A completely automated pipeline for 3D reconstruction of human heart from 2D cine magnetic resonance slices

Cardiac magnetic resonance (CMR) imaging is a valuable modality in the diagnosis and characterization of cardiovascular diseases, since it can identify abnormalities in structure and function of the myocardium non-invasively and without the need for ionizing radiation. However, in clinical practice, it is commonly acquired as a collection of separated and independent 2D image planes, which limits its accuracy in 3D analysis. This paper presents a completely automated pipeline for generating patient-specific 3D biventricular heart models from cine magnetic resonance (MR) slices. Our pipeline automatically selects the relevant cine MR images, segments them using a deep learning-based method to extract the heart contours, and aligns the contours in 3D space correcting possible misalignments due to breathing or subject motion first using the intensity and contours information from the cine data and next with the help of a statistical shape model. Finally, the sparse 3D representation of the contours is used to generate a smooth 3D biventricular mesh. The computational pipeline is applied and evaluated in a CMR dataset of 20 healthy subjects. Our results show an average reduction of misalignment artefacts from 1.82 ± 1.60 mm to 0.72 ± 0.73 mm over 20 subjects, in terms of distance from the final reconstructed mesh. The high-resolution 3D biventricular meshes obtained with our computational pipeline are used for simulations of electrical activation patterns, showing agreement with non-invasive electrocardiographic imaging. The automatic methodologies presented here for patient-specific MR imaging-based 3D biventricular representations contribute to the efficient realization of precision medicine, enabling the enhanced interpretability of clinical data, the digital twin vision through patient-specific image-based modelling and simulation, and augmented reality applications. This article is part of the theme issue ‘Advanced computation in cardiovascular physiology: new challenges and opportunities’.

Classification of Cardiomyopathies from MR Cine Images Using Convolutional Neural Network with Transfer Learning

The automatic classification of various types of cardiomyopathies is desirable but has never been performed using a convolutional neural network (CNN). The purpose of this study was to evaluate currently available CNN models to classify cine magnetic resonance (cine-MR) images of cardiomyopathies. Method: Diastolic and systolic frames of 1200 cine-MR sequences of three categories of subjects (395 normal, 411 hypertrophic cardiomyopathy, and 394 dilated cardiomyopathy) were selected, preprocessed, and labeled. Pretrained, fine-tuned deep learning models (VGG) were used for image classification (sixfold cross-validation and double split testing with hold-out data). The heat activation map algorithm (Grad-CAM) was applied to reveal salient pixel areas leading to the classification. Results: The diastolic–systolic dual-input concatenated VGG model cross-validation accuracy was 0.982 ± 0.009. Summed confusion matrices showed that, for the 1200 inputs, the VGG model led to 22 errors. The classification of a 227-input validation group, carried out by an experienced radiologist and cardiologist, led to a similar number of discrepancies. The image preparation process led to 5% accuracy improvement as compared to nonprepared images. Grad-CAM heat activation maps showed that most misclassifications occurred when extracardiac location caught the attention of the network. Conclusions: CNN networks are very well suited and are 98% accurate for the classification of cardiomyopathies, regardless of the imaging plane, when both diastolic and systolic frames are incorporated. Misclassification is in the same range as inter-observer discrepancies in experienced human readers.

Myocardial Deformation in Fontan Patients Assessed by Cardiac Magnetic Resonance Feature Tracking: Correlation with Function, Clinical Course, and Biomarkers

Cardiac MR (CMR) is a standard modality for assessing ventricular function of single ventricles. CMR feature-tracking (CMR-FT) is a novel application enabling strain measurement on cine MR images and is used in patients with congenital heart diseases. We sought to assess the feasibility of CMR-FT in Fontan patients and analyze the correlation between CMR-FT strain values and conventional CMR volumetric parameters, clinical findings, and biomarkers. Global circumferential (GCS) and longitudinal (GLS) strain were retrospectively measured by CMR-FT on Steady-State Free Precession cine images. Data regarding post-operative course at Fontan operation, and medication, exercise capacity, invasive hemodynamics, and blood biomarkers at a time interval ± 6 months from CMR were collected. Forty-seven patients underwent CMR 11 ± 6 years after the Fontan operation; age at CMR was 15 ± 7 years. End-diastolic volume (EDV) of the SV was 93 ± 37 ml/m2, end-systolic volume (ESV) was 46 ± 23 ml/m2, and ejection fraction (EF) was 51 ± 11%. Twenty (42%) patients had a single right ventricle (SRV). In single left ventricle (SLV), GCS was higher (p < 0.001), but GLS was lower (p = 0.04) than in SRV. GCS correlated positively with EDV (p = 0.005), ESV (p < 0.001), and EF (p ≤ 0.0001). GLS correlated positively with EF (p = 0.002), but not with ventricular volumes. Impaired GCS correlated with decreased ventricular function (p = 0.03) and atrioventricular valve regurgitation (p = 0.04) at echocardiography, direct atriopulmonary connection (p = 0.02), post-operative complications (p = 0.05), and presence of a rudimentary ventricle (p = 0.01). A reduced GCS was associated with increased NT-pro-BNP (p = 0.05). Myocardial deformation can be measured by CMR-FT in Fontan patients. SLVs have higher GCS, but lower GLS than SRVs. GCS correlates with ventricular volumes and EF, whereas GLS correlates with EF only. Myocardial deformation shows a relationship with several clinical parameters and NT-pro-BNP.

Cine MR-Based Radiomics to Predict Myocardial Segments With Infarction

Background: Although myocardial infarction (MI) can be assessed quantitatively and qualitatively by using late gadolinium-enhanced (LGE) cardiovascular magnetic resonance (CMR) imaging, intravenous administration of gadolinium can expose patients to high risk of nephrogenic systemic fibrosis, especially in those with cardiovascular diseases. The purpose of this study is to harness cine CMR-based radiomics for predicting MI without introducing gadolinium.Methods: In this retrospective study, we included 48 patients with acute myocardial infarction (AMI) confirmed by later gadolinium enhancement (LGE) at CMR. CMR examinations were performed within 2 to 6 days after PCI. According to the LGE, each myocardial segment was dichotomized into with and without MI. Radiomic features of myocardial segments were extracted from cine CMR images and the myocardial segments were divided into training and validation sets randomly at a ratio of 0.7:0.3. Pearson correlation and Mann-Whitney U rank test were used to eliminate redundant and irrelevant features. A least absolute shrinkage and selection operator (LASSO) algorithm was used for features selection in the training set. Radiomic signatures were constructed in both the training and validation sets and its predictive performance was assessed using area under the cure of receiver operating characteristic (AUC-ROC).Results：Of 768 myocardial segments in the 48 patients, there were 291 (38%) segments with MI and 477 (62%) segments without MI. After univariate analysis, there were 22 RFs related to MI with statistical significance. LASSO regression selected 18 RFs for radiomics signature builting. AUC-ROC of radiomic signatures in prediction of segments with MI was 0.74(95% CI:0.69-0.78)and 0.68 (95%CI: 0.60-0.75) in the training and validation sets, respectively. The difference was not statistically significant (p=0.14).Conclusion: Cine MR-based radiomics signature can achieve a good prediction performance for MI, which showed the potential to be a promising imaging biomarker for MI without the administration of contrast agent.

CMR feature tracking cloud assess right ventricular functional reserve with pulmonary arterial hypertension

Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): JSPS KAKENHI, Grant-in-Aid for Early-Career Scientists. Background Pulmonary arterial hypertension (PAH) remains a fatal disorder characterized by elevated pulmonary arterial pressure. Survival of the patients with PAH is determined from right ventricular (RV) function. CMR has become an attractive modality for following up and providing prognosis in such patients, and CMR feature tracking has been used as a newer useful parameter to assess RV function. However, it has not yet been determined whether CMR feature tracking can assess RV functional reserve in patients with PAH. Purpose We investigated whether CMR feature tracking can estimate RV functional reserve using a rat model with PAH. Methods Rats were received injections with monocrotaline (MCT-rats, n = 19) or solvent (Ctr-rats, n = 5). Four weeks after the injections, we performed CMR on 7-T MRI scanner and imaged retrospective ECG-gated cine MR (16 phases/beat). RV ejection fraction (RVEF) and RV strain were analyzed before and after addition of 0.5∼3 nmol endothelin-1 (ET-1). After the measurements, we dissected trabeculae (length = 1.45 ± 0.07 mm, width = 334 ± 27 µm, thickness = 114 ± 6 µm) from the RVs of rat hearts. Trabeculae were electrically stimulated with 2-s intervals at extracellular Ca2+ of 0.7 and 2.0 mmol/L (24°C). Force and maximum dF/dt (dF/dtmax) were then measured using a silicon strain gauge in the absence and presence of 0.1 µM ET-1. Results MCT-rats showed higher systolic RV pressure (RVP), lower RVEF, and lower RV global longitudinal strain (RVGLS) in CMR imaging and showed lower developed force and lower dF/dtmax in their trabeculae. Correlation between RVGLS and dF/dtmax was higher (r = 0.53, p &lt; 0.05) than that between RVEF and dF/dtmax (r = 0.24). In 5 MCT-rats with preserved RVEF (&gt;50%), RVGLS had already been reduced, suggesting that RVGLS is reduced earlier than RVEF. ET-1 increased developed force and dF/dtmax in trabeculae from MCT-rats (12.2 ± 5.7 to 17.4 ± 3.1 mN/mm2 and 0.08 ± 0.03 to 0.14 ± 0.06 mN/mm2/sec, respectively, n = 6), and ET-1 also increased RVP in MCT-rats and Ctr-rats (49.0 ± 19.3 to 59.7 ± 16.8 mmHg in MCT-rats, n = 6, 17.3 ± 7.5 to 20.4 ± 7.8 mmHg in Ctr-rats, n = 2). According to RV global circumferential strain (RVGCS) and RVEF, we could divide MCT-rats into three groups as follows: MCT-rats with reduced-RVGCS (&gt; -20%)/preserved-RVEF (&gt; 50%), MCT-rats with increased-RVGCS (&lt; -30%)/preserved-RVEF and MCT-rats with reduced-RVGCS/reduced-RVEF. ET-1 reduced RVGCS in MCT-rats with reduced-RVGCS/preserved-RVEF, while ET-1 did not change RVGCS in MCT-rats with increased-RVGCS/preserved-RVEF. MCT-rats with reduced-RVGCS/reduced-RVEF died after injection of ET-1. In Ctr-rats, ET-1 did not change RVGCS and RVEF. These results suggest that RVGCS can be useful to assess RV functional reserve. Conclusion CMR feature tracking can estimate RV functional reserve earlier and more accurately than RVEF in rats with PAH. RV strain may become an important parameter to assess RV functional reserve in patients with PAH.

The use of cardiac MRI texture analysis to investigate left ventricle tissue characterisation in a healthy UK population

Funding Acknowledgements Type of funding sources: Other. Main funding source(s): Medical Research Scotland(MRS) Guerbet Group Background Cardiac MR texture analysis (TA) has the potential to distinguish subtle differences amongst myocardial diseases, but with limited evidence in a healthy population. The aim of this study was to assess the application of TA to cardiac CINE MR Images of the left ventricle to evaluate variability and consistency in a large-scale healthy population. Method A cohort of 600 healthy volunteers was recruited from the Tayside Screening for Prevention of Cardiac Events study. All subjects underwent short axis CINE CMR at 3T (including LVM), using a 2D ECG-gated breath-hold segmented steady-state gradient echo sequence with constant field-of-view and pixel size. For each subject, the mid short-axis slices of the left ventricle, at ED and ES, were extracted for image analysis. The TA parameters (n = 50) for all images were derived using Mazda v4.7 by a single observer. Three different regions-of-interest (ROI) were applied to the LV myocardium at ED and ES, as follows: 1) ‘whole wall’ (figure 1 a and d); 2) ‘septal wall’ (figure 1 b and e); and 3) ‘lateral wall’ (figure 1 c and f). Statistical comparisons were made for all texture features to establish how they varied between ED and ES, males versus females, different age ranges (40-45 years, 46-54 years, 55-63 years, and ≥ 64 years) and also between those with small (66.8 ± 5.88 g/m2), media (98.2 ± 15.28 g/m2) and large (146.2 ± 16.81 g/m2) LVM. Finally, the images of 30 volunteers were analysed by a second observer to derive test-retest inter-observer variation as an index of measurement repeatability for each TA feature. Results Of the original 50 TA features tested, the means of n = 45-49 features (number dependent on whether whole wall, septal wall or lateral wall) were significantly different when compared between ED and ES (p &lt; 0.05). For comparisons with gender, the means of n = 36-43 of the original features were significantly different (p &lt; 0.05). Additionally, the means of n = 15-29 features were significantly different when tested between sub-cohorts of different ages (p &lt; 0.05). When these data were combined together, ‘lateral wall’ was less sensitive, but for ‘whole wall’ and ‘septal wall’, the means of n = 7 of the original TA features were able to identify statistically significant differences between (i) ED and ES, (ii) males and females, (iii) different sub-cohorts of age, and (iv) different sub-cohorts of LVM (figure 2). Further, the root mean square test-retest inter-observer coefficient of variation associated with the repeatability of these measures was &lt;10%. The n = 7 most useful TA features were ‘GrMean’, ‘GrNonZeros’, ‘Average_RLNonUni’, ‘Average_LngREmph’, ‘Average_ShrtREmph’, ‘Average_Fraction’ and ‘S5_Average_Entropy’. Conclusion These findings showed the repeatability of CMR texture anaysis and capability to identify differences in a healthy population. Further work may identify whether these features are able to differentiate between different cardiovascular diseases.