Machine learning of ECG waveforms and cardiac magnetic resonance for response and survival after cardiac resynchronization therapy.

Cardiac resynchronization therapy (CRT) can lead to marked symptom reduction and improved survival in selected patients with heart failure with reduced ejection fraction (HFrEF); however, many candidates for CRT based on clinical guidelines do not have a favorable response. A better way to identify patients expected to benefit from CRT that applies machine learning to accessible and cost-effective diagnostic tools such as the 12-lead electrocardiogram (ECG) could have a major impact on clinical care in HFrEF by helping providers personalize treatment strategies and avoid delays in initiation of other potentially beneficial treatments. This study addresses this need by demonstrating that a novel approach to ECG waveform analysis using functional principal component decomposition (FPCD) performs better than measures that require manual ECG analysis with the human eye and also at least as well as a previously validated but more expensive approach based on cardiac magnetic resonance (CMR). Analyses are based on five-fold cross validation of areas under the curve (AUCs) for CRT response and survival time after the CRT implant using Cox proportional hazards regression with stratification of groups using a Gaussian mixture model approach. Furthermore, FPCD and CMR predictors are shown to be independent, which demonstrates that the FPCD electrical findings and the CMR mechanical findings together provide a synergistic model for response and survival after CRT. In summary, this study provides a highly effective approach to prognostication after CRT in HFrEF using an accessible and inexpensive diagnostic test with a major expected impact on personalization of therapies.

StrainNet: Improved Myocardial Strain Analysis of Cine MRI by Deep Learning from DENSE.

Purpose: To develop a three-dimensional (two dimensions + time) convolutional neural network trained with displacement encoding with stimulated echoes (DENSE) data for displacement and strain analysis of cine MRI. Materials and Methods: In this retrospective multicenter study, a deep learning model (StrainNet) was developed to predict intramyocardial displacement from contour motion. Patients with various heart diseases and healthy controls underwent cardiac MRI examinations with DENSE between August 2008 and January 2022. Network training inputs were a time series of myocardial contours from DENSE magnitude images, and ground truth data were DENSE displacement measurements. Model performance was evaluated using pixelwise end-point error (EPE). For testing, StrainNet was applied to contour motion from cine MRI. Global and segmental circumferential strain (Ecc) derived from commercial feature tracking (FT), StrainNet, and DENSE (reference) were compared using intraclass correlation coefficients (ICCs), Pearson correlations, Bland-Altman analyses, paired t tests, and linear mixed-effects models. Results: The study included 161 patients (110 men; mean age, 61 years ± 14 [SD]), 99 healthy adults (44 men; mean age, 35 years ± 15), and 45 healthy children and adolescents (21 males; mean age, 12 years ± 3). StrainNet showed good agreement with DENSE for intramyocardial displacement, with an average EPE of 0.75 mm ± 0.35. The ICCs between StrainNet and DENSE and FT and DENSE were 0.87 and 0.72, respectively, for global Ecc and 0.75 and 0.48, respectively, for segmental Ecc. Bland-Altman analysis showed that StrainNet had better agreement than FT with DENSE for global and segmental Ecc. Conclusion: StrainNet outperformed FT for global and segmental Ecc analysis of cine MRI.Keywords: Image Postprocessing, MR Imaging, Cardiac, Heart, Pediatrics, Technical Aspects, Technology Assessment, Strain, Deep Learning, DENSE Supplemental material is available for this article. © RSNA, 2023.

Improved computation of Lagrangian tissue displacement and strain for cine DENSE MRI using a regularized spatiotemporal least squares method.

Introduction: In displacement encoding with stimulated echoes (DENSE), tissue displacement is encoded in the signal phase such that the phase of each pixel in space and time provides an independent measurement of absolute tissue displacement. Previously for DENSE, estimation of Lagrangian displacement used two steps: first a spatial interpolation and, second, least squares fitting through time to a Fourier or polynomial model. However, there is no strong rationale for such a through-time model. Methods: To compute the Lagrangian displacement field from DENSE phase data, a minimization problem is introduced to enforce fidelity with the acquired Eulerian displacement data while simultaneously providing model-independent regularization in space and time, enforcing only spatiotemporal smoothness. A regularized spatiotemporal least squares (RSTLS) method is used to solve the minimization problem, and RSTLS was tested using two-dimensional DENSE data from 71 healthy volunteers. Results: The mean absolute percent error (MAPE) between the Lagrangian displacements and the corresponding Eulerian displacements was significantly lower for the RSTLS method vs. the two-step method for both x- and y-directions (0.73±0.59 vs 0.83 ±0.1, p < 0.05) and (0.75±0.66 vs 0.82 ±0.1, p < 0.05), respectively. Also, peak early diastolic strain rate (PEDSR) was higher (1.81±0.58 (s-1) vs. 1.56±0. 63 (s-1), p<0.05) and the strain rate during diastasis was lower (0.14±0.18 (s-1) vs 0.35±0.2 (s-1), p < 0.05) for the RSTLS vs. the two-step method, with the former suggesting that the two-step method was over-regularized. Discussion: The proposed RSTLS method provides more realistic measurements of Lagrangian displacement and strain from DENSE images without imposing arbitrary motion models.

Reproducibility of global and segmental myocardial strain using cine DENSE at 3 T: a multicenter cardiovascular magnetic resonance study in healthy subjects and patients with heart disease.

BACKGROUND: While multiple cardiovascular magnetic resonance (CMR) methods provide excellent reproducibility of global circumferential and global longitudinal strain, achieving highly reproducible segmental strain is more challenging. Previous single-center studies have demonstrated excellent reproducibility of displacement encoding with stimulated echoes (DENSE) segmental circumferential strain. The present study evaluated the reproducibility of DENSE for measurement of whole-slice or global circumferential (Ecc), longitudinal (Ell) and radial (Err) strain, torsion, and segmental Ecc at multiple centers. METHODS: Six centers participated and a total of 81 subjects were studied, including 60 healthy subjects and 21 patients with various types of heart disease. CMR utilized 3 T scanners, and cine DENSE images were acquired in three short-axis planes and in the four-chamber long-axis view. During one imaging session, each subject underwent two separate DENSE scans to assess inter-scan reproducibility. Each subject was taken out of the scanner and repositioned between the scans. Intra-user, inter-user-same-site, inter-user-different-site, and inter-user-Human-Deep-Learning (DL) comparisons assessed the reproducibility of different users analyzing the same data. Inter-scan comparisons assessed the reproducibility of DENSE from scan to scan. The reproducibility of whole-slice or global Ecc, Ell and Err, torsion, and segmental Ecc were quantified using Bland-Altman analysis, the coefficient of variation (CV), and the intraclass correlation coefficient (ICC). CV was considered excellent for CV ≤ 10%, good for 10% < CV ≤ 20%, fair for 20% < CV ≤ 40%, and poor for CV > 40. ICC values were considered excellent for ICC > 0.74, good for ICC 0.6 < ICC ≤ 0.74, fair for ICC 0.4 < ICC ≤ 0.59, poor for ICC < 0.4. RESULTS: Based on CV and ICC, segmental Ecc provided excellent intra-user, inter-user-same-site, inter-user-different-site, inter-user-Human-DL reproducibility and good-excellent inter-scan reproducibility. Whole-slice Ecc and global Ell provided excellent intra-user, inter-user-same-site, inter-user-different-site, inter-user-Human-DL and inter-scan reproducibility. The reproducibility of torsion was good-excellent for all comparisons. For whole-slice Err, CV was in the fair-good range, and ICC was in the good-excellent range. CONCLUSIONS: Multicenter data show that 3 T CMR DENSE provides highly reproducible whole-slice and segmental Ecc, global Ell, and torsion measurements in healthy subjects and heart disease patients.

Fully-automated global and segmental strain analysis of DENSE cardiovascular magnetic resonance using deep learning for segmentation and phase unwrapping.

BACKGROUND: Cardiovascular magnetic resonance (CMR) cine displacement encoding with stimulated echoes (DENSE) measures heart motion by encoding myocardial displacement into the signal phase, facilitating high accuracy and reproducibility of global and segmental myocardial strain and providing benefits in clinical performance. While conventional methods for strain analysis of DENSE images are faster than those for myocardial tagging, they still require manual user assistance. The present study developed and evaluated deep learning methods for fully-automatic DENSE strain analysis. METHODS: Convolutional neural networks (CNNs) were developed and trained to (a) identify the left-ventricular (LV) epicardial and endocardial borders, (b) identify the anterior right-ventricular (RV)-LV insertion point, and (c) perform phase unwrapping. Subsequent conventional automatic steps were employed to compute strain. The networks were trained using 12,415 short-axis DENSE images from 45 healthy subjects and 19 heart disease patients and were tested using 10,510 images from 25 healthy subjects and 19 patients. Each individual CNN was evaluated, and the end-to-end fully-automatic deep learning pipeline was compared to conventional user-assisted DENSE analysis using linear correlation and Bland Altman analysis of circumferential strain. RESULTS: LV myocardial segmentation U-Nets achieved a DICE similarity coefficient of 0.87 ± 0.04, a Hausdorff distance of 2.7 ± 1.0 pixels, and a mean surface distance of 0.41 ± 0.29 pixels in comparison with manual LV myocardial segmentation by an expert. The anterior RV-LV insertion point was detected within 1.38 ± 0.9 pixels compared to manually annotated data. The phase-unwrapping U-Net had similar or lower mean squared error vs. ground-truth data compared to the conventional path-following method for images with typical signal-to-noise ratio (SNR) or low SNR (p < 0.05), respectively. Bland-Altman analyses showed biases of 0.00 ± 0.03 and limits of agreement of - 0.04 to 0.05 or better for deep learning-based fully-automatic global and segmental end-systolic circumferential strain vs. conventional user-assisted methods. CONCLUSIONS: Deep learning enables fully-automatic global and segmental circumferential strain analysis of DENSE CMR providing excellent agreement with conventional user-assisted methods. Deep learning-based automatic strain analysis may facilitate greater clinical use of DENSE for the quantification of global and segmental strain in patients with cardiac disease.

A Neonatal Bimodal MR-CT Head Template.

Neonatal MR templates are appropriate for brain structural analysis and spatial normalization. However, they do not provide the essential accurate details of cranial bones and fontanels-sutures. Distinctly, CT images provide the best contrast for bone definition and fontanels-sutures. In this paper, we present, for the first time, an approach to create a fully registered bimodal MR-CT head template for neonates with a gestational age of 39 to 42 weeks. Such a template is essential for structural and functional brain studies, which require precise geometry of the head including cranial bones and fontanels-sutures. Due to the special characteristics of the problem (which requires inter-subject inter-modality registration), a two-step intensity-based registration method is proposed to globally and locally align CT images with an available MR template. By applying groupwise registration, the new neonatal CT template is then created in full alignment with the MR template to build a bimodal MR-CT template. The mutual information value between the CT and the MR template is 1.17 which shows their perfect correspondence in the bimodal template. Moreover, the average mutual information value between normalized images and the CT template proposed in this study is 1.24±0.07. Comparing this value with the one reported in a previously published approach (0.63±0.07) demonstrates the better generalization properties of the new created template and the superiority of the proposed method for the creation of CT template in the standard space provided by MR neonatal head template. The neonatal bimodal MR-CT head template is freely downloadable from https://www.u-picardie.fr/labo/GRAMFC.

Skull Segmentation and Reconstruction From Newborn CT Images Using Coupled Level Sets.

This study presents a new approach for segmentation and reconstruction of newborn's skull including bones, fontanels, and sutures from computed tomography (CT) images. The segmentation approach relies on propagation of a pair of interacting smooth surfaces based on geodesic active regions. These surfaces evolve in opposite directions; the exterior surface moves inward while the interior one moves in outward direction. The moving surfaces are forced to stop when arriving at the outer or the inner surface of the cranial bones using edge information. Since fontanels and sutures are not directly detectable in CT images, this method imposes specific propagation constraints for coupled interfaces to prevent the moving surfaces from intersecting each other and penetrating into the opposite region. Finally, an algorithm for level set initialization is introduced which enforces the evolving surfaces to conform to the shape of the head. The proposed method was evaluated using 18 neonatal CT images. The segmentation results achieved by the suggested method have been compared with manual segmentations by two different raters, performed to establish a reliable reference. The comparison of the two segmentation results using the Dice similarity coefficient and modified Hausdorff distance shows that the proposed approach provides satisfactory results.