ECG Diagnosis Research Articles

Circadian variation pattern of sudden cardiac arrest occurred in Chinese community

BackgroundThe circadian variation pattern of sudden cardiac arrest (SCA) occurred in Chinese community including both community healthcare centres and primary hospitals remains unknown. This study analysed the circadian variation of...

Wave masked autoencoder: An electrocardiogram signal diagnosis model based on wave making strategy

ECG signal labeling is time-consuming and laborious, significantly constraining the performance of deep learning models on ECG diagnosis. Self-supervised learning methods offer a means to mitigate the reliance on annotated ECG data. Nevertheless, the direct transposition of these methods to ECG diagnostics may encounter challenges stemming from deficient domain knowledge and incongruous model architectures. Therefore, a self-supervised learning model that introduces ECG prior knowledge, called wave masked autoencoder (WMAE), is proposed. Firstly, considering that various waves that make up the ECG signal are key features for diagnosing arrhythmias, we design a wave masking strategy as an auxiliary task for self-supervised learning to introduce ECG-related knowledge. Secondly, sliding window slicing is utilized to divide each ECG signal into a series of subsequences, which are mapped into deep features with a convolutional feature mapper to increase the information density of basic semantic units and learn complete waveform representations. In this way, the proposed WMAE can capture robust and transferable ECG signal features from unlabeled data, thereby alleviating the problem of insufficient annotated ECG data and improving the performance of the model on cardiovascular disease diagnosis. Experimental results show that WMAE is significantly better than other benchmark models and has high practical value.

Two Cases of Acute Myocarditis with Suspected Acute Myocardial Infarction

Case 1: The patient is a 19-year-old male who was admitted to the hospital with a cough for three days and chest pain for a day, aggravated for two hours. Electrocardiogram showed 1. Sinus tachycardia 2. First-degree atrioventricular block 3. ST-T changes, not excluding acute extensive anterior wall myocardial infarction. Case 2: The patient was a 47-year-old woman admitted to the hospital with a fever for three days. Electrocardiogram showed: 1. Sinus tachycardia, 2. Complete block of the right bundle branch, 3. ST-T changes, not excluding acute extensive anterior myocardial infarction. The study of these two cases strengthened the departmental staff’s knowledge of ECG in acute myocarditis and provided a rapid and accurate ECG diagnosis in the clinic.

ROLE OF CORONARY COMPUTED TOMOGRAPHY ANGIOGRAPHY (CCTA) IN SPONTANEOUS CORONARY ARTERY DISSECTION (SCAD)

Abstract Spontaneous Coronary Artery Dissection (SCAD) is a rare but insidious cause of myocardial infarction and sudden cardiac death especially in those who lacks of traditional cardiovascular risk factors. Although the most appropriate management of patients with SCAD is not clearly defined, it‘s essential to make an early diagnosis in order to follow a conservative approach. Nowadays, the role of Coronary Computed Tomography Angiography (CCTA) is consolidated in the study of atherosclerotic coronary pathology but its contribution in the diagnosis and follow–up of patients with SCAD does not apperar to be defined. The purpose of this study was to retrospectively investigate the diagnostic performance of CCTA and its role in the follow–up of patients with diagnosis of SCAD. We examined 307 patients admitted to emergency department for acute chest pain and an ECG diagnosis of STEMI from March 2021 to July 2023. 1 patient was diagnosed with type 2 SCAD on invasive coronary angiography (ICA) and 2 patients received a suspected diagnosis of SCAD at ICA and CCTA examination was performed to confirm the suspicion. CCTA examination showed the acute presence of Intramural hematoma (IMH) of the wall of left anterior descending artery (LAD) in 2 patients, confirming the angiographic suspect of SCAD type 2. All 3 patients were treated with conservatie therapy and subsequentely underwent CCTA of follow–up. The examination performed respectively 3 months, 45 days and 1 month after the acute event demonstrated the complete resolution of IMH in all patients and the patency of the vessel lumens. CCTA seems to be a feasible and effective imaging modality to detect the main findings of SCAD. It may play a crucial role in the management of patients with acute chest pain at low–intermediate risk of ACS in order to reduce the number of misdiagnoses. Furthermore, CCTA could facilitate follow–up of patients with diagnosis of SCAD given the high recurrency rate of this pathology. However, further large prospective studies are needed to clarify the role of CCTA in the diagnosis and in the follow–up of SCAD.

Automatic multi-label diagnosis of single-lead ECG using novel hybrid residual recurrent convolutional neural networks

Arrhythmia is a prevalent cardiovascular disease that can unveil various heart health issues. In recent times, the rise of wearable devices has garnered attention towards the advantages of portable single-lead ECG devices in heart health monitoring. Achieving end-to-end ECG automatic diagnosis through deep learning is essential due to the time-consuming and labour-intensive characteristics of manual diagnosis of ECG records. This paper presents the design of a hybrid residual recurrent convolutional neural network (HRRCNN) for automatically diagnosing single-lead ECG with multiple labels. Firstly, the convolutional layers and residual networks extract the spatial and short-term temporal features. Subsequently, a hybrid recurrent neural network (RNN) comprised of a double-layer bidirectional gated recurrent unit (GRU) and a bidirectional long short-term memory (LSTM) is employed to capture long and short-term temporal features. Finally, an output layer, comprising an attention layer and a fully connected layer, converts the extracted features into classification results. To evaluate the performance of HRRCNN, a dataset from the Shanghai First People’s Hospital Affiliated to Shanghai JiaoTong University(Shanghai General Hospital, SGH) is utilized. The experimental results demonstrate that HRRCNN achieves an accuracy of 95.08 %, F1-macro of 87.22 %, Sensitivity of 88.51 %, and a positive prediction rate of 86.13 %. Comparative analysis with existing methods underscores the superior performance of HRRCNN, thereby establishing its potential as an out-of-hospital heart health monitoring tool and an auxiliary tool in clinical settings.

Fast CNN-Based Electrocardiogram Signal Quality Assessment Using Fourier Magnitude Spectrum for Resource-Constrained ECG Diagnosis Devices

Fast CNN-Based Electrocardiogram Signal Quality Assessment Using Fourier Magnitude Spectrum for Resource-Constrained ECG Diagnosis Devices

Tele-ECG improves diagnosis of acute coronary syndrome and ST-elevation myocardial infarction in Côte d'Ivoire.

The emergence of cardiovascular risk factors in sub-Saharan Africa suggests an increasing incidence of acute coronary syndromes and STEMI. The aim of the study was to define the prevalence of STEMI and to describe the characteristics of patients diagnosed with STEMI within the tele-electrocardiogram (ECG) network in Côte d'Ivoire. A retrospective study was conducted from January 2015 to August 2019. All adult patients managed by one of the six hospitals within the telemedicine network who benefited from a remote interpretation of their ECG by the cardiology department of Bouaké University Hospital were included. The main reason for ECG interpretation, patient and ECG characteristics, diagnosis, response time and treatment were described. A total of 5649 patients were included. The prevalence of STEMI was 0.7% (n = 44 cases) with a mean age of 58.6 ± 11.8 years and a M/F sex ratio of 1.93. Among STEMI patients, chest pain was the main reason for ECG testing (56.8%). Most ECGs were interpreted within 12 hours (72.8%). The anterior inter-ventricular artery location (59.1%, n = 26) was predominant. The Q wave of necrosis was absent in 18% (n = 8) of cases. All patients received double anti-platelet aggregation and 50% (n = 22) additional heparin therapy. No patient underwent primary angioplasty or thrombolysis, 65.9% (n = 29) were referred to the Bouaké Cardiology Department and 34.1% (n = 15) to the Abidjan Heart Institute. Scheduled angioplasty was performed in 20% (n = 3) of patients in Abidjan. Tele-ECG was an effective means of STEMI screening in Côte d'Ivoire. Systematic telethrombolysis of all patients diagnosed could improve their prognosis.

ECG classification efficient modeling with artificial bee colony optimization data augmentation and attention mechanism.

In addressing the key issues of the data imbalance within ECG signals and modeling optimization, we employed the TimeGAN network and a local attention mechanism based on the artificial bee colony optimization algorithm to enhance the performance and accuracy of ECG modeling. Initially, the TimeGAN network was introduced to rectify data imbalance and create a balanced dataset. Furthermore, the artificial bee colony algorithm autonomously searched hyperparameter configurations by minimizing Wasserstein distance. Control experiments revealed that data augmentation significantly boosted classification accuracy to 99.51%, effectively addressing challenges with unbalanced datasets. Moreover, to overcome bottlenecks in the existing network, the introduction of the Efficient network was adopted to enhance the performance of modeling optimized with attention mechanisms. Experimental results demonstrated that this integrated approach achieved an impressive overall accuracy of 99.70% and an average positive prediction rate of 99.44%, successfully addressing challenges in ECG signal identification, classification, and diagnosis.

Single center retrospective clinical audit and comparison of outcome between epicardial and transvenous endocardial permanent pacemaker implantations in dogs.

The aim of this retrospective cohort study was to provide a single-center clinical audit of complications for single chamber permanent pacemaker implantation (PPI) techniques and determine if the clinical parameters, PPI technique or complications were associated with outcome. The electronic medical records were searched for dogs treated for bradyarrhythmia with PPI. Data related to presenting complaint, signalment of the dog, ECG diagnosis, echocardiographic findings, PPI technique, and programing of the pacemaker were recorded. Survival length (days) was recorded as the last veterinary visit; if the dog was dead the reason was documented. Cumulative survival of each pacemaker was examined by a Kaplan-Meier survival curve and the two techniques compared with a logrank test. Chi-square was used to determine the association between major complications and death. A total of 66 dogs with 52 transvenous and 30 epicardial PPIs were included. All epicardial pacemakers were implanted via transdiaphragmatic approach. A total of 31 life-threatening complications were reported. There were nine deaths related to major complications (13.6% of the study sample). The median follow-up period was 366 days, with a median survival of 255 days, and a significant difference in cumulative survival of each pacemaker (P = 0.01) between epicardial (93 days, range 0-1882 days) and transvenous (334 days, range 0-2745) PPIs but no significant difference in cumulative survival between the two techniques when only the first pacemaker was considered (p = 0.07). The presence of a major complications had a significant association with death due to pacemaker complications (P<0.001). The decision to perform epicardial PPI in failed transvenous PPI patients may have skewed the cumulative survival as was evident in the lack of significant difference in survival when only first PPI were examined. Major complication rates between the two techniques were similar and the authors consider both techniques equally reliable to manage symptomatic bradycardia in dogs.

ISE/ISHNE expert consensus statement on the ECG diagnosis of left ventricular hypertrophy: The change of the paradigm.

The ECG diagnosis of LVH is predominantly based on the QRS voltage criteria. The classical paradigm postulates that the increased left ventricular mass generates a stronger electrical field, increasing the leftward and posterior QRS forces, reflected in the augmented QRS amplitude. However, the low sensitivity of voltage criteria has been repeatedly documented. We discuss possible reasons for this shortcoming and proposal of a new paradigm. The theoretical background for voltage measured at the body surface is defined by the solid angle theorem, which relates the measured voltage to spatial and non-spatial determinants. The spatial determinants are represented by the extent of the activation front and the distance of the recording electrodes. The non-spatial determinants comprise electrical characteristics of the myocardium, which are comparatively neglected in the interpretation of the QRS patterns. Various clinical conditions are associated with LVH. These conditions produce considerable diversity of electrical properties alterations thereby modifying the resultant QRS patterns. The spectrum of QRS patterns observed in LVH patients is quite broad, including also left axis deviation, left anterior fascicular block, incomplete and complete left bundle branch blocks, Q waves, and fragmented QRS. Importantly, the QRS complex can be within normal limits. The new paradigm stresses the electrophysiological background in interpreting QRS changes, i.e., the effect of the non-spatial determinants. This postulates that the role of ECG is not to estimate LV size in LVH, but to understand and decode the underlying electrical processes, which are crucial in relation to cardiovascular risk assessment.

Evaluation of a smartphone electrocardiograph in healthy foals and comparison to standard base-apex electrocardiography

Smartphone-based technology for ECG recording has recently spread as a complementary tool for electrocardiographic screening and monitoring in adult horses and in other animal species. The present study aimed to assess the feasibility and accuracy of a smartphone-based ECG in healthy foals. This was a prospective observational study (authorization n. 45,865/2016) including 22 foals aged less than 21 days. A reference standard base-apex ECG (rECG) was acquired, and a smartphone ECG (sECG) was recorded immediately after by using a smartphone-based single lead electrocardiograph. All ECG tracings were evaluated in a blind fashion by a single board-certified cardiologist, who judged whether the tracings were acceptable for interpretation and performed ECG measurements and diagnosis. The Spearman correlation coefficient, the Cohen’s k test and the Bland-Altman test were used to assess the agreement between sECG and rECG. All sECG tracings were acceptable for interpretation. All foals showed sinus rhythm on both rRCG and sECG tracings, with perfect agreement in heart rate classification (κ = 0.87; p < 0.001). No clinically relevant differences were found in the assessment of waves and intervals duration. Concerning P wave and QRS complex polarity, the percentage of agreement between rECG and sECG was 78% and 83%, respectively. About ECG tracing quality, rECG and sECG showed a substantial agreement (κ = 0.624; p < 0.001). In conclusion, the smartphone-based ECG device tested in the present study recorded good quality single-lead ECG tracings in foals, reliable for heart rate and ECG measurements, but different polarity of P waves and QRS complexes was found in some foals in comparison to rECG.

A filter design method based on evolutionary learning for R-peak detection in portable ECG testing devices

Portable ECG testing devices can develop individuals’ understanding of abnormal heart conditions in a timely manner and are of great value in disease prevention. R-peak detection is the first step in achieving fast and accurate ECG diagnosis. Currently, most R-peak (and QRS complex) detection algorithms target medical-grade multilead ECG measurement devices and cannot be applied to portable single-lead ECG measurement bracelets in low computing power and strong noise environments. Therefore, a filter design method for ECG R-peak detection was proposed in this paper to improve ECG quality with limited hardware resources. Besides, evolutionary learning was adopted in the filter design of the R-peak detection algorithm to reveal the optimal filter through the good global search ability of the genetic algorithm and the morphological features of the ECG. Finally, the precision, recall, and F1 score on the self-built ECG dataset reached 97.03%, 98.49%, and 97.67%, respectively, and the superiority of the method in some performance metrics compared with other excellent filters was verified.

Changing the diagnostic paradigm of the ECG diagnosis of left ventricular hypertrophy

Changing the diagnostic paradigm of the ECG diagnosis of left ventricular hypertrophy

Use of Oral Anticoagulants in Patients with Atrial Fibrillation: Preliminary Data from the Italian Atrial Fibrillation (ITALY-AF) Registry.

Atrial fibrillation (AFIB), the most frequent cardiac arrhythmia, is a major risk factor for stroke, heart failure, and death. Because of the recent advances in AFIB management and the availability of new oral anticoagulants (OACs), there is a need for a systematic and predefined collection of contemporary data regarding its management and treatment. The objective of the ongoing ITALY-AFIB registry is to evaluate the long-term morbidity and mortality in patients with AFIB and to verify the implementation of the current guidelines for stroke prevention in these patients. The registry includes consecutive in- and out-patients with first diagnosed, paroxysmal, persistent, or permanent AFIB. In patients in sinus rhythm at entry, the qualifying episode of AFIB, confirmed by ECG diagnosis, had to have occurred within 1 year before entry. The clinical record form is web-based and accessible by personal keyword. Enrolment into the registry started in the year 2013. In a current cohort of 2470 patients (mean age 75 ± 11 years, males 56%), the mean CHA2DS2-VASc score was 3.7 ± 1.8, and the mean HAS-BLED was 1.6 ± 0.9. There were no significant sex differences in the AFIB subtypes. At the end of the inclusion visit and after receiving knowledge of the web-based electronic estimate of risk for stroke and bleeding, the proportion of patients discharged with OACs was 80%. After exclusion of patients with first diagnosed AFIB (n = 397), the proportion of patients with prescription of OACs rose from 66% before the visit to 82% on discharge (p < 0.0001). Prescription of aspirin or other antiplatelet drugs fell from 18% before the visit to 10% on discharge (p < 0.0001). A web-based management of AFIB with automated estimation of risk profiles appears to favorably affect adherence to AFIB guidelines, based on a high proportion of patients treated with OACs and a substantial decline in the use of antiplatelet drugs.

A resource-efficient ECG diagnosis model for mobile health devices

Mobile health devices with automatic electrocardiogram diagnosis models facilitate long-term cardiac monitoring and enhance the sensitivity of detecting paroxysmal cardiovascular disease. However, these devices often encounter limitations in terms of computational power and battery life. Moreover, iterations of mobile health device versions may result in changes to their computational limitations, leading to additional costs for model design and training. Therefore, it is crucial to develop lightweight models that enable efficient inference under resource constraints and are compatible with various devices. To address these challenges, we propose a resource-efficient and device-adaptive electrocardiogram diagnosis model based on a multi-classifier dynamic neural network. This model achieves efficient computation by exiting simple samples early and can adapt to the computational limits of various devices by adjusting predefined thresholds. Furthermore, we propose two novel approaches, namely rethinking structure and feature-based knowledge distillation, which enhance the collaboration among classifiers and improve the overall performance. The proposed model is evaluated using three electrocardiogram datasets: PTB-XL (AUC = 0.9706), ICBEB2018 (AUC = 0.964), and MIT-BIH arrhythmia (ACC = 0.994, F1 = 0.961). The results consistently demonstrate the superiority of the proposed model over the comparative models across all three datasets, with average improvements of 2.4%, 0.5%, and 1.3% respectively.

Risk Stratification in Pediatric Wolff-Parkinson-White: Practice Variation Among Pediatric Cardiologists and Electrophysiologists.

Published guidelines provide recommendations for risk stratification in pediatric Wolff-Parkinson-White (WPW). There are no data on provider concordance with these guidelines. We hypothesized that significant practice variation exists between pediatric cardiologists (PC) and electrophysiologists (EP). The records of all patients, age 8 to 21 years, with a new ECG diagnosis of WPW between 1/1/2013 and 12/31/2018, from a single center, were retrospectively reviewed. Subjects were categorized on the basis of symptoms and resting ECG findings as one of the following: asymptomatic intermittent WPW, asymptomatic persistent WPW, or symptomatic WPW. The performance and results of diagnostic testing, including Holter monitor, event monitor, exercise stress test (EST), and electrophysiology study (EPS), were recorded. The primary outcome was concordance with published guidelines. A secondary outcome was documentation of a discussion of sudden cardiac death (SCD) risk. 615 patient encounters were analyzed in 231 patients with newly diagnosed WPW pattern on ECG (56% male; mean age at diagnosis 13.9 ± 2.5 years). EP were observed to have a significantly higher rate of guideline concordance than PC (95% vs. 71%, p < 0.001). There was significant practice variation between PC and EP in the documentation of a discussion of SCD risk: 96% in EP vs. 39% in PC (p < 0.001). Significant practice variation exists in the non-invasive and invasive risk stratification of pediatric WPW, with lower concordance to published guidelines amongst PC, when compared to EP. This report highlights the need to promote awareness of current WPW guidelines in the pediatric cardiology community at large.

A case report: anteroseptal ST elevation due to acute isolated right ventricular infarction

BackgroundElectrocardiogram (ECG) is the first diagnostic tool physicians use in diagnosing acute myocardial infarction (MI). In this case report, we present a case where the initial ECG diagnosis was that of an acute anteroseptal MI but emergency coronary angiography showed that the infarct-related artery was a small non-dominant right coronary artery (RCA) instead of the anticipated left anterior descending artery (LAD). Isolated right ventricular (RV) infarction from a non-dominant RCA is rarely seen in clinical practice, and it may exhibit ECG changes that can be confused with an acute anteroseptal MI. It is important to appreciate the subtle differences in the ECG changes that occur in either of these two types of MI for appropriate diagnosis and treatment.Case presentationA 49-year-old non-smoking male with prior coronary stent implantation in LAD presented with acute chest pain and his pre-hospital ECG indicated an anteroseptal STEMI possibly due to stent thrombosis, but an emergency angiogram showed patent LAD and Circumflex arteries. There was however thrombotic occlusion of the right, non-dominant coronary artery, which was revascularized with a drug-eluting stent. The patient’s chest pain and ST elevations resolved, and subsequent echo showed moderate RV systolic dysfunction in keeping with RV myocardial infarction.DiscussionRV myocardial infarction is usually due to an occlusion of the dominant RCA proximal to the origin of its RV wall branch, which often results in inferior ST elevation with reciprocal anterior ST depression. The ST elevation over V1 which would accompany RV infarction is often masked due to the more dominant electrical forces of inferior and posterior LV wall infarction. Our case demonstrates that in isolated RV infarction due to non-dominant proximal RCA occlusion, anterior ST elevation can be seen over V1-3, being most prominent in V1, which overlies the right ventricle, and resolved after restoring flow to the RCA. Spatial vector analysis of the ECG or right-sided ECG leads would be helpful to aid the diagnosis of RV infarction when clinical suspicion is present, for example when there is significant hypotension, raised jugular venous pressure but clear lung fields or deterioration after nitrate administration.

MEDICAL STUDENTS’ EVALUATION BY SERIOUS GAME IN THE ERA OF COVID-19 INFECTION

Objective: The aim of our study was to compare the results of medical students’ evaluation by virtual simulation using online SG and clinical case-based multiple choice questions (MCQ), and to assess the degree of satisfaction with these two evaluation methods. Design and method: Medical students from the same level of study voluntarily participated in this study. SG group had an evaluation by SG dealing with diagnosis and management of ST-segment elevation myocardial infarction (STEMI), already prepared on the following website: www.Mediactiv.com. Clinical case-based MCQ group was evaluated by clinical case-based MCQ having the same topic as SG group. Results of the two groups were compared. A satisfaction questionnaire was filled out by the two groups. The satisfaction degree was compared between the two groups. Results: In this study, 64 medical students (G1:31 and G2: 33) were enrolled. Obtaining complete notes in the questioning and clinical examination parts was significantly more important in SG group compared to clinical case-based MCQ group. No significant difference was noted between the two groups in terms of ECG interpretation and diagnosis of STEMI parts. A greater number of students in SG group obtained a full mark compared to clinical case-based MCQ group in the initial treatment part (54.8% vs. 39.4%; p = 0.04). A total of thirty learners (96.8%) in SG group obtained a total score more than 50% versus 69.7% in clinical case-based MCQ group (p = 0.004). The full score of 100% was obtained by three learners in SG group; however, no student scored 100% in clinical case-based MCQ group (p = 0.027). Medical evaluation using SG was reported to be more innovative, fun, and realistic compared to evaluation by clinical case-based MCQ. No significant difference was noted between the two methods in terms of stress. Conclusions: Simulation by SG could be an innovative and effective method in evaluating medical students, with a higher degree of satisfaction.

СARDIAC ARRHYTHMIAS ANALYSIS BASED ON THE PHASE PORTRAIT OF ECG SIGNAL

The article illustrates the development of virtual system in LabVIEW software environment for processing and analysis of electrocardiographic signals. In the system, the issues of processing ECG signal with classical methods and in phase space have been considered, the comparison of these methods have been carried out, characteristic changes of the shapes of ECG phase portraits corresponding to the change of diagnostic signs haveb been analyzed, and additional diagnostic signs determined on the the base of phase portrait in order to increase the reliability of ECG diagnosis have been interpreted. As a result of the analysis of the sequence of RR intervals using the developed program, a histogram, a scatterogram, a fast Fourier transform, an autoregressive spectrum, a short-term Fourier transform and the corresponding informative parameters characterizing heart rate variability were obtained: the average length and standard deviation of RR intervals, the mean value and standard deviation heart rate, spectral parameters, etc. The results obtained during the processing and analysis of ECG files taken from the MIT BIH databases using the HRV analysis program and the determination of informative parameters implemented in the LabVIEW environment confirm the adequacy of the developed program. Keywords: ECG signal, phase space, phase portrait, RR intervals, LabVIEW software, heart rate, QRS complexes.

PO-01-214 DEEP LEARNING EVALUATION OF ECHOCARDIOGRAMS TO IDENTIFY OCCULT ATRIAL FIBRILLATION

Atrial fibrillation (AF) can often be missed by routine screening given its frequently paroxysmal and asymptomatic presentation. Deep learning algorithms can identify patients with paroxysmal AF from sinus electrocardiograms. Transthoracic echocardiograms (TTEs) may provide additional structural information complementary to ECGs that could also be used to help identify occult AF. We sought to determine whether deep learning evaluation of echocardiograms of patients in normal sinus rhythm could predict concurrent atrial fibrillation. We identified patients who had TTE performed between 6/2004 and 6/2021. Cases were defined as patients who had a TTE in normal sinus rhythm (NSR) with AF documented on ECG within 90 days before/after the TTE. Controls had a NSR TTE (by report and confirmed by lack of noted mitral A wave velocity) and no documented AF within 90 days by ECG or ICD diagnosis. We trained a video-based convolutional neural network using TTE parasternal long axis (PLAX) videos to predict concurrent AF. Model performance was compared to AF prediction by logistic regression using clinical variables (age, sex, heart failure, hypertension, cerebrovascular disease, peripheral arterial disease, diabetes, height, weight, smoker), CHADS2VASc score, and left atrial (LA) size on TTE. Model was trained on 57,681 TTEs of patients who were on average 66.2 years old (SD 16.9), 44.9% female, with a mean CHADS2VASc score of 3.1 (SD 2.1) and LA area of 19.7cm2 (SD 5.8). When tested on a held-out sample of 7,283 TTEs, the model predicted the occurrence of atrial fibrillation within 90 days of a sinus TTE with an AUC of 0.71 (0.69-0.73). Using the max accuracy threshold, the PPV was 0.20 and the NPV was 0.95. The model performed better than predicting AF using clinical risk factors (AUC 0.67), LA area (AUC 0.63), and CHADS2VASc (AUC 0.61). A deep learning model predicted the occurrence of atrial fibrillation from TTEs moderately well, better than clinical variables or LA size alone. This may open up additional opportunities for screening patients for occult AF.