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.

Characterizing advanced heart failure risk and hemodynamic phenotypes using interpretable machine learning.

BACKGROUND: Although previous risk models exist for advanced heart failure with reduced ejection fraction (HFrEF), few integrate invasive hemodynamics or support missing data. This study developed and validated a heart failure (HF) hemodynamic risk and phenotyping score for HFrEF, using Machine Learning (ML). METHODS: Prior to modeling, patients in training and validation HF cohorts were assigned to 1 of 5 risk categories based on the composite endpoint of death, left ventricular assist device (LVAD) implantation or transplantation (DeLvTx), and rehospitalization in 6 months of follow-up using unsupervised clustering. The goal of our novel interpretable ML modeling approach, which is robust to missing data, was to predict this risk category (1, 2, 3, 4, or 5) using either invasive hemodynamics alone or a rich and inclusive feature set that included noninvasive hemodynamics (all features). The models were trained using the ESCAPE trial and validated using 4 advanced HF patient cohorts collected from previous trials, then compared with traditional ML models. Prediction accuracy for each of these 5 categories was determined separately for each risk category to generate 5 areas under the curve (AUCs, or C-statistics) for belonging to risk category 1, 2, 3, 4, or 5, respectively. RESULTS: Across all outcomes, our models performed well for predicting the risk category for each patient. Accuracies of 5 separate models predicting a patient's risk category ranged from 0.896 +/- 0.074 to 0.969 +/- 0.081 for the invasive hemodynamics feature set and 0.858 +/- 0.067 to 0.997 +/- 0.070 for the all features feature set. CONCLUSION: Novel interpretable ML models predicted risk categories with a high degree of accuracy. This approach offers a new paradigm for risk stratification that differs from prediction of a binary outcome. Prospective clinical evaluation of this approach is indicated to determine utility for selecting the best treatment approach for patients based on risk and prognosis.

Echocardiographic derived pulmonary artery wedge pressure is associated with mortality, heart hospitalizations, and functional capacity in chronic systolic heart failure: insights from the HF-ACTION trial.

BACKGROUND: Heart Failure (HF) is associated with increased morbidity and mortality. Identification of patients at risk for adverse events could lead to improved outcomes. Few studies address the association of echocardiographic-derived PAWP with exercise capacity, readmissions, and mortality in HF. METHODS: HF-ACTION enrolled 2331 outpatients with HF with reduced ejection fraction (HFrEF) who were randomized to aerobic exercise training versus usual care. All patients underwent baseline echocardiography. Echocardiographic-derived PAWP (ePAWP) was assessed using the Nagueh formula. We evaluated the relationship between ePAWP to clinical outcomes. RESULTS: Among the 2331 patients in the HF-ACTION trial, 2125 patients consented and completed follow-up with available data. 807 of these patients had complete echocardiographic data that allowed the calculation of ePAWP. Of this cohort, mean age (SD) was 58 years (12.7), and 255 (31.6%) were female. The median ePAWP was 14.06 mmHg. ePAWP was significantly associated with cardiovascular death or HF hospitalization (Hazard ratio [HR] 1.02, coefficient 0.016, CI 1.002-1.030, p = 0.022) and all-cause death or HF hospitalization (HR 1.01, coefficient 0.010, CI 1.001-1.020, p = 0.04). Increased ePAWP was also associated with decreased exercise capacity leading to lower peak VO2 (p = < 0.001), high Ve/VCO2 slope (p = < 0.001), lower exercise duration (p = < 0.001), oxygen uptake efficiency (p = < 0.001), and shorter 6-MWT distance (p = < 0.001). CONCLUSIONS: Among HFrEF patients, echocardiographic-derived PAWP was associated with increased mortality, reduced functional capacity and heart failure hospitalization. ePAWP may be a viable noninvasive marker to risk stratify HFrEF patients.

Change in Systemic Arterial Pulsatility Index (SAPi) during Heart Failure Hospitalization is Associated with Improved Outcomes.

Study Objective: To identify Change in Systemic Arterial Pulsatitlity index (ΔSAPi) as a novel hemodynamic marker associated with outcomes in heart failure (HF). Design: The ESCAPE trial was a randomized controlled trial. Setting: The ESCAPE trial was conducted at 26 sites. Participants: 134 patients were analyzed (mean age 56.8 ± 13.4 years, 29% female). Interventions: We evaluated the change in SAPi, ([systemic pulse pressure/pulmonary artery wedge pressure) obtained at baseline and at the final hemodynamic measurement in the ESCAPE trial. Main Outcome Measures: Change in SAPi, (ΔSAPi), was analyzed for the primary outcomes of death, heart transplant, left ventricular assist device (DTxLVAD) or hospitalization, (DTxLVADHF) and secondary outcome of DTxLVAD using Cox proportional hazards regression. Results: Median change in SAPi was 0.81 (IQR 0.20-1.68). ΔSAPi in uppermost quartile was associated with reductions in DTxLVADHF (HR 0.55 [95% CI 0.32, 0.93]). ΔSAPi in the uppermost and lowermost quartiles combined was similarly associated with significant reductions in DTxLVADHF (HR 0.62 [95% CI 0.41, 0.94]). ΔSAPi higher than 1.17 was associated with improved DTxLVADHF. ΔSAPi was also associated with troponin levels at discharge (regression coefficient p = 0.001) and trended with 6-minute walk at discharge (Spearman correlation r = 0.179, p = 0.058). Conclusion: ΔSAPi was strongly associated with improved HF clinical profile and adverse outcomes. These findings support further exploration of Δ SAPi in the risk stratification of HF.

Cardiac Magnetic Resonance, Electromechanical Activation, Kidney Function, and Natriuretic Peptides in Cardiac Resynchronization Therapy Upgrades.

As the mechanism for worse prognosis after cardiac resynchronization therapy (CRT) upgrades in heart failure patients with RVP dependence (RVP-HF) has clinical implications for patient selection and CRT implementation approaches, this study's objective was to evaluate prognostic implications of cardiac magnetic resonance (CMR) findings and clinical factors in 102 HF patients (23.5% female, median age 66.5 years old, median follow-up 4.8 years) with and without RVP dependence undergoing upgrade and de novo CRT implants. Compared with other CRT groups, RVP-HF patients had decreased survival (p = 0.02), more anterior late-activated LV pacing sites (p = 0.002) by CMR, more atrial fibrillation (p = 0.0006), and higher creatinine (0.002). CMR activation timing at the LV pacing site predicted post-CRT LV functional improvement (p < 0.05), and mechanical activation onset < 34 ms by CMR at the LVP site was associated with decreased post-CRT survival in a model with higher pre-CRT creatinine and B-type natriuretic peptide (AUC 0.89; p < 0.0001); however, only the higher pre-CRT creatinine partially mediated (37%) the decreased survival in RVP-HF patients. In conclusion, RVP-HF had a distinct CMR phenotype, which has important implications for the selection of LV pacing sites in CRT upgrades, and only chronic kidney disease mediated the decreased survival after CRT in RVP-HF.

Impact of a Center of Excellence in Confirming or Excluding a Diagnosis of Hypertrophic Cardiomyopathy.

Tertiary hospitals with expertise in hypertrophic cardiomyopathy (HCM) are assuming a greater role in confirming and correcting HCM diagnoses at referring centers. The objectives were to establish the frequency of alternate diagnoses from referring centers and identify predictors of accuracy of an HCM diagnosis from the referring centers. Imaging findings from echocardiography, cardiac computed tomography, and cardiac magnetic resonance imaging (CMR) in 210 patients referred to an HCM Center of Excellence between September 2020 and October 2022 were reviewed. Clinical and imaging characteristics from pre-referral studies were used to construct a model for predictors of ruling out HCM or confirming the diagnosis using machine learning methods (least absolute shrinkage and selection operator logistic regression). Alternative diagnoses were found in 38 of the 210 patients (18.1%) (median age 60 years, 50% female). A total of 17 of the 38 patients (44.7%) underwent a new CMR after their initial visit, and 14 of 38 patients (36.8%) underwent review of a previous CMR. Increased left ventricular end-diastolic volume, indexed, greater septal thickness measurements, greater left atrial size, asymmetric hypertrophy on echocardiography, and the presence of an implantable cardioverter-defibrillator were associated with higher odds ratios for confirming a diagnosis of HCM, whereas increasing age and the presence of diabetes were more predictive of rejecting a diagnosis of HCM (area under the curve 0.902, p <0.0001). In conclusion, >1 in 6 patients with presumed HCM were found to have an alternate diagnosis after review at an HCM Center of Excellence, and both clinical findings and imaging parameters predicted an alternate diagnosis.

Noninvasive Electrical Mapping Compared with the Paced QRS Complex for Optimizing CRT Programmed Settings and Predicting Multidimensional Response.

The aim was to test the hypothesis that left ventricular (LV) and right ventricular (RV) activation from body surface electrical mapping (CardioInsight 252-electrode vest, Medtronic) identifies optimal cardiac resynchronization therapy (CRT) pacing strategies and outcomes in 30 patients. The LV80, RV80, and BIV80 were defined as the times to 80% LV, RV, or biventricular electrical activation. Smaller differences in the LV80 and RV80 (|LV80-RV80|) with synchronized LV pacing predicted better LV function post-CRT (p = 0.0004) than the LV-paced QRS duration (p = 0.32). Likewise, a lower RV80 was associated with a better pre-CRT RV ejection fraction by CMR (r = - 0.40, p = 0.04) and predicted post-CRT improvements in myocardial oxygen uptake (p = 0.01) better than the biventricular-paced QRS (p = 0.38), while a lower LV80 with BIV pacing predicted lower post-CRT B-type natriuretic peptide (BNP) (p = 0.02). RV pacing improved LV function with smaller |LV80-RV80| (p = 0.009). In conclusion, 3-D electrical mapping predicted favorable post-CRT outcomes and informed effective pacing strategies.

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.

First-in-human noninvasive left ventricular ultrasound pacing: A potential screening tool for cardiac resynchronization therapy.

Background: A screening tool to predict response to cardiac resynchronization therapy (CRT) could improve patient selection and outcomes. Objective: The purpose of this study was to investigate the feasibility and safety of noninvasive CRT via transcutaneous ultrasonic left ventricular (LV) pacing applied as a screening test before CRT implants. Methods: P-wave-triggered ultrasound stimuli were delivered during bolus dosing of an echocardiographic contrast agent to simulate CRT noninvasively. Ultrasound pacing was delivered at a variety of LV locations with a range of atrioventricular delays to achieve fusion with intrinsic ventricular activation. Three-dimensional cardiac activation maps were acquired via the Medtronic CardioInsight 252-electrode mapping vest during baseline, ultrasound pacing, and after CRT implantation. A separate control group received only the CRT implants. Results: Ultrasound pacing was achieved in 10 patients with a mean of 81.2 ± 50.8 ultrasound paced beats per patient and up to 20 consecutive beats of ultrasound pacing. QRS width at baseline (168.2 ± 17.8 ms) decreased significantly to 117.3 ± 21.5 ms (P <.001) in the best ultrasound paced beat and to 125.8 ± 13.3 ms (P <.001) in the best CRT beat. Electrical activation patterns were similar between CRT pacing and ultrasound pacing with stimulation from the same area of the LV. Troponin results were similar between the ultrasound pacing and the control groups (P = .96), confirming safety. Conclusion: Noninvasive ultrasound pacing before CRT is safe and feasible, and it estimates the degree of electrical resynchronization achievable with CRT. Further study of this promising technique to guide CRT patient selection is warranted.

Use of Acellular Biologic Matrix Envelope for Cardiac Implantable Electronic Device Placement to Correct Migration into Submuscular Breast Implant Pocket.

Breast implants whether used for cosmetic or reconstructive purposes can be placed in pockets either above or below the pectoralis major muscle, depending on clinical circumstances such as subcutaneous tissue volume, history of radiation, and patient preference. Likewise, cardiac implantable electronic devices (CIEDs) can be placed above or below the pectoralis major muscle. When a patient has both devices, knowledge of the pocket location is important for procedural planning and for durability of device placement and performance. Here, we report a patient who previously failed subcutaneous CIED placement due to incision manipulation with prior threatened device exposure requiring plane change to subpectoral pocket. Her course was complicated by submuscular migration of the CIED into her breast implant periprosthetic pocket. With subcutaneous plane change being inadvisable due to patient noncompliance, soft tissue support of subpectoral CIED placement with an acellular biologic matrix (ABM) was performed. Similar to soft tissue support used for breast implants, submuscular CIED neo-pocket creation with ABM was performed with durable CIED device positioning confirmed at 9 months postprocedure.

Compensation for respiratory motion-induced signal loss and phase corruption in free-breathing self-navigated cine DENSE using deep learning.

PURPOSE: To introduce a model that describes the effects of rigid translation due to respiratory motion in displacement encoding with stimulated echoes (DENSE) and to use the model to develop a deep convolutional neural network to aid in first-order respiratory motion compensation for self-navigated free-breathing cine DENSE of the heart. METHODS: The motion model includes conventional position shifts of magnetization and further describes the phase shift of the stimulated echo due to breathing. These image-domain effects correspond to linear and constant phase errors, respectively, in k-space. The model was validated using phantom experiments and Bloch-equation simulations and was used along with the simulation of respiratory motion to generate synthetic images with phase-shift artifacts to train a U-Net, DENSE-RESP-NET, to perform motion correction. DENSE-RESP-NET-corrected self-navigated free-breathing DENSE was evaluated in human subjects through comparisons with signal averaging, uncorrected self-navigated free-breathing DENSE, and breath-hold DENSE. RESULTS: Phantom experiments and Bloch-equation simulations showed that breathing-induced constant phase errors in segmented DENSE leads to signal loss in magnitude images and phase corruption in phase images of the stimulated echo, and that these artifacts can be corrected using the known respiratory motion and the model. For self-navigated free-breathing DENSE where the respiratory motion is not known, DENSE-RESP-NET corrected the signal loss and phase-corruption artifacts and provided reliable strain measurements for systolic and diastolic parameters. CONCLUSION: DENSE-RESP-NET is an effective method to correct for breathing-associated constant phase errors. DENSE-RESP-NET used in concert with self-navigation methods provides reliable free-breathing DENSE myocardial strain measurement.

Positional obstructive sleep apnea in patients with atrial fibrillation.

PURPOSE: Obstructive sleep apnea (OSA) is a common, potentially modifiable condition implicated in the pathogenesis of atrial fibrillation (AF). The presence and severity of OSA is largely sleep position-dependent, yet there is high variability in positional dependence among patients with OSA. We investigated the prevalence of positional OSA (POSA) and examined associated factors in patients with AF. METHODS: We recruited an equal number of patients with and without AF who underwent diagnostic polysomnography. Patients included had ≥ 120 min of total sleep time with 30 min of sleep in both supine and lateral positions. POSA was defined as an overall apnea hypopnea index (AHI) ≥ 5/h, supine AHI (sAHI) ≥ 5/h, and sAHI greater than twice the non-supine AHI. POSA prevalence was compared in patients with and without AF adjusting for age, sex, OSA severity, and heart failure. RESULTS: A total of patients (male: 56%, mean age 62 years) were included. POSA prevalence was similar between the two groups (46% vs. 39%; p = 0.33). Obesity and severe OSA (AHI ≥ 30/h) were associated with low likelihood of POSA (OR [CI] of 0.17 [0.09-0.32] and 0.28 [0.12-0.62]). In patients with AF, male sex was associated with a higher likelihood of POSA (OR [CI] of 3.16 [1.06-10.4]). CONCLUSION: POSA is common, affecting more than half of patients with AF, but the prevalence was similar in those without AF. Obesity and more severe OSA are associated with lower odds of POSA. Positional therapy should be considered in patients with mild OSA and POSA.

Academic cardiac electrophysiologists' perspectives on sleep apnea care.

PURPOSE: Obstructive sleep apnea syndrome (OSAS) is an important, modifiable risk factor in the pathophysiology of arrhythmias including atrial fibrillation (AF). The purpose of the study was to evaluate cardiac electrophysiologists' (EPs) perception of OSAS. METHODS: We designed a 27-item online Likert scale-based survey instrument entailing several domains: (1) relevance of OSAS in EP practice, (2) OSAS screening and diagnosis, (3) perception on treatments for OSAS, (4) opinion on the OSAS care model. The survey was distributed to 89 academic EP programs in the USA and Canada. While the survey instrument questions refer to the term sleep apnea (SA), our discussion of the diagnosis, management, and research on the sleep disorder is more accurately described with the term OSAS. RESULTS: A total of 105 cardiac electrophysiologists from 49 institutions responded over a 9-month period. The majority of respondents agreed that sleep apnea (SA) is a major concern in their practice (94%). However, 42% reported insufficient education on SA during training. Many (58%) agreed that they would be comfortable managing SA themselves with proper training and education and 66% agreed cardiac electrophysiologists should become more involved in management. Half of EPs (53%) were not satisfied with the sleep specialist referral process. Additionally, a majority (86%) agreed that trained advanced practice providers should be able to assess and manage SA. Time constraints, lack of knowledge, and the referral process are identified as major barriers to EPs becoming more involved in SA care. CONCLUSIONS: We found that OSAS is widely recognized as a major concern for EP. However, incorporation of OSAS care in training and routine practice lags. Barriers to increased involvement include time constraints and education. This study can serve as an impetus for innovation in the cardiology OSAS care model.

Systemic arterial pulsatility (SAPi) in advanced heart failure: a novel hemodynamic risk stratification tool.

Systemic arterial pulsatility index (SAPi) is a novel hemodynamic marker for ventriculo-arterial coupling (VAC), as it integrates the contractile properties of the left ventricle with the aortic impendence. SAPi can identify heart failure patients at increased risk for adverse events. Systemic pulsatility decreases as heart failure progresses, and there is a decrease in pulse pressure accompanied by an increase in left ventricular filling pressure. Decreasing SAPi is associated with worse prognosis in advanced heart failure patients.

Coronary perfusion pressure is associated with adverse outcomes in advanced heart failure.

BACKGROUND: Myocardial perfusion is an important determinant of cardiac function. We hypothesized that low coronary perfusion pressure (CPP) would be associated with adverse outcomes in heart failure. Myocardial perfusion impacts the contractile efficiency thus a low CPP would signal low myocardial perfusion in the face of increased cardiac demand as a result of volume overload. METHODS: We analyzed patients with complete hemodynamic data in the Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness trial using Cox Proportional Hazards regression for the primary outcome of the composite risk of death, heart transplantation, or left ventricular assist device [(LVAD). DT × LVAD] and the secondary outcome of the composite risk of DT × LVAD and heart failure hospitalization (DT × LVADHF). CPP was calculated as the difference between diastolic blood pressure and pulmonary artery wedge pressure. Heart failure categories (ischemic vs non-ischemic) were also stratified based on CPP strata. RESULTS: The 158 patients (56.7 ± 13.6 years, 28.5% female) studied had a median CPP of 40 mmHg (IQR 35-52 mmHg). During 6 months of follow-up, 35 (22.2%) had the composite primary outcome and 109 (69.0%) had the composite secondary outcome. When these outcomes were then stratified based on the median, CPP was associated with these outcomes. Increasing CPP was associated with lower risk of both the primary outcome of DT × LVAD (HR 0.96, 95% CI 0.94-0.99 p = .002) and as well as the secondary outcome of DT × LVADHF (p = .0008) There was significant interaction between CPP and ischemic etiology (p = .04). CONCLUSION: A low coronary artery perfusion pressure below (median) 40mmHg in patients with advanced heart failure undergoing invasive hemodynamic monitoring with a pulmonary artery catheter was associated with adverse outcomes. CPP could useful in guiding risk stratification of advanced heart failure patients and timely evaluation of advanced heart failure therapies.

Right atrial structural remodeling predict worse outcomes in transcatheter mitral valve repair.

BACKGROUND: In the current study, we assess the predictive role of right and left atrial volume indices (RAVI and LAVI) as well as the ratio of RAVI/LAVI (RLR) on mortality following transcatheter mitral valve repair (TMVr). METHODS: Transthoracic echocardiograms of 158 patients who underwent TMVr at a single academic medical center from 2011 to 2018 were reviewed retrospectively. RAVI and LAVI were calculated using Simpson's method. Patients were stratified based on etiology of mitral regurgitation (MR). Cox proportional-hazard regression was created utilizing MR type, STS-score, and RLR to assess the independent association of RLR with survival. Kaplan-Meier analysis was used to analyze the association between RAVI and LAVI with all-cause mortality. Hemodynamic values from preprocedural right heart catheterization were also compared between RLR groups. RESULTS: Among 123 patients included (median age 81.3 years; 52.5% female) there were 50 deaths during median follow-up of 3.0 years. Patients with a high RAVI and low LAVI had significantly higher all-cause mortality while patients with high LAVI and low RAVI had significantly improved all-cause mortality compared to other groups (p = 0.0032). RLR was significantly associated with mortality in patients with both functional and degenerative MR (p = 0.0038). Finally, Cox proportion-hazard modeling demonstrated that an elevated RLR above the median value was an independent predictor of all-cause mortality [HR = 2.304; 95% CI = 1.26-4.21, p = 0.006] when MR type and STS score were accounted for. CONCLUSION: Patients with a high RAVI and low LAVI had significantly increased mortality than other groups following TMVr suggesting RA remodeling may predict worse outcomes following the procedure. Concordantly, RLR was predictive of mortality independent of MR type and preprocedural STS-score. These indices may provide additional risk stratification in patients undergoing evaluation for TMVr.

Diffusion Models To Predict 3D Late Mechanical Activation From Sparse 2D Cardiac MRIs.

Identifying regions of late mechanical activation (LMA) of the left ventricular (LV) myocardium is critical in determining the optimal pacing site for cardiac resynchronization therapy in patients with heart failure. Several deep learning-based approaches have been developed to predict 3D LMA maps of LV myocardium from a stack of sparse 2D cardiac magnetic resonance imaging (MRIs). However, these models often loosely consider the geometric shape structure of the myocardium. This makes the reconstructed activation maps suboptimal; hence leading to a reduced accuracy of predicting the late activating regions of hearts. In this paper, we propose to use shape-constrained diffusion models to better reconstruct a 3D LMA map, given a limited number of 2D cardiac MRI slices. In contrast to previous methods that primarily rely on spatial correlations of image intensities for 3D reconstruction, our model leverages object shape as priors learned from the training data to guide the reconstruction process. To achieve this, we develop a joint learning network that simultaneously learns a mean shape under deformation models. Each reconstructed image is then considered as a deformed variant of the mean shape. To validate the performance of our model, we train and test the proposed framework on a publicly available mesh dataset of 3D myocardium and compare it with state-of-the-art deep learning-based reconstruction models. Experimental results show that our model achieves superior performance in reconstructing the 3D LMA maps as compared to the state-of-the-art models.

JOINT DEEP LEARNING FOR IMPROVED MYOCARDIAL SCAR DETECTION FROM CARDIAC MRI.

Automated identification of myocardial scar from late gadolinium enhancement cardiac magnetic resonance images (LGE-CMR) is limited by image noise and artifacts such as those related to motion and partial volume effect. This paper presents a novel joint deep learning (JDL) framework that improves such tasks by utilizing simultaneously learned myocardium segmentations to eliminate negative effects from non-region-of-interest areas. In contrast to previous approaches treating scar detection and myocardium segmentation as separate or parallel tasks, our proposed method introduces a message passing module where the information of myocardium segmentation is directly passed to guide scar detectors. This newly designed network will efficiently exploit joint information from the two related tasks and use all available sources of myocardium segmentation to benefit scar identification. We demonstrate the effectiveness of JDL on LGE-CMR images for automated left ventricular (LV) scar detection, with great potential to improve risk prediction in patients with both ischemic and non-ischemic heart disease and to improve response rates to cardiac resynchronization therapy (CRT) for heart failure patients. Experimental results show that our proposed approach outperforms multiple state-of-the-art methods, including commonly used two-step segmentation-classification networks, and multitask learning schemes where subtasks are indirectly interacted.

Machine learning for multidimensional response and survival after cardiac resynchronization therapy using features from cardiac magnetic resonance.

Background: Cardiac resynchronization therapy (CRT) response is complex, and better approaches are required to predict survival and need for advanced therapies. Objective: The objective was to use machine learning to characterize multidimensional CRT response and its relationship with long-term survival. Methods: Associations of 39 baseline features (including cardiac magnetic resonance [CMR] findings and clinical parameters such as glomerular filtration rate [GFR]) with a multidimensional CRT response vector (consisting of post-CRT left ventricular end-systolic volume index [LVESVI] fractional change, post-CRT B-type natriuretic peptide, and change in peak VO2) were evaluated. Machine learning generated response clusters, and cross-validation assessed associations of clusters with 4-year survival. Results: Among 200 patients (median age 67.4 years, 27.0% women) with CRT and CMR, associations with more than 1 response parameter were noted for the CMR CURE-SVD dyssynchrony parameter (associated with post-CRT brain natriuretic peptide [BNP] and LVESVI fractional change) and GFR (associated with peak VO2 and post-CRT BNP). Machine learning defined 3 response clusters: cluster 1 (n = 123, 90.2% survival [best]), cluster 2 (n = 45, 60.0% survival [intermediate]), and cluster 3 (n = 32, 34.4% survival [worst]). Adding the 6-month response cluster to baseline features improved the area under the receiver operating characteristic curve for 4-year survival from 0.78 to 0.86 (P = .02). A web-based application was developed for cluster determination in future patients. Conclusion: Machine learning characterizes distinct CRT response clusters influenced by CMR features, kidney function, and other factors. These clusters have a strong and additive influence on long-term survival relative to baseline features.