We present the taxonomy of data augmentation for electrocardiogram (ECG) after reviewing various ECG augmentation methods. On the basis of the taxonomy, we demonstrate the effect of augmentation methods on the ECG classification via extensive experiments. Initially, we examine the performance trend as the magnitude of distortion increases and identify the optimal distortion magnitude. Secondly, we investigate the synergistic combinations of the transformations and identify the pairs of transformations with the greatest positive effect. Finally, based on our experimental findings, we propose an efficient augmentation policy and demonstrate that it outperforms previous augmentation policies.

Peripartum cardiomyopathy (PPCM) is a fatal maternal complication, with left ventricular systolic dysfunction (LVSD; Left ventricular ejection fraction 45% or less) occurring at the end of pregnancy or in the months following delivery. The scarcity of screening tools for PPCM leads to a delayed diagnosis and increases its mortality and morbidity. We aim to evaluate an electrocardiogram (ECG)-deep learning model (DLM) for detecting cardiomyopathy in the peripartum period. For the DLM development and internal performance test for detecting LVSD, we obtained a dataset of 122,733 ECG-echocardiography pairs from 58,530 male and female patients from two community hospitals. For the DLM external validation, this study included 271 ECG-echocardiography pairs (157 unique pregnant and postpartum period women) examined in the Ajou University Medical Center (AUMC) between January 2007 and May 2020. All included cases underwent an ECG within two weeks before or after the day of transthoracic echocardiography, which was performed within a month before delivery, or within five months after delivery. Based on the diagnostic criteria of PPCM, we analyzed the area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) to evaluate the model effectiveness. The ECG-based DLM detected PPCM with an AUROC of 0.877. Moreover, its sensitivity, specificity, PPV, and NPV for the detection of PPCM were 0.877, 0.833, 0.809, 0.352, and 0.975, respectively. An ECG-based DLM non-invasively and effectively detects cardiomyopathies occurring in the peripartum period and could be an ideal screening tool for PPCM.

We developed and validated an artificial intelligence (AI)-enabled smartwatch ECG to detect heart failure-reduced ejection fraction (HFrEF). This was a cohort study involving two hospitals (A and B). We developed the AI in two steps. First, we developed an AI model (ECGT2T) to synthesize ten-lead ECG from the asynchronized 2-lead ECG (Lead I and II). ECGT2T is a deep learning model based on a generative adversarial network, which translates source ECGs to reference ECGs by learning styles of the reference ECGs. For this, we included adult patients aged ≥18 years from hospital A with at least one digitally stored 12-lead ECG. Second, we developed an AI model to detect HFrEF using a 10 s 12-lead ECG. The AI model was based on convolutional neural network. For this, we included adult patients who underwent ECG and echocardiography within 14 days. To validate the AI, we included adult patients from hospital B who underwent two-lead smartwatch ECG and echocardiography on the same day. The AI model generates a 10 s 12-lead ECG from a two-lead smartwatch ECG using ECGT2T and detects HFrEF using the generated 12-lead ECG. We included 137,673 patients with 458,745 ECGs and 38,643 patients with 88,900 ECGs from hospital A for developing the ECGT2T and HFrEF detection models, respectively. The area under the receiver operating characteristic curve of AI for detecting HFrEF using smartwatch ECG was 0.934 (95% confidence interval 0.913-0.955) with 755 patients from hospital B. The sensitivity, specificity, positive predictive value, and negative predictive value of AI were 0.897, 0.860, 0.258, and 0.994, respectively. An AI-enabled smartwatch 2-lead ECG could detect HFrEF with reasonable performance.

In recent years, self-supervised learning methods have shown significant improvement for pre-training with unlabeled data and have proven helpful for electrocardiogram signals. However, most previous pre-training methods for electrocardiogram focused on capturing only global contextual representations. This inhibits the models from learning fruitful representation of electrocardiogram, which results in poor performance on downstream tasks. Additionally, they cannot fine-tune the model with an arbitrary set of electrocardiogram leads unless the models were pre-trained on the same set of leads. In this work, we propose an ECG pre-training method that learns both local and global contextual representations for better generalizability and performance on down-stream tasks. In addition, we propose random lead masking as an ECG-specific augmentation method to make our proposed model robust to an arbitrary set of leads. Experimental results on two downstream tasks, cardiac arrhythmia classification and patient identification, show that our proposed approach outperforms other state-of-the-art methods.

Although renal failure is a major healthcare burden globally and the cornerstone for preventing its irreversible progression is an early diagnosis, an adequate and noninvasive tool to screen renal impairment (RI) reliably and economically does not exist. We developed an interpretable deep learning model (DLM) using electrocardiography (ECG) and validated its performance. This retrospective cohort study included two hospitals. We included 115,361 patients who had at least one ECG taken with an estimated glomerular filtration rate measurement within 30 min of the index ECG. A DLM was developed using 96,549 ECGs of 55,222 patients. The internal validation included 22,949 ECGs of 22,949 patients. Furthermore, we conducted an external validation with 37,190 ECGs of 37,190 patients from another hospital. The endpoint was to detect a moderate to severe RI (estimated glomerular filtration rate < 45 ml/min/1.73m2). The area under the receiver operating characteristic curve (AUC) of a DLM using a 12-lead ECG for detecting RI during the internal and external validation was 0.858 (95% confidence interval 0.851–0.866) and 0.906 (0.900–0.912), respectively. In the initial evaluation of 25,536 individuals without RI patients whose DLM was defined as having a higher risk had a significantly higher chance of developing RI than those in the low-risk group (17.2% vs. 2.4%, p < 0.001). The sensitivity map indicated that the DLM focused on the QRS complex and T-wave for detecting RI. The DLM demonstrated high performance for RI detection and prediction using 12-, 6-, single-lead ECGs.

Although overt hyperthyroidism adversely affects a patient’s prognosis, thyroid function tests (TFTs) are not routinely conducted. Furthermore, vague symptoms of hyperthyroidism often lead to hyperthyroidism being overlooked. An electrocardiogram (ECG) is a commonly used screening test, and the association between thyroid function and ECG is well known. However, it is difficult for clinicians to detect hyperthyroidism through subtle ECG changes. For early detection of hyperthyroidism, we aimed to develop and validate an electrocardiographic biomarker based on a deep learning model (DLM) for detecting hyperthyroidism. This multicentre retrospective cohort study included patients who underwent ECG and TFTs within 24 h. For model development and internal validation, we obtained 174 331 ECGs from 113 194 patients. We extracted 48 648 ECGs from 33 478 patients from another hospital for external validation. Using 500 Hz raw ECG, we developed a DLM with 12-lead, 6-lead (limb leads, precordial leads), and single-lead (lead I) ECGs to detect overt hyperthyroidism. We calculated the model’s performance on the internal and external validation sets using the area under the receiver operating characteristic curve (AUC). The AUC of the DLM using a 12-lead ECG was 0.926 (0.913–0.94) for internal validation and 0.883(0.855–0.911) for external validation. The AUC of DLMs using six and a single-lead were in the range of 0.889–0.906 for internal validation and 0.847–0.882 for external validation. CWe developed a DLM using ECG for non-invasive screening of overt hyperthyroidism. We expect this model to contribute to the early diagnosis of diseases and improve patient prognosis.

Emergency department (ED) overcrowding is a national crisis in which pediatric patients are often prioritized at lower levels. Because the prediction of prognosis for pediatric patients is important but difficult, we developed and validated a deep learning algorithm to predict the need for critical care in pediatric EDs. We conducted a retrospective observation cohort study using data from the Korean National Emergency Department Information System, which collected data in real time from 151 EDs. The study subjects were pediatric patients who visited EDs from 2014 to 2016. The data were divided by date into derivation and test data. The primary end point was critical care, and the secondary endpoint was hospitalization. We used age, sex, chief complaint, symptom onset to arrival time, arrival mode, trauma, and vital signs as predicted variables. The study subjects consisted of 2,937,078 pediatric patients of which 18,253 were critical care and 375,078 were hospitalizations. For critical care, the area under the receiver operating characteristics curve of the deep learning algorithm was 0.908 (95% confidence interval, 0.903-0.910). This result significantly outperformed that of the pediatric early warning score (0.812 [0.803-0.819]), conventional triage and acuity system (0.782 [0.773-0.790]), random forest (0.881 [0.874-0.890]), and logistic regression (0.851 [0.844-0.858]). For hospitalization, the deep-learning algorithm (0.782 [0.780-0.783]) significantly outperformed the other methods. The deep learning algorithm predicted the critical care and hospitalization of pediatric ED patients more accurately than the conventional early warning score, triage tool, and machine learning methods.

Although heart failure with reduced ejection fraction (HFrEF) is a common clinical syndrome and can be modified by the administration of appropriate medical therapy, there is no adequate tool available to perform reliable, economical, early-stage screening. To meet this need, we developed an interpretable artificial intelligence (AI) algorithm for HFrEF screening using electrocardiography (ECG) and validated its performance. This retrospective cohort study included two hospitals. An AI algorithm based on a convolutional neural network was developed using 39,371 ECG results from 17,127 patients. The internal validation included 3,470 ECGs from 2,908 patients. Furthermore, we conducted external validation using 4,362 ECGs from 4,176 patients from another hospital to verify the applicability of the algorithm across different centers. The end-point was to detect HFrEF, defined as an ejection fraction <40%. We also visualized the regions in 12 lead ECG that affected HFrEF detection in the AI algorithm and compared this to the previously documented literature. During the internal and external validation, the areas under the curves of the AI algorithm using a 12 lead ECG for detecting HFrEF were 0.913 (95% confidence interval, 0.902–0.925) and 0.961 (0.951–0.971), respectively, and the areas under the curves of the AI algorithm using a single-lead ECG were 0.874 (0.859–0.890) and 0.929 (0.911–0.946), respectively. The deep learning-based AI algorithm performed HFrEF detection well using not only a 12 lead but also a single-lead ECG. These results suggest that HFrEF can be screened not only using a 12 lead ECG, as is typical of a conventional ECG machine, but also with a single-lead ECG performed by a wearable device employing the AI algorithm, thereby preventing irreversible disease progression and mortality.

Early detection and intervention of atrial fibrillation (AF) is a cornerstone for effective treatment and prevention of mortality. Diverse deep learning models (DLMs) have been developed, but they could not be applied in clinical practice owing to their lack of interpretability. We developed an explainable DLM to detect AF using ECG and validated its performance using diverse formats of ECG. We conducted a retrospective study. The Sejong ECG dataset comprising 128,399 ECGs was used to develop and internally validated the explainable DLM. DLM was developed with two feature modules, which could describe the reason for DLM decisions. DLM was external validated using data from 21,837, 10,605, and 8528 ECGs from PTB-XL, Chapman, and PhysioNet non-restricted datasets, respectively. The predictor variables were digitally stored ECGs, and the endpoints were AFs. During internal and external validation of the DLM, the area under the receiver operating characteristic curves (AUCs) of the DLM using a 12‑lead ECG in detecting AF were 0.997-0.999. The AUCs of the DLM with VAE using a 6‑lead and single‑lead ECG were 0.990-0.999. The AUCs of explainability about features such as rhythm irregularity and absence of P-wave were 0.961-0.993 and 0.983-0.993, respectively. Our DLM successfully detected AF using diverse ECGs and described the reason for this decision. The results indicated that an explainable artificial intelligence methodology could be adopted to the DLM using ECG and enhance the transparency of the DLM for its application in clinical practice.

Although heart failure with preserved ejection fraction (HFpEF) is a rapidly emerging global health problem, an adequate tool to screen it reliably and economically does not exist. We developed an interpretable deep learning model (DLM) using electrocardiography (ECG) and validated its performance. This retrospective cohort study included two hospitals. 34 103 patients who underwent echocardiography and ECG within 1 week and indicated normal left ventricular systolic function were included in this study. A DLM based on an ensemble neural network was developed using 32 671 ECGs of 20 169 patients. The internal validation included 1979 ECGs of 1979 patients. Furthermore, we conducted an external validation with 11 955 ECGs of 11955 patients from another hospital. The endpoint was to detect HFpEF. During the internal and external validation, the area under the receiver operating characteristic curves of a DLM using 12-lead ECG for detecting HFpEF were 0.866 (95% confidence interval 0.850–0.883) and 0.869 (0.860–0.877), respectively. In the 1412 individuals without HFpEF at initial echocardiography, patients whose DLM was defined as having a higher risk had a significantly higher chance of developing HFpEF than those in the low-risk group (33.6% vs. 8.4%, P < 0.001). Sensitivity map showed that the DLM focused on the QRS complex and T-wave. The DLM demonstrated high performance for HFpEF detection using not only a 12-lead ECG but also 6- single lead ECG. These results suggest that HFpEF can be screened using conventional ECG devices and diverse life-type ECG machines employing the DLM, thereby preventing disease progression.