Utility of Coronary Computed Tomography Angiography in Patients Undergoing Transcatheter Aortic Valve Implantation: A Meta-Analysis and Meta-Regression Based on Published Data from 7458 Patients.

BACKGROUND: Coronary CT angiography (CCTA) may detect coronary artery disease (CAD) in transcatheter aortic valve implantation (TAVI) patients and may obviate invasive coronary angiography (ICA) in selected patients. We assessed the diagnostic accuracy of CCTA for detecting CAD in TAVI patients based on published data. METHODS: Meta-analysis and meta-regression were performed based on a comprehensive electronic search, including relevant studies assessing the diagnostic accuracy of CCTA in the setting of TAVI patients compared to ICA. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), were calculated on a patient and per segment level. RESULTS: Overall, 27 studies (total of 7458 patients) were included. On the patient level, the CCTA's pooled sensitivity and NPV were 95% (95% CI: 93-97%) and 97% (95% CI: 95-98%), respectively, while the specificity and PPV were at 73% (95% CI: 62-82%) and 64% (95% CI: 57-71%), respectively. On the segmental coronary vessel level, the sensitivity and NPV were 90% (95% CI: 79-96%) and 98% (95% CI: 97-99%). CONCLUSIONS: This meta-analysis highlights CCTA's potential as a first-line diagnostic tool although its limited PPV and specificity may pose challenges when interpreting heavily calcified arteries. This study underscores the need for further research and protocol standardization in this area.

Immunodeficiency is prevalent in congenital heart disease and associated with increased risk of emergency admissions and death.

AIMS: To provide population-based data on the prevalence and clinical significance of immune deficiency syndromes (IDS) associated with congenital heart disease (CHD). METHODS AND RESULTS: Utilizing administrative German Health System data the prevalence of increased susceptibility to infection (ISI) or confirmed IDS was assessed in CHD patients and compared with an age-matched non-congenital control group. Furthermore, the prognostic significance of IDS was assessed using all-cause mortality and freedom from emergency hospital admission. A total of 54 449 CHD patients were included. Of these 14 998 (27.5%) had ISI and 3034 (5.6%) had a documented IDS (compared with 2.9% of the age-matched general population). During an observation period of 394 289 patient-years, 3824 CHD patients died, and 31 017 patients experienced a combined event of all-cause mortality or emergency admission. On multivariable Cox proportional-hazard analysis, the presence of ISI [hazard ratio (HR): 2.14, P < 0.001] or documented IDS (HR: 1.77, P = 0.035) emerged as independent predictors of all-cause mortality. In addition, ISI and confirmed IDS were associated with a significantly higher risk of emergency hospital admission (P = 0.01 for both on competing risk analysis) during follow-up. CONCLUSION: Limited immune competence is common in CHD patients and associated with an increased risk of morbidity and mortality. This highlights the need for structured IDS screening and collaboration with immunology specialists as immunodeficiency may be amenable to specific therapy. Furthermore, studies are required to assess whether IDS patients might benefit from intensified antibiotic shielding or tailored prophylaxis.

A framework of deep learning networks provides expert-level accuracy for the detection and prognostication of pulmonary arterial hypertension.

AIMS: To test the hypothesis that deep learning (DL) networks reliably detect pulmonary arterial hypertension (PAH) and provide prognostic information. METHODS AND RESULTS: Consecutive patients with PAH, right ventricular (RV) dilation (without PAH), and normal controls were included. An ensemble of deep convolutional networks incorporating echocardiographic views and estimated RV systolic pressure (RVSP) was trained to detect (invasively confirmed) PAH. In addition, DL-networks were trained to segment cardiac chambers and extracted geometric information throughout the cardiac cycle. The ability of DL parameters to predict all-cause mortality was assessed using Cox-proportional hazard analyses. Overall, 450 PAH patients, 308 patients with RV dilatation (201 with tetralogy of Fallot and 107 with atrial septal defects) and 67 normal controls were included. The DL algorithm achieved an accuracy and sensitivity of detecting PAH on a per patient basis of 97.6 and 100%, respectively. On univariable analysis, automatically determined right atrial area, RV area, RV fractional area change, RV inflow diameter and left ventricular eccentricity index (P < 0.001 for all) were significantly related to mortality. On multivariable analysis DL-based RV fractional area change (P < 0.001) and right atrial area (P = 0.003) emerged as independent predictors of outcome. Statistically, DL parameters were non-inferior to measures obtained manually by expert echocardiographers in predicting prognosis. CONCLUSION: The study highlights the utility of DL algorithms in detecting PAH on routine echocardiograms irrespective of RV dilatation. The algorithms outperform conventional echocardiographic evaluation and provide prognostic information at expert-level. Therefore, DL methods may allow for improved screening and optimized management of PAH.

Incidence and predictors of left atrial appendage thrombus on transesophageal echocardiography before elective cardioversion.

Guidelines recommend transesophageal echocardiography (TEE) before cardioversion in thrombogenic arrhythmias when the requirement of ≥ 3 weeks of anticoagulation is not met. Current data to support this approach, especially with direct oral anticoagulants (DOAC), are scarce. We analyzed consecutive elective pre-cardioversion TEE in a high-volume electrophysiology center for the occurrence of left atrial appendage (LAA) thrombi or reduced LAA flow velocity. Possible predictors were recorded and compared in a multivariate logistic regression analysis. Consecutive pre-cardioversion TEE in 512 patients (148 female, median age 69 years) were included. In all patients, indication for TEE was either intake of anticoagulation < 3 weeks before cardioversion or uncertain adherence to the prescribed anticoagulation regimen. Of the 512 TEE, 19 (3.7%) depicted a LAA thrombus. An additional 41 patients (8.0%) showed either a reduced LAA flow velocity (≤ 20 cm/s), LAA sludge, or both. In a multivariate logistic regression analysis, QRS width on admission 12-lead ECG emerged as a possible predictor of LAA thrombus and reduced LAA flow (p = 0.008). Noteworthy, a high CHA2DS2-VASc score was not associated with an increased risk of reduced LAA emptying velocity and LAA thrombi were even found in patients with a CHA2DS2-VASc score of 0 (n = 1) and 1 (n = 1). The presence of LAA thrombus before an elective cardioversion is a rare event in the age of direct oral anticoagulants. However, LAA thrombi occurred even in supposed low-risk individuals according to the CHA2DS2-VASc score. QRS width may aid in identifying patients at risk of reduced LAA flow velocity.

Incidence and Predictors of Left Atrial Appendage Thrombus before Catheter Ablation of Thrombogenic Arrhythmias.

Introduction: Transesophageal echocardiography (TEE) is routinely performed before catheter ablation of atrial tachyarrhythmias to rule out the presence of left atrial thrombi. However, data to support this practice are inconsistent. Methods: We analyzed consecutive pre-procedural TEE in a high-volume electrophysiology center for the presence of left atrial thrombi and a relevant flow reduction in the left atrial appendage (LAA) defined as LAA sludge or LAA emptying velocity (LAAEV) < 20 cm/s. The possible predictors of reduced flow were recorded and compared in a multivariate logistic regression analysis. Results: 1676 TEE were included (1122 before pulmonary vein isolation, 436 before atrial flutter ablation, 166 before other ablations). 543 patients (32%) were female and 991 (59%) were on DOAC. Nine patients (0.5%) had an LAA thrombus on pre-procedural TEE. Ninety-five further patients (5.7%) had a relevant reduction in LAA flow. The underlying rhythm showed a significant association with the presence of LAA thrombus or reduced LAA flow (p = 0.003). Patients in sinus rhythm and cavotricuspid isthmus-dependent atrial flutter exhibited the lowest risk. Additionally, reduced kidney function was associated with a reduction in LAA flow velocities (p = 0.04). Of note, two LAA thrombi occurred in patients in sinus rhythm and six out of nine patients with an LAA thrombus were on vitamin-K antagonists. Conclusions: LAA thrombus is a rare occurrence before an elective catheter ablation. The underlying rhythm and kidney function may serve as markers of a higher likelihood of significantly reduced LAAEV and LAA thrombus.

Accuracy of Deep Learning Echocardiographic View Classification in Patients with Congenital or Structural Heart Disease: Importance of Specific Datasets.

INTRODUCTION: Automated echocardiography image interpretation has the potential to transform clinical practice. However, neural networks developed in general cohorts may underperform in the setting of altered cardiac anatomy. METHODS: Consecutive echocardiographic studies of patients with congenital or structural heart disease (C/SHD) were used to validate an existing convolutional neural network trained on 14,035 echocardiograms for automated view classification. In addition, a new convolutional neural network for view classification was trained and tested specifically in patients with C/SHD. RESULTS: Overall, 9793 imaging files from 262 patients with C/SHD (mean age 49 years, 60% male) and 62 normal controls (mean age 45 years, 50.0% male) were included. Congenital diagnoses included among others, tetralogy of Fallot (30), Ebstein anomaly (18) and transposition of the great arteries (TGA, 48). Assessing correct view classification based on 284,250 individual frames revealed that the non-congenital model had an overall accuracy of 48.3% for correct view classification in patients with C/SHD compared to 66.7% in patients without cardiac disease. Our newly trained convolutional network for echocardiographic view detection based on over 139,910 frames and tested on 35,614 frames from C/SHD patients achieved an accuracy of 76.1% in detecting the correct echocardiographic view. CONCLUSIONS: The current study is the first to validate view classification by neural networks in C/SHD patients. While generic models have acceptable accuracy in general cardiology patients, the quality of image classification is only modest in patients with C/SHD. In contrast, our model trained in C/SHD achieved a considerably increased accuracy in this particular cohort.

Frequency, Mortality, and Predictors of Adverse Outcomes for Endocarditis in Patients with Congenital Heart Disease: Results of a Nationwide Analysis including 2512 Endocarditis Cases.

BACKGROUND: Infective endocarditis (IE) represents a major complication in patients with congenital heart disease (CHD) and is associated with high morbidity and mortality. The aim of this study was to analyse the frequency and outcome of IE in contemporary CHD patients based on all IE hospital admissions in Germany over a 10-year period. METHODS: Based on data of all hospital admissions in Germany from 2009 to 2018, we identified all CHD cases with a diagnosis of IE. The data contained information on patient demographics, diagnoses, surgical procedures, and mortality. The primary endpoint of the study was endocarditis-associated mortality as well as major adverse events (defined as death or myocardial infarction, stroke, pulmonary embolism, sepsis, renal dialysis, resuscitation, or intubation). RESULTS: Overall, 309,245 CHD inpatient cases were included in the analysis (underlying heart defects of simple complexity 55%, moderate complexity 23%, and complex heart defects 22%, respectively). Of those, 2512 (0.8% of all inpatient cases) were treated for IE. The mortality rate of IE inpatient cases was 6% with a major adverse events rate of 46%, and 41.5% of cases required surgical intervention. The overall IE associated mortality was lower in adult CHD cases compared to the 153,242 in adult IE cases without CHD (7.1% vs. 16.1%, p < 0.001). After adjustments using multivariable logistic regression analysis, the presence or complexity of CHD was not associated with the outcomes. Meanwhile, age, male sex, and co-morbidities emerged as significant predictors of adverse outcomes. CONCLUSIONS: IE accounts for a minority of CHD related hospitalizations but remains a deadly disease, and major adverse events are common in this setting. Due to different demographic and co-morbidity spectrums, adult CHD patients tend to have better survival prospects when compared to non-CHD IE patients. Acquired co-morbidities emerged as the main predictors of adverse outcomes.

Clinical Value of Stress Transaortic Flow Rate During Dobutamine Echocardiography in Reduced Left Ventricular Ejection Fraction, Low-Gradient Aortic Stenosis: A Multicenter Study.

BACKGROUND: Low rest transaortic flow rate (FR) has been shown previously to predict mortality in low-gradient aortic stenosis. However limited prognostic data exists on stress FR during low-dose dobutamine stress echocardiography. We aimed to assess the value of stress FR for the detection of aortic valve stenosis (AS) severity and the prediction of mortality. METHODS: This is a multicenter cohort study of patients with reduced left ventricular ejection fraction and low-gradient aortic stenosis (aortic valve area <1 cm2 and mean gradient <40 mm Hg) who underwent low-dose dobutamine stress echocardiography to identify the AS severity and presence of flow reserve. The outcome assessed was all-cause mortality. RESULTS: Of the 287 patients (mean age, 75±10 years; males, 71%; left ventricular ejection fraction, 31±10%) over a mean follow-up of 24±30 months there were 127 (44.3%) deaths and 147 (51.2%) patients underwent aortic valve intervention. Higher stress FR was independently associated with reduced risk of mortality (hazard ratio, 0.97 [95% CI, 0.94-0.99]; P=0.01) after adjusting for age, chronic kidney disease, heart failure symptoms, aortic valve intervention, and rest left ventricular ejection fraction. The minimum cutoff for prediction of mortality was stress FR 210 mL/s. Following adjustment to the same important clinical and echocardiographic parameters, among the three criteria of AS severity during stress, ie, the guideline definition of aortic valve area <1cm2 and aortic valve mean gradient ≥40 mm Hg, or aortic valve mean gradient ≥40 mm Hg, or the novel definition of aortic valve area <1 cm2 at stress FR ≥210 mL/s, only the latter was independently associated with mortality (hazard ratio, 1.72 [95% CI, 1.05-2.82]; P=0.03). Furthermore aortic valve area <1cm2 at stress FR ≥210 mL/s was the only severe aortic stenosis criterion that was associated with improved outcome following aortic valve intervention (P<0.001). Guideline-defined stroke volume flow reserve did not predict mortality. CONCLUSIONS: Stress FR during low-dose dobutamine stress echocardiography was useful for the detection of both AS severity and flow reserve and was associated with improved prediction of outcome following aortic valve intervention.

Quantifying Left Atrial Size in the Context of Atrial Fibrillation Ablation: Which Echocardiographic Method Correlates to Outcome of Pulmonary Venous Isolation?

INTRODUCTION: Multiple studies have shown that left atrial (LA) enlargement is a strong predictor of poor outcome after catheter ablation of atrial fibrillation (AF). LA size is commonly approximated as the diameter in the parasternal long axis. It remains unknown whether more precise echocardiographic measurements of LA size allow for better correlation with outcome after pulmonary vein isolation (PVI). METHODS AND RESULTS: We performed a retrospective study of 131 consecutive patients (43 females, 60% paroxysmal AF, mean CHA2DS2-Vasc score 1.6, mean age 61 ± 12 years) referred for PVI. Measurements of the LA were carried out by a single observer in transthoracic echocardiograms (TTE) performed prior to ablation. We calculated diameter of the LA in the parasternal long axis (PLAX), LA area in the 2- as well as 4-Chamber (CH) view. LA volume was computed using the disc summation technique (LAV) and indexed to body surface area (LAVI). Procedural and follow-up data were gathered from a prospective AF database. Ablation was performed exclusively using the second generation cryoballoon by the same operators. Follow-up visits at 3, 6 and 12 months showed freedom from AF in 76%, 73% and 73% respectively. Mean values of LA calculations were LAPLAX: 37.9 mm ± 6.3 mm, 2CH area: 22.5 cm2 ± 6.7 cm2, 4CH area: 21.4 cm2 ± 5.5 cm2, LAV: 73.7 mL ± 26.1 mL and LAVI: 36.2 mL/m2 ± 12.7 mL/m2, respectively. C statistic revealed the best concordance of LAVI with outcome after 12 months (C = 0.67), LAV also exhibited a satisfactory value (C = 0.61) in comparison to surfaces in 2CH (C = 0.59) and 4CH (C = 0.57). PLAX showed the worst correlation (C = 0.51). Additionally, different binary logistic regression models identified three independent predictors of AF outcome after cryoballoon PVI: gender (OR = 0.95 per year; p = 0.01); LAV (OR = 1.3/10mL; p = 0.02) and LAVI (OR = 1.58/10 mL/m2; p = 0.02). In all models, PLAX and area measurements were not predictive. CONCLUSIONS: Our data add further to evidence that LA size lends itself well as a predictor of PVI outcome. LAVI and LAV were independently predictive of rhythm outcome after PVI. This did not hold true for more commonly used measurements, such as PLAX diameter and surfaces of the LA, irrespective of the view chosen.

Lack of specialist care is associated with increased morbidity and mortality in adult congenital heart disease: a population-based study.

AIMS: The aim of this study was to provide population-based data on the healthcare provision for adults with congenital heart disease (ACHD) and the impact of cardiology care on morbidity and mortality in this vulnerable population. METHODS AND RESULTS: Based on administrative data from one of the largest German Health Insurance Companies, all insured ACHD patients (<70 years of age) were included. Patients were stratified into those followed exclusively by primary care physicians (PCPs) and those with additional cardiology follow-up between 2014 and 2016. Associations between level of care and outcome were assessed by multivariable/propensity score Cox analyses. Overall, 24 139 patients (median age 43 years, 54.8% female) were included. Of these, only 49.7% had cardiology follow-up during the 3-year period, with 49.2% of patients only being cared for by PCPs and 1.1% having no contact with either. After comprehensive multivariable and propensity score adjustment, ACHD patients under cardiology follow-up had a significantly lower risk of death [hazard ratio (HR) 0.81, 95% confidence interval (CI) 0.67-0.98; P = 0.03) or major events (HR 0.85, 95% CI 0.78-0.92; P < 0.001) compared to those only followed by PCPs. At 3-year follow-up, the absolute risk difference for mortality was 0.9% higher in ACHD patients with moderate/severe complexity lesions cared by PCPs compared to those under cardiology follow-up. CONCLUSION: Cardiology care compared with primary care is associated with superior survival and lower rates of major complications in ACHD. It is alarming that even in a high resource setting with well-established specialist ACHD care approximately 50% of contemporary ACHD patients are still not linked to regular cardiac care. Almost all patients had at least one contact with a PCP during the study period, suggesting that opportunities to refer patients to cardiac specialists were missed at PCP level. More efforts are required to alert PCPs and patients to appropriate ACHD care.

Management considerations in the adult with surgically modified d-transposition of the great arteries.

Dextro-transposition of the great arteries (D-TGA) has undergone a significant evolution in surgical repair, leading to survivors with vastly different postsurgical anatomy which in turn guides their long-term cardiovascular morbidity and mortality. Atrial switch repair survivors are limited by a right ventricle in the systemic position, arrhythmia and atrial baffles prone to obstruction or leak. Functional assessment of the systemic right ventricle is complex, requiring multimodality imaging to include specialised echocardiography and cross-sectional imaging (MRI and CT). In the current era, most neonates undergo the arterial switch operation with increasing understanding of near-term and long-term outcomes specific to their cardiac anatomy. Long-term observations of the Lecompte manoeuvre or coronary stenoses following transfer continue, with evolving understanding to improve surveillance. Ultimately, an understanding of postsurgical anatomy, specialised imaging techniques and interventional and electrophysiological procedures is essential to comprehensive care of D-TGA survivors.

Heart Failure Results in Inspiratory Muscle Dysfunction Irrespective of Left Ventricular Ejection Fraction.

BACKGROUND: Exercise intolerance in heart failure with reduced ejection fraction (HFrEF) or heart failure with preserved ejection fraction (HFpEF) results from both cardiac dysfunction and skeletal muscle weakness. Respiratory muscle dysfunction with restrictive ventilation disorder may be present irrespective of left ventricular ejection fraction and might be mediated by circulating pro-inflammatory cytokines. OBJECTIVE: To determine lung and respiratory muscle function in patients with HFrEF/HFpEF and to determine its associations with exercise intolerance and markers of systemic inflammation. METHODS: Adult patients with HFrEF (n = 22, 19 male, 61 ± 14 years) and HFpEF (n = 8, 7 male, 68 ± 8 years) and 19 matched healthy control subjects underwent spirometry, measurement of maximum mouth occlusion pressures, diaphragm ultrasound, and recording of transdiaphragmatic and gastric pressures following magnetic stimulation of the phrenic nerves and the lower thoracic nerve roots. New York Heart Association (NYHA) class and 6-min walking distance (6MWD) were used to quantify exercise intolerance. Levels of circulating interleukin 6 (IL-6) and tumor necrosis factor-α (TNF-α) were measured using ELISAs. RESULTS: Compared with controls, both patient groups showed lower forced vital capacity (FVC) (p < 0.05), maximum inspiratory pressure (PImax), maximum expiratory pressure (PEmax) (p < 0.05), diaphragm thickening ratio (p = 0.01), and diaphragm strength (twitch transdiaphragmatic pressure in response to supramaximal cervical magnetic phrenic nerve stimulation) (p = 0.01). In patients with HFrEF, NYHA class and 6MWD were both inversely correlated with FVC, PImax, and PEmax. In those with HFpEF, there was an inverse correlation between amino terminal pro B-type natriuretic peptide levels and FVC (r = -0.77, p = 0.04). In all HF patients, IL-6 and TNF-α were statistically related to FVC. CONCLUSIONS: Irrespective of left ventricular ejection fraction, HF is associated with respiratory muscle dysfunction, which is associated with increased levels of circulating IL-6 and TNF-α.

Mortality and morbidity in patients with congenital heart disease hospitalised for viral pneumonia.

OBJECTIVES: Data on the clinical outcome of patients with congenital heart disease (CHD) affected by severe viral pneumonia are limited. We analysed morbidity and mortality of viral pneumonia and evaluated the association between medical conditions, medication, vaccination and outcome specifically in patients with CHD requiring hospitalisation for viral pneumonia. METHODS: Based on data from one of Germany's largest health insurers, all cases of viral pneumonia requiring hospital admission (2005-2018) were studied. Mortality, and composites of death, transplantation, mechanical circulatory support, ventilation or extracorporeal lung support served as endpoints. RESULTS: Overall, 26 262 viral pneumonia cases occurred in 24 980 patients. Of these, 1180 cases occurred in patients with CHD. Compared with patients without CHD, mortality rate was elevated in patients with CHD. As a group, patients with CHD aged 20-59 years even exceeded mortality rates in patients without CHD aged >60 years. No mortality was observed in patients with CHD with simple defects <60 years of age without associated cardiovascular risk factors. On multivariable logistic regression analysis, age, CHD complexity, chromosomal anomalies, cardiac medication, use of immunosuppressants and absence of vaccination for influenza emerged as risk factors of adverse outcome. CONCLUSIONS: We present timely data on morbidity and mortality of severe viral pneumonia requiring hospital admission in patients with CHD. Need for mechanical ventilation and risk of death in CHD increase early in life, reaching a level equivalent to non-CHD individuals >60 years of age. Our data suggest that except for patients with isolated simple defects, patients with CHD should be considered higher-risk individuals when faced with severe viral pneumonia.

Comparison of Early Surgical or Transcatheter Aortic Valve Replacement Versus Conservative Management in Low-Flow, Low-Gradient Aortic Stenosis Using Inverse Probability of Treatment Weighting: Results From the TOPAS Prospective Observational Cohort Study.

Background No randomized comparison of early (ie, ≤3 months) aortic valve replacement (AVR) versus conservative management or of transcatheter AVR (TAVR) versus surgical AVR has been conducted in patients with low-flow, low-gradient (LFLG) aortic stenosis (AS). Methods and Results A total of 481 consecutive patients (75±10 years; 71% men) with LFLG AS (aortic valve area ≤0.6 cm2/m2 and mean gradient <40 mm Hg), 72% with classic LFLG and 28% with paradoxical LFLG, were prospectively recruited in the multicenter TOPAS (True or Pseudo Severe Aortic Stenosis) study. True-severe AS or pseudo-severe AS was adjudicated by flow-independent criteria. During follow-up (median [IQR] 36 [11-60] months), 220 patients died. Using inverse probability of treatment weighting to address the bias of nonrandom treatment assignment, early AVR (n=272) was associated with a major overall survival benefit (hazard ratio [HR], 0.34 [95% CI, 0.24-0.50]; P<0.001). This benefit was observed in patients with true-severe AS but also with pseudo-severe AS (HR, 0.38 [95% CI, 0.18-0.81]; P=0.01), and in classic (HR, 0.33 [95% CI, 0.22-0.49]; P<0.001) and paradoxical LFLG AS (HR, 0.42 [95% CI, 0.20-0.92]; P=0.03). Compared with conservative management in the conventional multivariate model, trans femoral TAVR was associated with the best survival (HR, 0.23 [95% CI, 0.12-0.43]; P<0.001), followed by surgical AVR (HR, 0.36 [95% CI, 0.23-0.56]; P<0.001) and alternative-access TAVR (HR, 0.51 [95% CI, 0.31-0.82]; P=0.007). In the inverse probability of treatment weighting model, trans femoral TAVR appeared to be superior to surgical AVR (HR [95% CI] 0.28 [0.11-0.72]; P=0.008) with regard to survival. Conclusions In this large prospective observational study of LFLG AS, early AVR appeared to confer a major survival benefit in both classic and paradoxical LFLG AS. This benefit seems to extend to the subgroup with pseudo-severe AS. Our findings suggest that TAVR using femoral access might be the best strategy in these patients. Registration URL: https://www.clinicaltrials.gov; Unique identifier: NCT01835028.

Utility of deep learning networks for the generation of artificial cardiac magnetic resonance images in congenital heart disease.

BACKGROUND: Deep learning algorithms are increasingly used for automatic medical imaging analysis and cardiac chamber segmentation. Especially in congenital heart disease, obtaining a sufficient number of training images and data anonymity issues remain of concern. METHODS: Progressive generative adversarial networks (PG-GAN) were trained on cardiac magnetic resonance imaging (MRI) frames from a nationwide prospective study to generate synthetic MRI frames. These synthetic frames were subsequently used to train segmentation networks (U-Net) and the quality of the synthetic training images, as well as the performance of the segmentation network was compared to U-Net-based solutions trained entirely on patient data. RESULTS: Cardiac MRI data from 303 patients with Tetralogy of Fallot were used for PG-GAN training. Using this model, we generated 100,000 synthetic images with a resolution of 256 × 256 pixels in 4-chamber and 2-chamber views. All synthetic samples were classified as anatomically plausible by human observers. The segmentation performance of the U-Net trained on data from 42 separate patients was statistically significantly better compared to the PG-GAN based training in an external dataset of 50 patients, however, the actual difference in segmentation quality was negligible (< 1% in absolute terms for all models). CONCLUSION: We demonstrate the utility of PG-GANs for generating large amounts of realistically looking cardiac MRI images even in rare cardiac conditions. The generated images are not subject to data anonymity and privacy concerns and can be shared freely between institutions. Training supervised deep learning segmentation networks on this synthetic data yielded similar results compared to direct training on original patient data.