Biosignal data captured by patient monitoring systems could provide key evidence for detecting or predicting critical clinical events; however, noise in these data hinders their use. Because deep learning algorithms can extract features without human annotation, this study hypothesized that they could be used to screen unacceptable electrocardiograms (ECGs) that include noise. To test that, a deep learning-based model for unacceptable ECG screening was developed, and its screening results were compared with the interpretations of a medical expert. To develop and apply the screening model, we used a biosignal database comprising 165,142,920 ECG II (10-second lead II electrocardiogram) data gathered between August 31, 2016 and September 30, 2018 from a trauma intensive-care unit. Then, 2,700 and 300 ECGs (ratio of 9:1) were reviewed by a medical expert and used for 9-fold cross-validation (training and validation) and test datasets. A convolutional neural network-based model for unacceptable ECG screening was developed based on the training and validation datasets. The model exhibiting the lowest cross-validation loss was subsequently selected as the final model. Its performance was evaluated through comparison with a test dataset. When the screening results of the proposed model were compared to the test dataset, the area under the receiver operating characteristic curve and the F1-score of the model were 0.93 and 0.80 (sensitivity = 0.88, specificity = 0.89, positive predictive value = 0.74, and negative predictive value = 0.96). The deep learning-based model developed in this study is capable of detecting and screening unacceptable ECGs efficiently.

Most existing electrocardiogram (ECG) feature extraction methods rely on rule-based approaches. It is difficult to manually define all ECG features. We propose an unsupervised feature learning method using a convolutional variational autoencoder (CVAE) that can extract ECG features with unlabeled data. We used 596,000 ECG samples from 1,278 patients archived in biosignal databases from intensive care units to train the CVAE. Three external datasets were used for feature validation using two approaches. First, we explored the features without an additional training process. Clustering, latent space exploration, and anomaly detection were conducted. We confirmed that CVAE features reflected the various types of ECG rhythms. Second, we applied CVAE features to new tasks as input data and CVAE weights to weight initialization for different models for transfer learning for the classification of 12 types of arrhythmias. The f1-score for arrhythmia classification with extreme gradient boosting was 0.86 using CVAE features only. The f1-score of the model in which weights were initialized with the CVAE encoder was 5% better than that obtained with random initialization. Unsupervised feature learning with CVAE can extract the characteristics of various types of ECGs and can be an alternative to the feature extraction method for ECGs.

Although concern regarding the increased risk for new-onset diabetes mellitus (NODM) after statin treatment has been raised, there has been a lack of evidence in real-world clinical practice, particularly in East Asians. We investigated whether statin use is associated with risk for NODM in Koreans. We conducted a retrospective cohort study using the clinical research database from electronic health records. The study cohort consisted of 8265 statin-exposed and 33,060 matched nonexposed patients between January 1996 and August 2013. Matching at a 1:4 ratio was performed using a propensity score based on age, gender, baseline glucose levels (mg/dL), and hypertension. The comparative risks for NODM with various statins (atorvastatin, fluvastatin, pitavastatin, pravastatin, rosuvastatin, and simvastatin) were estimated by both statin exposure versus matched nonexposed and within-class comparisons. The incidence of NODM among the statin-exposed group (6.000 per 1000 patient-years [PY]) was higher than that of the nonexposed group (3.244 per 1000 PY). The hazard ratio (HR) of NODM after statin exposure was 1.872 (95% confidence interval [CI], 1.432-2.445). Male gender (HR, 1.944; 95% CI, 1.497-2.523), baseline glucose per mg/dL (HR, 1.014; 95% CI, 1.013-1.016), hypertension (HR, 2.232; 95% CI, 1.515-3.288), and thiazide use (HR, 1.337; 95% CI, 1.081-1.655) showed an increased risk for NODM, while angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker showed a decreased risk (HR, 0.774; 95% CI, 0.668-0.897). Atorvastatin-exposed patients showed a higher risk for NODM than their matched nonexposed counterparts (HR, 1.939; 95% CI, 1.278-2.943). However, the risk for NODM was not significantly different among statins in within-class comparisons. In conclusion, an increased risk for NODM was observed among statin users in a practical healthcare setting in Korea.

Background: When using a smartwatch to obtain electrocardiogram (ECG) signals from multiple leads, the device has to be placed on different parts of the body sequentially. The ECG signals measured from different leads are asynchronous. Artificial intelligence (AI) models for asynchronous ECG signals have barely been explored. Objective: We aimed to develop an AI model for detecting acute myocardial infarction using asynchronous ECGs and compare its performance with that of the automatic ECG interpretations provided by a commercial ECG analysis software. We sought to evaluate the feasibility of implementing multiple lead–based AI-enabled ECG algorithms on smartwatches. Moreover, we aimed to determine the optimal number of leads for sufficient diagnostic power. Methods: We extracted ECGs recorded within 24 hours from each visit to the emergency room of Ajou University Medical Center between June 1994 and January 2018 from patients aged 20 years or older. The ECGs were labeled on the basis of whether a diagnostic code corresponding to acute myocardial infarction was entered. We derived asynchronous ECG lead sets from standard 12-lead ECG reports and simulated a situation similar to the sequential recording of ECG leads via smartwatches. We constructed an AI model based on residual networks and self-attention mechanisms by randomly masking each lead channel during the training phase and then testing the model using various targeting lead sets with the remaining lead channels masked. Results: The performance of lead sets with 3 or more leads compared favorably with that of the automatic ECG interpretations provided by a commercial ECG analysis software, with 8.1%-13.9% gain in sensitivity when the specificity was matched. Our results indicate that multiple lead-based AI-enabled ECG algorithms can be implemented on smartwatches. Model performance generally increased as the number of leads increased (12-lead sets: area under the receiver operating characteristic curve [AUROC] 0.880; 4-lead sets: AUROC 0.858, SD 0.008; 3-lead sets: AUROC 0.845, SD 0.011; 2-lead sets: AUROC 0.813, SD 0.018; single-lead sets: AUROC 0.768, SD 0.001). Considering the short amount of time needed to measure additional leads, measuring at least 3 leads—ideally more than 4 leads—is necessary for minimizing the risk of failing to detect acute myocardial infarction occurring in a certain spatial location or direction. Conclusions: By developing an AI model for detecting acute myocardial infarction with asynchronous ECG lead sets, we demonstrated the feasibility of multiple lead-based AI-enabled ECG algorithms on smartwatches for automated diagnosis of cardiac disorders. We also demonstrated the necessity of measuring at least 3 leads for accurate detection. Our results can be used as reference for the development of other AI models using sequentially measured asynchronous ECG leads via smartwatches for detecting various cardiac disorders.

Elevated visit-to-visit blood pressure variability (BPV), independent of mean BP, has been associated with cardiovascular events. However, its impact after ST-elevation myocardial infarction (STEMI) has not been established. This study aimed to investigate the prognostic impact of BPV on patients after STEMI. We analyzed the data and clinical outcomes of STEMI survivors who underwent successful primary coronary intervention from 2003 to 2009. BP was measured at discharge and at 1, 3, 6, 12, 24, and 36 months, and we calculated BPV as the intra-individual standard deviations (SDs) of systolic BP (SBP) across these measurements. We classified the patients as high and low-BPV group, and evaluated the outcomes: occurrence of major adverse cardiovascular events (MACEs), death, recurrent myocardial infarction, and target vessel revascularization within 60 months. We enrolled 343 patients, and mean follow-up duration was 68 ± 34 months (median: 76 months). Mean and median SBP SDs were 13.2 and 12.3 mmHg, and patients were divided into one of the two groups based on the median (high-BPV group = SD ≥ 12.3 mmHg; low-BPV group = SD < 12.3 mmHg). The MACE-free survival in the high-BPV group was significantly worse than that in low-BPV group (log-rank p = 0.035). For the high-BPV group, the risk of a MACE significantly increased by 57% (95% confidence interval: 1.03–2.39; p = 0.038). Visit-to-visit systolic BPV was associated with increased rates of adverse clinical outcomes in patients after STEMI. Careful assessment of BP and attempts to reduce BPV might be also important in STEMI survivors.

(1) Background: Currently, limited data are available regarding the relationship between epicardial fat and plaque composition. The aim of this study was to assess the relationship between visceral fat surrounding the heart and the lipid core burden in patients with coronary artery diseases; (2) Methods: Overall, 331 patients undergoing coronary angiography with combined near-infrared spectroscopy and intravascular ultrasound imaging were evaluated for epicardial adipose tissue (EAT) thickness using transthoracic echocardiography. Patients were divided into thick EAT and thin EAT groups according to the median value; (3) Results: There was a positive correlation between EAT thickness and maxLCBI_4mm, and maxLCBI_4mm was significantly higher in the thick EAT group compared to the thin EAT group (437 vs. 293, p < 0.001). EAT thickness was an independent predictor of maxLCBI_4mm ≥ 400 along with age, low-density lipoprotein-cholesterol level, acute coronary syndrome presentation, and plaque burden in a multiple linear regression model. Receiver operating characteristic curve analysis showed that EAT thickness was a predictor for maxLCBI_4mm ≥ 400; (4) Conclusions: In the present study, EAT thickness is related to the lipid core burden assessed by NIRS-IVUS in patients with CAD which suggests that EAT may affect the stability of the plaques in coronary arteries.