Search

Abstract The efficacy of an implantable cardioverter-defibrillator (ICD) in patients with a non-ischaemic cardiomyopathy for primary prevention of sudden cardiac death is increasingly debated. We developed a multimodal deep learning model for arrhythmic risk prediction that integrated late gadolinium enhanced (LGE) cardiac magnetic resonance imaging (MRI), electrocardiography (ECG) and clinical data. Short-axis LGE-MRI scans and 12-lead ECGs were retrospectively collected from a cohort of 289 patients prior to ICD implantation, across two tertiary hospitals. A residual variational autoencoder was developed to extract physiological features from LGE-MRI and ECG, and used as inputs for a machine learning model (DEEP RISK) to predict malignant ventricular arrhythmia onset. In the validation cohort, the multimodal DEEP RISK model predicted malignant ventricular arrhythmias with an area under the receiver operating characteristic curve (AUROC) of 0.84 (95% confidence interval (CI) 0.71–0.96), a sensitivity of 0.98 (95% CI 0.75–1.00) and a specificity of 0.73 (95% CI 0.58–0.97). The models trained on individual modalities exhibited lower AUROC values compared to DEEP RISK [MRI branch: 0.80 (95% CI 0.65–0.94), ECG branch: 0.54 (95% CI 0.26–0.82), Clinical branch: 0.64 (95% CI 0.39–0.87)]. These results suggest that a multimodal model achieves high prognostic accuracy in predicting ventricular arrhythmias in a cohort of patients with non-ischaemic systolic heart failure, using data collected prior to ICD implantation.

Summary: Background: Risk stratification for ventricular arrhythmias currently relies on static measurements that fail to adequately capture dynamic interactions between arrhythmic substrate and triggers over time. We trained and internally validated a dynamic machine learning (ML) model and neural network that extracted features from longitudinally collected electrocardiograms (ECG), and used these to predict the risk of malignant ventricular arrhythmias. Methods: A multicentre study in patients implanted with an implantable cardioverter-defibrillator (ICD) between 2007 and 2021 in two academic hospitals was performed. Variational autoencoders (VAEs), which combine neural networks with variational inference principles, and can learn patterns and structure in data without explicit labelling, were trained to encode the mean ECG waveforms from the limb leads into 16 variables. Supervised dynamic ML models using these latent ECG representations and clinical baseline information were trained to predict malignant ventricular arrhythmias treated by the ICD. Model performance was evaluated on a hold-out set, using time-dependent receiver operating characteristic (ROC) and calibration curves. Findings: 2942 patients (61.7 ± 13.9 years, 25.5% female) were included, with a total of 32,129 ECG recordings during a mean follow-up of 43.9 ± 35.9 months. The mean time-varying area under the ROC curve for the dynamic model was 0.738 ± 0.07, compared to 0.639 ± 0.03 for a static (i.e. baseline-only model). Feature analyses indicated dynamic changes in latent ECG representations, particularly those affecting the T-wave morphology, were of highest importance for model predictions. Interpretation: Dynamic ML models and neural networks effectively leverage routinely collected longitudinal ECG recordings for personalised and updated predictions of malignant ventricular arrhythmias, outperforming static models. Funding: This publication is part of the project DEEP RISK ICD (with project number 452019308) of the research programme Rubicon which is (partly) financed by the Dutch Research Council (NWO). This research is partly funded by the Amsterdam Cardiovascular Sciences (personal grant F.V.Y.T).

Summary: Background: Ventricular arrhythmia (VA) precipitating sudden cardiac arrest (SCD) is among the most frequent causes of death and pose a high burden on public health systems worldwide. The increasing availability of electrophysiological signals collected through conventional methods (e.g. electrocardiography (ECG)) and digital health technologies (e.g. wearable devices) in combination with novel predictive analytics using machine learning (ML) and deep learning (DL) hold potential for personalised predictions of arrhythmic events. Methods: This systematic review and exploratory meta-analysis assesses the state-of-the-art of ML/DL models of electrophysiological signals for personalised prediction of malignant VA or SCD, and studies potential causes of bias (PROSPERO, reference: CRD42021283464). Five electronic databases were searched to identify eligible studies. Pooled estimates of the diagnostic odds ratio (DOR) and summary area under the curve (AUROC) were calculated. Meta-analyses were performed separately for studies using publicly available, ad-hoc datasets, versus targeted clinical data acquisition. Studies were scored on risk of bias by the PROBAST tool. Findings: 2194 studies were identified of which 46 were included in the systematic review and 32 in the meta-analysis. Pooling of individual models demonstrated a summary AUROC of 0.856 (95% CI 0.755–0.909) for short-term (time-to-event up to 72 h) prediction and AUROC of 0.876 (95% CI 0.642–0.980) for long-term prediction (time-to-event up to years). While models developed on ad-hoc sets had higher pooled performance (AUROC 0.919, 95% CI 0.867–0.952), they had a high risk of bias related to the re-use and overlap of small ad-hoc datasets, choices of ML tool and a lack of external model validation. Interpretation: ML and DL models appear to accurately predict malignant VA and SCD. However, wide heterogeneity between studies, in part due to small ad-hoc datasets and choice of ML model, may reduce the ability to generalise and should be addressed in future studies. Funding: This publication is part of the project DEEP RISK ICD (with project number 452019308) of the research programme Rubicon which is (partly) financed by the Dutch Research Council (NWO). This research is partly funded by the Amsterdam Cardiovascular Sciences (personal grant F.V.Y.T).

Background: Current implantable cardioverter-defibrillator (ICD) devices are equipped with a device-embedded accelerometer capable of capturing physical activity (PA). In contrast, wearable accelerometer-based methods enable the measurement of physical behavior (PB) that encompasses not only PA but also sleep behavior, sedentary time, and rest-activity patterns. Objective: This systematic review evaluates accelerometer-based methods used in patients carrying an ICD or at high risk of sudden cardiac death. Methods: Papers were identified via the OVID MEDLINE and OVID EMBASE databases. PB could be assessed using a wearable accelerometer or an embedded accelerometer in the ICD. Results: A total of 52 papers were deemed appropriate for this review. Out of these studies, 30 examined device-embedded accelerometry (189,811 patients), 19 examined wearable accelerometry (1601 patients), and 3 validated wearable accelerometry against device-embedded accelerometry (106 patients). The main findings were that a low level of PA after implantation of the ICD and a decline in PA were both associated with an increased risk of mortality, heart failure hospitalization, and appropriate ICD shock. Second, PA was affected by cardiac factors (eg, onset of atrial fibrillation, ICD shocks) and noncardiac factors (eg, seasonal differences, societal factors). Conclusion: This review demonstrated the potential of accelerometer-measured PA as a marker of clinical deterioration and ventricular arrhythmias. Notwithstanding that the evidence of PB assessed using wearable accelerometry was limited, there seems to be potential for accelerometers to improve early warning systems and facilitate preventative and proactive strategies.

Background: The Apple Watch (AW) is the first commercially available wearable device with built-in electrocardiogram (ECG) electrodes to perform a single-lead ECG to detect atrial fibrillation (AF). Methods: Patients with AF who were scheduled for electrical cardioversion (ECV) were included in this study. The AW ECGs were obtained pre-ECV and post-ECV. In case of an unclassified recording, the AW ECG was obtained up to 3 times. The 12-lead ECG was used as the reference standard. Sensitivity, specificity, and kappa coefficient were calculated. Results: In total, 74 patients were included. Mean age was 67.1 ± 12.3 years and 20.3% were female. In total 65 AF and 64 sinus rhythm measurements were obtained. The first measurement with the AW showed a sensitivity of 93.5% and specificity of 100% (κ = 0.94). A second measurement resulted in a sensitivity of 94.6% and specificity of 100% (κ = 0.95). A third measurement resulted in a sensitivity of 93% and a specificity of 96.5% (κ = 0.90). Adjudication of unclassified recordings by a physician reduced the total unclassified recordings from 27.9% to 1.6%, but also reduced the accuracy. The kappa coefficient for unclassified single-lead ECGs was 0.58. Conclusion: The single-lead ECG of the AW shows a high accuracy for identifying AF in a clinical setting. Repeating the recording once decreases the total of unclassified recordings; however, a third recording resulted in a lower accuracy and the occurrence of false-positive measurements. Unclassified results of the AW can be reduced by physicians’ interpretation of the single-lead ECG; however, the interrater agreement is only moderate.

Background: Patients with an implantable cardioverter-defibrillator (ICD) are at a high risk of malignant ventricular arrhythmias. The use of remote ICD monitoring, wearable devices, and patient-reported outcomes generate large volumes of potential valuable data. Artificial intelligence–based methods can be used to develop personalized prediction models and improve early-warning systems. Objective: The purpose of this study was to develop an integrated web-based personalized prediction engine for ICD therapy. Methods: This international, multicenter, prospective, observational study consists of 2 phases: (1) a development study and (2) a feasibility study. We plan to enroll 400 participants with an ICD (with or without cardiac resynchronization therapy) on remote monitoring: 300 participants in the development study and 100 in the feasibility study. During 12-month follow-up, electronic health record data, remote monitoring data, accelerometry-assessed physical behavior data, and patient-reported data are collected. By using machine- and deep-learning approaches, a prediction engine is developed to assess the risk probability of ICD therapy (shock and antitachycardia pacing). The feasibility of the prediction engine as a clinical tool, the SafeHeart Platform, is assessed during the feasibility study. Results: Development study recruitment commenced in 2021. The feasibility study starts in 2022. Conclusion: SafeHeart is the first study to prospectively collect a multimodal data set to construct a personalized prediction engine for ICD therapy. Moreover, SafeHeart explores the integration and added value of detailed objective accelerometer data in the prediction of clinical events. The translation of the SafeHeart Platform to clinical practice is examined during the feasibility study.

BackgroundPrevious studies report conflicting results about a higher incidence of ventricular arrhythmias in patients with a chronic total coronary occlusion (CTO). We aimed to investigate this association in a large cohort of implantable cardioverter defibrillator patients with long‐term follow‐up. Methods and ResultsAll consecutive patients from 1992 onwards who underwent implantable cardioverter defibrillator implantation for ischemic cardiomyopathy at the Leiden University Medical Center were evaluated. Coronary angiograms were reviewed for the presence of a CTO. The occurrence of ventricular arrhythmias and survival status at follow‐up were compared between patients with and patients without a CTO. A total of 722 patients constitute the study cohort (age 66±11 years; 84% males; 74% primary prevention, median left ventricular ejection fraction 30% [first–third quartile: 25–37], 44% received a cardiac resynchronization therapy defibrillator). At baseline, 240 patients (33%) had a CTO, and the CTOs were present for at least 44 (2–127) months. The median follow‐up duration was 4 (2–6) years. On long‐term follow‐up, CTO patients had a higher crude appropriate device therapy rate (37% versus 27%, P=0.010) and a lower crude survival rate (51% versus 67%, P