Machine learning prediction of atrial fibrillation in cardiovascular patients using cardiac magnetic resonance and electronic health information

Background

Atrial fibrillation (AF) is a commonly encountered cardiac arrhythmia associated with morbidity and substantial healthcare costs. While patients with cardiovascular disease experience the greatest risk of new-onset AF, no risk model has been developed to predict AF occurrence in this population. We hypothesized that a patient-specific model could be delivered using cardiovascular magnetic resonance (CMR) disease phenotyping, contextual patient health information, and machine learning.

Methods

Nine thousand four hundred forty-eight patients referred for CMR imaging were enrolled and followed over a 5-year period. Seven thousand, six hundred thirty-nine had no prior history of AF and were eligible to train and validate machine learning algorithms. Random survival forests (RSFs) were used to predict new-onset AF and compared to Cox proportional-hazard (CPH) models. The best performing features were identified from 115 variables sourced from three data domains: (i) CMR-based disease phenotype, (ii) patient health questionnaire, and (iii) electronic health records. We evaluated discriminative performance of optimized models using C-index and time-dependent AUC (tAUC).

Results

A RSF-based model of 20 variables (CIROC-AF-20) delivered an overall C-index of 0.78 for the prediction of new-onset AF with respective tAUCs of 0.80, 0.79, and 0.78 at 1-, 2- and 3-years. This outperformed a novel CPH-based model and historic AF risk scores. At 1-year of follow-up, validation cohort patients classified as high-risk of future AF by CIROC-AF-20 went on to experience a 17.3% incidence of new-onset AF, being 24.7-fold higher risk than low risk patients.

Conclusions

Using phenotypic data available at time of CMR imaging we developed and validated the first described risk model for the prediction of new-onset AF in patients with cardiovascular disease. Complementary value was provided by variables from patient-reported measures of health and the electronic health record, illustrating the value of multi-domain phenotypic data for the prediction of AF.

Machine Learning Patient-Specific Prediction of Heart Failure Hospitalization Using Cardiac MRI-Based Phenotype and Electronic Health Information

Background

Heart failure (HF) hospitalization is a dominant contributor of morbidity and healthcare expenditures in patients with systolic HF. Cardiovascular magnetic resonance (CMR) imaging is increasingly employed for the evaluation of HF given capacity to provide highly reproducible phenotypic markers of disease. The combined value of CMR phenotypic markers and patient health information to deliver predictions of future HF events has not been explored. We sought to develop and validate a novel risk model for the patient-specific prediction of time to HF hospitalization using routinely reported CMR variables, patient-reported health status, and electronic health information.

Methods

Standardized data capture was performed for 1,775 consecutive patients with chronic systolic HF referred for CMR imaging. Patient demographics, symptoms, Health-related Quality of Life, pharmacy, and routinely reported CMR features were provided to both machine learning (ML) and competing risk Fine-Gray-based models (FGM) for the prediction of time to HF hospitalization.

Results

The mean age was 59 years with a mean LVEF of 36 ± 11%. The population was evenly distributed between ischemic (52%) and idiopathic non-ischemic cardiomyopathy (48%). Over a median follow-up of 2.79 years (IQR: 1.59–4.04) 333 patients (19%) experienced HF related hospitalization. Both ML and competing risk FGM based models achieved robust performance for the prediction of time to HF hospitalization. Respective 90-day, 1 and 2-year AUC values were 0.87, 0.83, and 0.80 for the ML model, and 0.89, 0.84, and 0.80 for the competing risk FGM-based model in a holdout validation cohort. Patients classified as high-risk by the ML model experienced a 34-fold higher occurrence of HF hospitalization at 90 days vs. the low-risk group.

Conclusion

In this study we demonstrated capacity for routinely reported CMR phenotypic markers and patient health information to be combined for the delivery of patient-specific predictions of time to HF hospitalization. This work supports an evolving migration toward multi-domain data collection for the delivery of personalized risk prediction at time of diagnostic imaging.

Right ventricular insertion site fibrosis in a dilated cardiomyopathy referral population: phenotypic associations and value for the prediction of heart failure admission or death

BACKGROUND

Dilated cardiomyopathy (DCM) is increasingly recognized as a heterogenous disease with distinct phenotypes on late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging. While mid-wall striae (MWS) fibrosis is a widely recognized phenotypic risk marker, other fibrosis patterns are prevalent but poorly defined. Right ventricular (RV) insertion (RVI) site fibrosis is commonly seen, but without objective criteria has been considered a non-specific finding. In this study we developed objective criteria for RVI fibrosis and studied its clinical relevance in a large cohort of patients with DCM.

METHODS

We prospectively enrolled 645 DCM patients referred for LGE-CMR. All underwent standardized imaging protocols and baseline health evaluations. LGE images were blindly scored using objective criteria, inclusive of RVI site and MWS fibrosis. Associations between LGE patterns and CMR-based markers of adverse chamber remodeling were evaluated. Independent associations of LGE fibrosis patterns with the primary composite clinical outcome of heart failure admission or death were determined by multivariable analysis.

RESULTS

The mean age was 56 ± 14 (28% female) with a mean left ventricular (LV) ejection fraction (LVEF) of 37%. At a median of 1061 days, 129 patients (20%) experienced the primary outcome. Any abnormal LGE was present in 306 patients (47%), inclusive of 274 (42%) meeting criteria for RVI site fibrosis and 167 (26%) for MWS fibrosis. All with MWS fibrosis showed RVI site fibrosis. Solitary RVI site fibrosis was associated with higher bi-ventricular volumes [LV end-systolic volume index (78 ± 39 vs. 66 ± 33 ml/m2, p = 0.01), RV end-diastolic volume index (94 ± 28 vs. 84 ± 22 ml/m2 (p < 0.01), RV end-systolic volume index (56 ± 26 vs. 45 ± 17 ml/m2, p < 0.01)], lower bi-ventricular function [LVEF 35 ± 12 vs. 39 ± 10% (p < 0.01), RV ejection fraction (RVEF) 43 ± 12 vs. 48 ± 10% (p < 0.01)], and higher extracellular volume (ECV). Patient with solitary RVI site fibrosis experienced a non-significant 1.4-fold risk of the primary outcome, increasing to a significant 2.6-fold risk when accompanied by MWS fibrosis.

CONCLUSIONS

RVI site fibrosis in the absence of MWS fibrosis is associated with bi-ventricular remodelling and intermediate risk of heart failure admission or death. Our study findings suggest RVI site fibrosis to be pre-requisite for the incremental development of MWS fibrosis, a more advanced phenotype associated with greater LV remodeling and risk of clinical events.

Effect of Active Cancer on the Cardiac Phenotype: A Cardiac Magnetic Resonance Imaging-Based Study of Myocardial Tissue Health and Deformation in Patients With Chemotherapy-Naïve Cancer

Background

The overlap between cancer and cardiovascular care continues to expand, with intersections emerging before, during, and following cancer therapies. To date, emphasis has been placed on how cancer therapeutics influence downstream cardiac health. However, whether active malignancy itself influences chamber volumes, function, or overall myocardial tissue health remains uncertain. We sought to perform a comprehensive cardiovascular magnetic resonance‐based evaluation of cardiac health in patients with chemotherapy‐naïve cancer with comparison with a healthy volunteer population.

Methods and Results

Three‐hundred and eighty‐one patients with active breast cancer or lymphoma before cardiotoxic chemotherapy exposure were recruited in addition to 102 healthy volunteers. Both cohorts underwent standardized cardiovascular magnetic resonance imaging with quantification of chamber volumes, ejection fraction, and native myocardial T1. Left ventricular mechanics were incrementally assessed using three‐dimensional myocardial deformation analysis, providing global longitudinal, circumferential, radial, and principal peak‐systolic strain amplitude and systolic strain rate. The mean age of patients with cancer was 53.8±13.4 years; 79% being women. Despite similar left ventricular ejection fraction, patients with cancer showed smaller chambers, increased strain amplitude, and systolic strain rate in both conventional and principal directions, and elevated native T1 versus sex‐matched healthy volunteers. Adjusting for age, sex, hypertension, and diabetes mellitus, the presence of cancer remained associated with these cardiovascular magnetic resonance parameters.

Conclusions

The presence of cancer is independently associated with alterations in cardiac chamber size, function, and objective markers of tissue health. Dedicated research is warranted to elucidate pathophysiologic mechanisms underlying these findings and to explore their relevance to the management of patients with cancer referred for cardiotoxic therapies.

Right Ventricular Ejection Fraction for the Prediction of Major Adverse Cardiovascular and Heart Failure-Related Events: A Cardiac MRI Based Study of 7131 Patients With Known or Suspected Cardiovascular Disease

BACKGROUND

There is increasing evidence that right ventricular ejection fraction (RVEF) may provide incremental value to left ventricular (LV) ejection fraction for the prediction of major adverse cardiovascular events. To date, generalizable utility for RVEF quantification in patients with cardiovascular disease has not been established. Using a large prospective clinical outcomes registry, we investigated the prognostic value of RVEF for the prediction of major adverse cardiovascular events- and heart failure-related outcomes.

METHODS

Seven thousand one hundred thirty-one consecutive patients with known or suspected cardiovascular disease undergoing cardiovascular magnetic resonance imaging were prospectively enrolled. Multichamber volumetric quantification was performed by standardized operational procedures. Patients were followed for the primary composite outcome of all-cause death, survived cardiac arrest, admission for heart failure, need for transplantation or LV assist device, acute coronary syndrome, need for revascularization, stroke, or transient ischemic attack. A secondary, heart failure focused outcome of heart failure admission, need for transplantation/LV assist device or death was also studied.

RESULTS

Mean age was 54±15 years. The mean LV ejection fraction was 55±14% (range 6%-90%) with a mean RVEF of 54±10% (range 9%-87%). At a median follow-up of 908 days, 870 (12%) patients experienced the primary composite outcome and 524 (7%) the secondary outcome. Each 10% drop in RVEF was associated with a 1.3-fold increased risk of the primary outcome (P<0.001) and 1.5-fold increased risk of the secondary outcome (P<0.001). RVEF was an independent predictor following comprehensive covariate adjustment, inclusive of LV ejection fraction. Patients with an RVEF<40% experienced a 3.1-fold risk of the primary outcome (P<0.001) with a 1-year cumulative event rate of 22% versus 7% above this cutoff.

CONCLUSIONS

RVEF is a powerful and independent predictor of major adverse cardiac events with broad generalizability across patients with known or suspected cardiovascular disease. These findings support migration towards biventricular phenotyping for the classification of risk in clinical practice. Registration: URL: https://www.clinicaltrials.gov; Unique identifier: NCT04367220.

Neural-Network-Based Diagnosis Using 3-Dimensional Myocardial Architecture and Deformation: Demonstration for the Differentiation of Hypertrophic Cardiomyopathy

The diagnosis of cardiomyopathy states may benefit from machine-learning (ML) based approaches, particularly to distinguish those states with similar phenotypic characteristics. Three-dimensional myocardial deformation analysis (3D-MDA) has been validated to provide standardized descriptors of myocardial architecture and deformation, and may therefore offer appropriate features for the training of ML-based diagnostic tools. We aimed to assess the feasibility of automated disease diagnosis using a neural network trained using 3D-MDA to discriminate hypertrophic cardiomyopathy (HCM) from its mimic states: cardiac amyloidosis (CA), Anderson–Fabry disease (AFD), and hypertensive cardiomyopathy (HTNcm). 3D-MDA data from 163 patients (mean age 53.1 ± 14.8 years; 68 females) with left ventricular hypertrophy (LVH) of known etiology was provided. Source imaging data was from cardiac magnetic resonance (CMR). Clinical diagnoses were as follows: 85 HCM, 30 HTNcm, 30 AFD, and 18 CA. A fully-connected-layer feed-forward neural was trained to distinguish HCM vs. other mimic states. Diagnostic performance was compared to threshold-based assessments of volumetric and strain-based CMR markers, in addition to baseline clinical patient characteristics. Threshold-based measures provided modest performance, the greatest area under the curve (AUC) being 0.70. Global strain parameters exhibited reduced performance, with AUC under 0.64. A neural network trained exclusively from 3D-MDA data achieved an AUC of 0.94 (sensitivity 0.92, specificity 0.90) when performing the same task. This study demonstrates that ML-based diagnosis of cardiomyopathy states performed exclusively from 3D-MDA is feasible and can distinguish HCM from mimic disease states. These findings suggest strong potential for computer-assisted diagnosis in clinical practice.

Detection of papillary muscle infarction by late gadolinium enhancement: incremental value of short-inversion time vs. standard imaging

AIMS

Late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging can detect myocardial scar in patients with myocardial infarction. The detection of papillary muscle infarction (PMI) may be difficult due to the bright blood signal. The aim of our study was to evaluate the incremental value of LGE CMR imaging using an inversion recovery (IR)-GRE with a short-inversion time (TI) over standard LGE imaging in identifying PMI.

METHODS AND RESULTS

Fifty-six patients with myocardial infarction were studied using a standard IR-GRE LGE sequence with an adjusted TI to null the signal intensity of normal myocardium and with a 3D IR-GRE with a short TI (<180 ms). Signal-to-noise and contrast-to-noise ratios (CNR) and the frequency of PMI were determined. Image quality and infarction sharpness were evaluated. The short-TI LGE sequence detected a higher number of PMI compared with standard LGE sequence (19/54 vs. 15/54) with an increased sharpness of PMI (84.2 vs. 53.3%). The CNR was higher between infarcted myocardium and blood (77.9 ± 60 vs. 19.3 ± 16, P < 0.001) and between PMI and blood (69.4 ± 51 vs. 39.4 ± 26, respectively, P = 0.0157).

CONCLUSIONS

Our data indicate that in patients with myocardial infarction, LGE CMR imaging using a short TI may be more sensitive than standard LGE imaging for the detection of PMI.