ECG-only explainable deep learning algorithm predicts the risk for malignant ventricular arrhythmia in phospholamban cardiomyopathy.

BACKGROUND: Phospholamban (PLN) p.(Arg14del) variant carriers are at risk for development of malignant ventricular arrhythmia (MVA). Accurate risk stratification allows timely implantation of intracardiac defibrillators and is currently performed with a multimodality prediction model. OBJECTIVE: This study aimed to investigate whether an explainable deep learning-based approach allows risk prediction with only electrocardiogram (ECG) data. METHODS: A total of 679 PLN p.(Arg14del) carriers without MVA at baseline were identified. A deep learning-based variational auto-encoder, trained on 1.1 million ECGs, was used to convert the 12-lead baseline ECG into its FactorECG, a compressed version of the ECG that summarizes it into 32 explainable factors. Prediction models were developed by Cox regression. RESULTS: The deep learning-based ECG-only approach was able to predict MVA with a C statistic of 0.79 (95% CI, 0.76-0.83), comparable to the current prediction model (C statistic, 0.83 [95% CI, 0.79-0.88]; P = .054) and outperforming a model based on conventional ECG parameters (low-voltage ECG and negative T waves; C statistic, 0.65 [95% CI, 0.58-0.73]; P < .001). Clinical simulations showed that a 2-step approach, with ECG-only screening followed by a full workup, resulted in 60% less additional diagnostics while outperforming the multimodal prediction model in all patients. A visualization tool was created to provide interactive visualizations (https://pln.ecgx.ai). CONCLUSION: Our deep learning-based algorithm based on ECG data only accurately predicts the occurrence of MVA in PLN p.(Arg14del) carriers, enabling more efficient stratification of patients who need additional diagnostic testing and follow-up.

Toward advanced diagnosis and management of inherited arrhythmia syndromes: Harnessing the capabilities of artificial intelligence and machine learning.

The use of advanced computational technologies, such as artificial intelligence (AI), is now exerting a significant influence on various aspects of life, including health care and science. AI has garnered remarkable public notice with the release of deep learning models that can model anything from artwork to academic papers with minimal human intervention. Machine learning, a method that uses algorithms to extract information from raw data and represent it in a model, and deep learning, a method that uses multiple layers to progressively extract higher-level features from the raw input with minimal human intervention, are increasingly leveraged to tackle problems in the health sector, including utilization for clinical decision support in cardiovascular medicine. Inherited arrhythmia syndromes are a clinical domain where multiple unanswered questions remain despite unprecedented progress over the past 2 decades with the introduction of large panel genetic testing and the first steps in precision medicine. In particular, AI tools can help address gaps in clinical diagnosis by identifying individuals with concealed or transient phenotypes; enhance risk stratification by elevating recognition of underlying risk burden beyond widely recognized risk factors; improve prediction of response to therapy, and further prognostication. In this contemporary review, we provide a summary of the AI models developed to solve challenges in inherited arrhythmia syndromes and also outline gaps that can be filled with the development of intelligent AI models.

ECG Changes during Adult Life in Fabry Disease: Results from a Large Longitudinal Cohort Study.

Background: Fabry disease (FD) is an X-linked, lysosomal storage disorder leading to severe cardiomyopathy in a significant proportion of patients. To identify ECG markers that reflect early cardiac involvement and disease progression, we conducted a long term retrospective study in a large cohort of FD patients. Methods: A total of 1995 ECGs from 133 patients with classical FD (64% females, 80% treated with enzyme replacement therapy), spanning 20 years of follow-up, were compared to ECGs from 3893 apparently healthy individuals. Generalized linear mixed models were used to evaluate the effect of age, FD and sex on: P-wave duration, PR-interval, QRS-duration, QTc, Cornell index, spatial QRS-T angle and frontal QRS-axis. Regression slopes and absolute values for each parameter were compared between FD patients and control subjects. Results: At a younger age (<40 years), the Cornell index was higher and frontal QRS-axis more negative in FD patients compared to controls (p < 0.05). For the other ECG parameters, the rate of change, more than the absolute value, was greater in FD patients compared to controls (p < 0.05). From the fifth decade (men) or sixth (women) onwards, absolute values for P-wave duration, QRS-duration, QTc and spatial QRS-T angle were longer and higher in FD patients compared to control subjects. Conclusions: ECG abnormalities indicative of FD are age and sex dependent. Tracking the rate of change in ECG parameters could be a good way to detect disease progression, guiding treatment initiation. Moreover, monitoring ECG changes in FD can be used to evaluate the effectiveness of treatment.

Clinical applicability of artificial intelligence for patients with an inherited heart disease: A scoping review.

The number of inherited heart disease (IHD) studies using artificial intelligence (AI) has increased rapidly over the last years. In this scoping review, we aimed to present an overview of the current literature available on the applicability of AI within the field of IHD. The literature search resulted in eighteen articles in which an AI model was trained and tested, mostly for diagnostic and predictive purposes. The areas under the receiver operating characteristic curves ranged from 0.78-0.96, but varied between IHD types, used methods and outcome measures. Only three out of eighteen did perform validation on an external dataset and most studies did not use explainable deep learning models. To be able to integrate AI as a tool to aid physicians in their diagnoses and clinical decisions within the IHD field, generalizability has to be better evaluated and explainability of DL models has to be increased.

A deep learning approach identifies new ECG features in congenital long QT syndrome.

BACKGROUND: Congenital long QT syndrome (LQTS) is a rare heart disease caused by various underlying mutations. Most general cardiologists do not routinely see patients with congenital LQTS and may not always recognize the accompanying ECG features. In addition, a proportion of disease carriers do not display obvious abnormalities on their ECG. Combined, this can cause underdiagnosing of this potentially life-threatening disease. METHODS: This study presents 1D convolutional neural network models trained to identify genotype positive LQTS patients from electrocardiogram as input. The deep learning (DL) models were trained with a large 10-s 12-lead ECGs dataset provided by Amsterdam UMC and externally validated with a dataset provided by University Hospital Leuven. The Amsterdam dataset included ECGs from 10000 controls, 172 LQTS1, 214 LQTS2, and 72 LQTS3 patients. The Leuven dataset included ECGs from 2200 controls, 32 LQTS1, and 80 LQTS2 patients. The performance of the DL models was compared with conventional QTc measurement and with that of an international expert in congenital LQTS (A.A.M.W). Lastly, an explainable artificial intelligence (AI) technique was used to better understand the prediction models. RESULTS: Overall, the best performing DL models, across 5-fold cross-validation, achieved on average a sensitivity of 84 ± 2%, 90 ± 2% and 87 ± 6%, specificity of 96 ± 2%, 95 ± 1%, and 92 ± 4%, and AUC of 0.90 ± 0.01, 0.92 ± 0.02, and 0.89 ± 0.03, for LQTS 1, 2, and 3 respectively. The DL models were also shown to perform better than conventional QTc measurements in detecting LQTS patients. Furthermore, the performances held up when the DL models were validated on a novel external cohort and outperformed the expert cardiologist in terms of specificity, while in terms of sensitivity, the DL models and the expert cardiologist in LQTS performed the same. Finally, the explainable AI technique identified the onset of the QRS complex as the most informative region to classify LQTS from non-LQTS patients, a feature previously not associated with this disease. CONCLUSIONS: This study suggests that DL models can potentially be used to aid cardiologists in diagnosing LQTS. Furthermore, explainable DL models can be used to possibly identify new features for LQTS on the ECG, thus increasing our understanding of this syndrome.

Sex- and age specific association of new-onset atrial fibrillation with in-hospital mortality in hospitalised COVID-19 patients.

BACKGROUND: Coronavirus disease 2019 (COVID-19) is a systemic disease with cardiovascular involvement, including cardiac arrhythmias. Notably, new-onset atrial fibrillation (AF) and atrial flutter (AFL) during hospitalisation in COVID-19 patients has been associated with increased mortality. However, how this risk is impacted by age and sex is still poorly understood. METHODS: For this multicentre cohort study, we extracted demographics, medical history, occurrence of electrical disorders and in-hospital mortality from the large international patient registry CAPACITY-COVID. For each electrical disorder, prevalence during hospitalisation was calculated. Subsequently, we analysed the incremental prognostic effect of developing AF/AFL on in-hospital mortality, using multivariable logistic regression analyses, stratified for sex and age. RESULTS: In total, 5782 patients (64% male; median age 67) were included. Of all patients 11.0% (95% CI 10.2-11.8) experienced AF and 1.6% (95% CI 1.3-1.9) experienced AFL during hospitalisation. Ventricular arrhythmias were rare (<0.8% (95% CI 0.6-1.0)) and a conduction disorder was observed in 6.3% (95% CI 5.7-7.0). An event of AF/AFL appeared to occur more often in patients with pre-existing heart failure. After multivariable adjustment for age and sex, new-onset AF/AFL was significantly associated with a poorer prognosis, exemplified by a two- to three-fold increased risk of in-hospital mortality in males aged 60-72 years, whereas this effect was largely attenuated in older male patients and not observed in female patients. CONCLUSION: In this large COVID-19 cohort, new-onset AF/AFL was associated with increased in-hospital mortality, yet this increased risk was restricted to males aged 60-72 years.

Improving electrocardiogram-based detection of rare genetic heart disease using transfer learning: An application to phospholamban p.Arg14del mutation carriers.

The pathogenic mutation p.Arg14del in the gene encoding Phospholamban (PLN) is known to cause cardiomyopathy and leads to increased risk of sudden cardiac death. Automatic tools might improve the detection of patients with this rare disease. Deep learning is currently the state-of-the-art in signal processing but requires large amounts of data to train the algorithms. In situations with relatively small amounts of data, like PLN, transfer learning may improve accuracy. We propose an ECG-based detection of the PLN mutation using transfer learning from a model originally trained for sex identification. The sex identification model was trained with 256,278 ECGs and subsequently finetuned for PLN detection (155 ECGs of patients with PLN) with two control groups: a balanced age/sex matched group and a randomly selected imbalanced population. The data was split in 10 folds and 20% of the training data was used for validation and early stopping. The models were evaluated with the area under the receiver operating characteristic curve (AUROC) of the testing data. We used gradient activation for explanation of the prediction models. The models trained with transfer learning outperformed the models trained from scratch for both the balanced (AUROC 0.87 vs AUROC 0.71) and imbalanced (AUROC 0.0.90 vs AUROC 0.65) population. The proposed approach was able to improve the accuracy of a rare disease detection model by transfer learning information from a non-manual annotated and abundant label with only limited data available.

Computer versus cardiologist: Is a machine learning algorithm able to outperform an expert in diagnosing a phospholamban p.Arg14del mutation on the electrocardiogram?

BACKGROUND: Phospholamban (PLN) p.Arg14del mutation carriers are known to develop dilated and/or arrhythmogenic cardiomyopathy, and typical electrocardiographic (ECG) features have been identified for diagnosis. Machine learning is a powerful tool used in ECG analysis and has shown to outperform cardiologists. OBJECTIVES: We aimed to develop machine learning and deep learning models to diagnose PLN p.Arg14del cardiomyopathy using ECGs and evaluate their accuracy compared to an expert cardiologist. METHODS: We included 155 adult PLN mutation carriers and 155 age- and sex-matched control subjects. Twenty-one PLN mutation carriers (13.4%) were classified as symptomatic (symptoms of heart failure or malignant ventricular arrhythmias). The data set was split into training and testing sets using 4-fold cross-validation. Multiple models were developed to discriminate between PLN mutation carriers and control subjects. For comparison, expert cardiologists classified the same data set. The best performing models were validated using an external PLN p.Arg14del mutation carrier data set from Murcia, Spain (n = 50). We applied occlusion maps to visualize the most contributing ECG regions. RESULTS: In terms of specificity, expert cardiologists (0.99) outperformed all models (range 0.53-0.81). In terms of accuracy and sensitivity, experts (0.28 and 0.64) were outperformed by all models (sensitivity range 0.65-0.81). T-wave morphology was most important for classification of PLN p.Arg14del carriers. External validation showed comparable results, with the best model outperforming experts. CONCLUSION: This study shows that machine learning can outperform experienced cardiologists in the diagnosis of PLN p.Arg14del cardiomyopathy and suggests that the shape of the T wave is of added importance to this diagnosis.

A morphology based deep learning model for atrial fibrillation detection using single cycle electrocardiographic samples.

BACKGROUND: Deep learning (DL) has shown promising results in improving atrial fibrillation (AF) detection algorithms. However, these models are often criticized because of their "black box" nature. AIM: To develop a morphology based DL model to discriminate AF from sinus rhythm (SR), and to visualize which parts of the ECG are used by the model to derive to the right classification. METHODS: We pre-processed raw data of 1469 ECGs in AF or SR, of patients with a history AF. Input data was generated by normalizing all single cycles (SC) of one ECG lead to SC-ECG samples by 1) centralizing the R wave or 2) scaling from R-to- R wave. Different DL models were trained by splitting the data in a training, validation and test set. By using a DL based heat mapping technique we visualized those areas of the ECG used by the classifier to come to the correct classification. RESULTS: The DL model with the best performance was a feedforward neural network trained by SC-ECG samples on a R-to-R wave basis of lead II, resulting in an accuracy of 0.96 and F1-score of 0.94. The onset of the QRS complex proved to be the most relevant area for the model to discriminate AF from SR. CONCLUSION: The morphology based DL model developed in this study was able to discriminate AF from SR with a very high accuracy. DL model visualization may help clinicians gain insights into which (unrecognized) ECG features are most sensitive to discriminate AF from SR.