

RESULTS

Included were 50 patients with HCM, 50 patients with cardiac amyloidosis, 26 patients with HFpEF, and 30 normal subjects. Age was 54.1 years (SD 16.7); mean MVs/MVd was 0.88 (SD 0.04) in the normal subgroup, 1.03 (SD 0.06) in HCM patients, 1.03 (SD 0.06) in cardiac amyloidosis patients, and 0.97 (SD 0.02) in HFpEF patients, with all pathology subgroups different from the normal subgroup (p < .0001 for each). The ratio of MVs/MVd discriminated diseased from normal with c statistic 0.989 (p < .001).

CONCLUSIONS

This study suggests that a novel and easily-captured metric of ventricular function, MVs/MVd, can differentiate normal ventricular function from multiple cardiomyopathies with normal ejection fractions.