25 July 2022 in European Heart Journal - Digital Health
Aims: Deep neural networks (DNNs) perform excellently in interpreting electrocardiograms (ECGs), both for conventional ECG interpretation and for novel applications such as detection of reduced ejection fraction (EF). Despite these promising developments, implementation is hampered by the lack of trustworthy techniques to explain the algorithms to clinicians. Especially, currently employed heatmap-based methods have shown to be inaccurate. Methods and results: We present a novel pipeline consisting of a variational auto-encoder (VAE) to learn the underlying factors of variation of the median beat ECG morphology (the FactorECG), which are subsequently used in common and interpretable prediction models. As the ECG factors can be made explainable by generating and visualizing ECGs on both the model and individual level, the pipeline provides improved explainability over heatmap-based methods. By training on a database with 1.1 million ECGs, the VAE can compress the ECG into 21 generative ECG factors, most of which are associated with physiologically valid underlying processes. Performance of the explainable pipeline was similar to ‘black box’ DNNs in conventional ECG interpretation [area under the receiver operating curve (AUROC) 0.94 vs. 0.96], detection of reduced EF (AUROC 0.90 vs. 0.91), and prediction of 1-year mortality (AUROC 0.76 vs. 0.75). Contrary to the ‘black box’ DNNs, our pipeline provided explainability on which morphological ECG changes were important for prediction. Results were confirmed in a population-based external validation dataset. Conclusions: Future studies on DNNs for ECGs should employ pipelines that are explainable to facilitate clinical implementation by gaining confidence in artificial intelligence and making it possible to identify biased models.
26 February 2021 in npj Digital Medicine
PHPeter G. M. van der Heijden
Peter G. M. van der Heijden
ARAnneline S. J. M. Te Riele
Anneline S. J. M. Te Riele
+2
Standard reference terminology of diagnoses and risk factors is crucial for billing, epidemiological studies, and inter/intranational comparisons of diseases. The International Classification of Disease (ICD) is a standardized and widely used method, but the manual classification is an enormously time-consuming endeavor. Natural language processing together with machine learning allows automated structuring of diagnoses using ICD-10 codes, but the limited performance of machine learning models, the necessity of gigantic datasets, and poor reliability of terminal parts of these codes restricted clinical usability. We aimed to create a high performing pipeline for automated classification of reliable ICD-10 codes in the free medical text in cardiology. We focussed on frequently used and well-defined three- and four-digit ICD-10 codes that still have enough granularity to be clinically relevant such as atrial fibrillation (I48), acute myocardial infarction (I21), or dilated cardiomyopathy (I42.0). Our pipeline uses a deep neural network known as a Bidirectional Gated Recurrent Unit Neural Network and was trained and tested with 5548 discharge letters and validated in 5089 discharge and procedural letters. As in clinical practice discharge letters may be labeled with more than one code, we assessed the single- and multilabel performance of main diagnoses and cardiovascular risk factors. We investigated using both the entire body of text and only the summary paragraph, supplemented by age and sex. Given the privacy-sensitive information included in discharge letters, we added a de-identification step. The performance was high, with F1 scores of 0.76–0.99 for three-character and 0.87–0.98 for four-character ICD-10 codes, and was best when using complete discharge letters. Adding variables age/sex did not affect results. For model interpretability, word coefficients were provided and qualitative assessment of classification was manually performed. Because of its high performance, this pipeline can be useful to decrease the administrative burden of classifying discharge diagnoses and may serve as a scaffold for reimbursement and research applications.
11 November 2020 in ESC Heart Failure
Aims Dilated cardiomyopathy (DCM) is a complex disease where genetics interplay with extrinsic factors. This study aims to compare the phenotype, management, and outcome of familial DCM (FDCM) and non‐familial (sporadic) DCM (SDCM) across Europe. Methods and results Patients with DCM that were enrolled in the prospective ESC EORP Cardiomyopathy & Myocarditis Registry were included. Baseline characteristics, genetic testing, genetic yield, and outcome were analysed comparing FDCM and SDCM; 1260 adult patients were studied (238 FDCM, 707 SDCM, and 315 not disclosed). Patients with FDCM were younger (P < 0.01), had less severe disease phenotype at presentation (P < 0.02), more favourable baseline cardiovascular risk profiles (P ≤ 0.007), and less medication use (P ≤ 0.042). Outcome at 1 year was similar and predicted by NYHA class (HR 0.45; 95% CI [0.25–0.81]) and LVEF per % decrease (HR 1.05; 95% CI [1.02–1.08]. Throughout Europe, patients with FDCM received more genetic testing (47% vs. 8%, P < 0.01) and had higher genetic yield (55% vs. 22%, P < 0.01). Conclusions We observed that FDCM and SDCM have significant differences at baseline but similar short‐term prognosis. Whether modification of associated cardiovascular risk factors provide opportunities for treatment remains to be investigated. Our results also show a prevalent role of genetics in FDCM and a non‐marginal yield in SDCM although genetic testing is largely neglected in SDCM. Limited genetic testing and heterogeneity in panels provides a scaffold for improvement of guideline adherence.
28 June 2022 in EP Europace
While electrocardiogram (ECG) characteristics have been associated with life-threatening ventricular arrhythmias (LTVA) in dilated cardiomyopathy (DCM), they typically rely on human-derived parameters. Deep neural networks (DNNs) can discover complex ECG patterns, but the interpretation is hampered by their ‘black-box’ characteristics. We aimed to detect DCM patients at risk of LTVA using an inherently explainable DNN. In this two-phase study, we first developed a variational autoencoder DNN on more than 1 million 12-lead median beat ECGs, compressing the ECG into 21 different factors (F): FactorECG. Next, we used two cohorts with a combined total of 695 DCM patients and entered these factors in a Cox regression for the composite LTVA outcome, which was defined as sudden cardiac arrest, spontaneous sustained ventricular tachycardia, or implantable cardioverter-defibrillator treated ventricular arrhythmia. Most patients were male (n = 442, 64%) with a median age of 54 years [interquartile range (IQR) 44–62], and median left ventricular ejection fraction of 30% (IQR 23–39). A total of 115 patients (16.5%) reached the study outcome. Factors F8 (prolonged PR-interval and P-wave duration, P < 0.005), F15 (reduced P-wave height, P = 0.04), F25 (increased right bundle branch delay, P = 0.02), F27 (P-wave axis P < 0.005), and F32 (reduced QRS-T voltages P = 0.03) were significantly associated with LTVA. Inherently explainable DNNs can detect patients at risk of LTVA which is mainly driven by P-wave abnormalities.
30 September 2022 in Circulation: Genomic and Precision Medicine
Background: This study aimed to describe the current practice and results of genetic evaluation in Dutch children with dilated cardiomyopathy and to evaluate genotype-phenotype correlations that may guide prognosis. Methods: We performed a multicenter observational study in children diagnosed with dilated cardiomyopathy, from 2010 to 2017. Results: One hundred forty-four children were included. Initial diagnostic categories were idiopathic dilated cardiomyopathy in 67 children (47%), myocarditis in 23 (16%), neuromuscular in 7 (5%), familial in 18 (13%), inborn error of metabolism in 4 (3%), malformation syndrome in 2 (1%), and “other” in 23 (16%). Median follow-up time was 2.1 years [IQR 1.0–4.3]. Hundred-seven patients (74%) underwent genetic testing. We found a likely pathogenic or pathogenic variant in 38 children (36%), most often in MYH7 (n = 8). In 1 patient initially diagnosed with myocarditis, a pathogenic LMNA variant was found. During the study, 39 patients (27%) reached study endpoint (SE: all-cause death or heart transplantation). Patients with a likely pathogenic or pathogenic variant were more likely to reach SE compared with those without (hazard ratio 2.8; 95% CI 1.3–5.8, P = 0.007), while transplant-free survival was significantly lower ( P = 0.006). Clinical characteristics at diagnosis did not differ between the 2 groups. Conclusions: Genetic testing is a valuable tool for predicting prognosis in children with dilated cardiomyopathy, with carriers of a likely pathogenic or pathogenic variant having a worse prognosis overall. Genetic testing should be incorporated in clinical work-up of all children with dilated cardiomyopathy regardless of presumed disease pathogenesis.