Recent global developments underscore the prominent role big data have in modern medical science. But privacy issues constitute a prevalent problem for collecting and sharing data between ...researchers. However, synthetic data generated to represent real data carrying similar information and distribution may alleviate the privacy issue. In this study, we present generative adversarial networks (GANs) capable of generating realistic synthetic DeepFake 10-s 12-lead electrocardiograms (ECGs). We have developed and compared two methods, named WaveGAN* and Pulse2Pulse. We trained the GANs with 7,233 real normal ECGs to produce 121,977 DeepFake normal ECGs. By verifying the ECGs using a commercial ECG interpretation program (MUSE 12SL, GE Healthcare), we demonstrate that the Pulse2Pulse GAN was superior to the WaveGAN* to produce realistic ECGs. ECG intervals and amplitudes were similar between the DeepFake and real ECGs. Although these synthetic ECGs mimic the dataset used for creation, the ECGs are not linked to any individuals and may thus be used freely. The synthetic dataset will be available as open access for researchers at OSF.io and the DeepFake generator available at the Python Package Index (PyPI) for generating synthetic ECGs. In conclusion, we were able to generate realistic synthetic ECGs using generative adversarial neural networks on normal ECGs from two population studies, thereby addressing the relevant privacy issues in medical datasets.
Abstract
Deep learning-based tools may annotate and interpret medical data more quickly, consistently, and accurately than medical doctors. However, as medical doctors are ultimately responsible for ...clinical decision-making, any deep learning-based prediction should be accompanied by an explanation that a human can understand. We present an approach called electrocardiogram gradient class activation map (ECGradCAM), which is used to generate attention maps and explain the reasoning behind deep learning-based decision-making in ECG analysis. Attention maps may be used in the clinic to aid diagnosis, discover new medical knowledge, and identify novel features and characteristics of medical tests. In this paper, we showcase how ECGradCAM attention maps can unmask how a novel deep learning model measures both amplitudes and intervals in 12-lead electrocardiograms, and we show an example of how attention maps may be used to develop novel ECG features.
Background Results from animal models and observational studies have raised concerns regarding the potential cataractogenic effects of statin treatment. We investigated whether common and rare ...genetic variants in
are associated with cataract risk, to gauge the likely long-term effects of statin treatment on lenticular opacities. Methods and Results We used genotyping data and exome sequencing data of unrelated European individuals in the UK Biobank to test the association between genetically proxied inhibition of
and cataract risk. First, we constructed an
genetic score consisting of 5 common variants weighted by their association with low-density lipoprotein cholesterol. Second, we analyzed exome sequencing data to identify carriers of predicted loss-of-function mutations in
. Common and rare variants in aggregate were then tested for association with cataract and cataract surgery. In an analysis of >402 000 individuals, a 38.7 mg/dL (1 mmol/L) reduction in low-density lipoprotein C by the
genetic score was associated with higher risk for cataract (odds ratio, 1.14 95% CI, 1.00-1.39,
=0.045) and cataract surgery (odds ratio, 1.25 95% CI, 1.06-1.48,
=0.009). Among 169 172 individuals with
sequencing data, we identified 32 participants (0.02%), who carried a rare
predicted loss-of-function variant. Compared with noncarriers, heterozygous carriers of
predicted loss-of-function had a higher risk of developing cataract (odds ratio, 4.54 95% CI, 1.96-10.53,
=0.001) and cataract surgery (odds ratio, 5.27 95% CI, 2.27-12.25
=5.37×10
). In exploratory analyses, we found no significant association between genetically proxied inhibition of
, or circulating low-density lipoprotein cholesterol levels (
>0.05 for all) and cataract risk. Conclusions We found that genetically proxied inhibition of the
gene mimicking long-term statin treatment associated with higher risk of cataract. Clinical trials with longer follow-up are needed to confirm these findings.
Atrial fibrillation (AF) has traditionally been considered an electrical heart disease. However, genetic studies have revealed that the structural architecture of the heart also play a significant ...role. We evaluated the functional and structural consequences of harboring a titin-truncating variant (TTNtv) in AF patients, using cardiac magnetic resonance (CMR). Seventeen early-onset AF cases carrying a TTNtv, were matched 1:1 with non-AF controls and a replication cohort of early-onset AF cases without TTNtv, and underwent CMR. Cardiac volumes and left atrial late gadolinium enhancement (LA LGE), as a fibrosis proxy, were measured by a blinded operator. Results: AF cases with TTNtv had significantly reduced left ventricular ejection fraction (LVEF) compared with controls (57 ± 4 vs 64 ± 5%, P < 0.001). We obtained similar findings in early-onset AF patients without TTNtv compared with controls (61 ± 4 vs 64 ± 5%, P = 0.02). We furthermore found a statistically significant increase in LA LGE when comparing early-onset AF TTNtv cases with controls. Using state-of-the-art CMR, we found that early-onset AF patients, irrespective of TTNtv carrier status, had reduced LVEF, indicating that early-onset AF might not be as benign as previously thought.
We aimed to investigate the pathogenicity of cardiac ion channel variants previously associated with SIDS. We reviewed SIDS-associated variants previously reported in databases and the literature in ...three large population-based cohorts; The ExAC database, the Inter99 study, and the UK Biobank (UKBB). Variants were classified according to the American College of Medical Genetics and Genomics (ACMG) guidelines. Of the 92 SIDS-associated variants, 59 (64%) were present in ExAC, 18 (20%) in Inter99, and 24 (26%) in UKBB. Using the Inter99 cohort, we found no difference in J-point amplitude and QTc-interval between carriers and non-carriers for 14/18 variants. There was no difference in the risk of syncope (P = 0.32), malignant ventricular arrhythmia (P = 0.96), and all-cause mortality (P = 0.59) between carriers and non-carriers. The ACMG guidelines reclassified 75% of all variants as variant-of-uncertain significance, likely benign, and benign. We identified ~2/3 of variants previously associated with SIDS and found no significant associations with electrocardiographic traits, syncope, malignant ventricular arrhythmia, or all-cause mortality. These data indicate that many of these variants are not highly penetrant, monogenic causes of SIDS and underline the importance of frequent reappraisal of genetic variants to avoid future misdiagnosis.
Background
The electrocardiographic interatrial block (IAB) has been associated with atrial fibrillation (AF). We aimed to test whether IAB can improve risk prediction of AF for the individual ...person.
Methods and Results
Digital ECGs of 152 759 primary care patients aged 50 to 90 years were collected from 2001 to 2011. We identified individuals with P‐wave ≥120 ms and the presence of none, 1, 2, or 3 biphasic P‐waves in inferior leads. Data on comorbidity, medication, and outcomes were obtained from nationwide registries. We observed a dose‐response relationship between the number of biphasic P‐waves in inferior leads and the hazard of AF during follow‐up. Discrimination of the 10‐year outcome of AF, measured by time‐dependent area under the curve, was increased by 1.09% (95% confidence interval 0.43–1.74%) for individuals with cardiovascular disease at baseline (CVD) and 1.01% (95% confidence interval 0.40–1.62%) for individuals without CVD, when IAB was added to a conventional risk model for AF. The highest effect of IAB on the absolute risk of AF was observed in individuals aged 60 to 70 years with CVD. In this subgroup, the 10‐year risk of AF was 50% in those with advanced IAB compared with 10% in those with a normal P‐wave. In general, individuals with advanced IAB and no CVD had a higher risk of AF than patients with CVD and no IAB.
Conclusions
IAB improves risk prediction of AF when added to a conventional risk model. Clinicians may consider monitoring patients with IAB more closely for the occurrence of AF, especially for high‐risk subgroups.
De novo variants in the exome occur at a rate of 1 per individual per generation, and because of the low reproductive fitness for de novo variants causing severe disease, the likelihood of finding ...these as standing variations in the general population is low. Therefore, this study sought to evaluate the pathogenicity of de novo variants previously associated with cardiac disease based on a large population-representative exome database.
We performed a literature search for previous publications on de novo variants associated with severe arrhythmias and structural heart diseases and investigated whether these variants were present in the Exome Aggregation Consortium (ExAC) database (n=60 706). We identified monogenic variants in single case reports and smaller studies (≤200 subjects) and variants considered to increase susceptibility of disease in 3 larger trio studies (>1000 subjects). Of the monogenic variants, 11% (23/211) were present in ExAC, whereas 26% (802/3050) variants believed to increase susceptibility of disease were identified in ExAC. Monogenic de novo variants in ExAC had a total allele count of 109 and with ≈844 expected cases in ExAC, these variants would account for 13% of all cases in the studied diseases if truly monogenetic.
We observed numerous de novo variants associated with cardiac disease as standing variation in ExAC, thus these variants are less likely monogenetic causes or major risk contributors for cardiac disease. This highlights the importance of investigating the pathogenicity of de novo variants because they are not as exclusive and pathogenically evident as presumed previously.
Atrial fibrillation (AF) is the most common cardiac arrhythmia, and it is associated with an increased risk of heart failure, stroke, dementia, and death. Recently, titin-truncating variants (TTNtv), ...which are predominantly associated with dilated cardiomyopathy (DCM), were associated with early-onset AF. Furthermore, genome-wide association studies (GWAS) associated AF with other structural genes. In this study, we investigated whether early-onset AF was associated with loss-of-function variants in DCM-associated genes encoding cytoskeletal proteins. Using targeted sequencing, we examined a cohort of 527 Scandinavian individuals with early-onset AF and a control group of individuals free of AF (
= 383). The patients had onset of AF before 50 years of age, normal echocardiogram, and no other cardiovascular disease at onset of AF. We identified six individuals with rare loss-of-function variants in three different genes (dystrophin (
), actin-associated
protein (
), and fukutin (
)), of which two variants were novel. Loss-of-function variants in cytoskeletal genes were significantly associated with early-onset AF when patients were compared with controls (
= 0.044). Using publicly available GWAS data, we performed genetic correlation analyses between AF and 13 other traits, e.g., showing genetic correlation between AF and non-ischemic cardiomyopathy (
= 0.0003). Our data suggest that rare loss-of-function variants in cytoskeletal genes previously associated with DCM may have a role in early-onset AF, perhaps through the development of an atrial cardiomyopathy.
A previous genome-wide association study found three genetic loci, rs9388451, rs10428132, and rs11708996, to increase the risk of Brugada Syndrome (BrS). Since the effect of these loci in the general ...population is unknown, we aimed to investigate the effect on electrocardiogram (ECG) parameters and outcomes in the general population.
A cohort of 6,161 individuals (median age 45 interquartile range (IQR) 40-50 years, 49% males), with available digital ECGs, was genotyped and subsequently followed for a median period of 13 IQR 12.6-13.4 years. Data on outcomes were collected from Danish administrative healthcare registries. Furthermore, ~400,000 persons from UK Biobank were investigated for associations between the three loci and cardiac arrest/ventricular fibrillation (VF).
Homozygote carriers of the C allele in rs6800541 intronic to
had a significantly larger J-point elevation (JPE) compared with wildtype carriers (11 vs. 6 μV,
< 0.001). There was an additive effect of carrying multiple BrS-associated risk alleles with an increased JPE in lead V1. None of the BrS-associated genetic loci predisposed to syncope, atrial fibrillation, or total mortality in the general Danish population. The rs9388451 genetic locus adjacent to the
gene was associated with cardiac arrest/VF in an analysis using the UK Biobank study (odds ratio = 1.13 (95% confidence interval: 1.08-1.18),
= 0.006).
BrS-associated risk alleles increase the JPE in lead V1 in an additive manner, but was not associated with increased mortality or syncope in the general population of Denmark. However, the
risk allele increased the risk of cardiac arrest/VF in the larger population study of UK Biobank indicating an important role of this common genetic locus.