Continuous Atrial Fibrillation Monitoring From Photoplethysmography: Comparison Between Supervised Deep Learning and Heuristic Signal Processing.

BACKGROUND: Continuous monitoring for atrial fibrillation (AF) using photoplethysmography (PPG) from smartwatches or other wearables is challenging due to periods of poor signal quality during motion or suboptimal wearing. As a result, many consumer wearables sample infrequently and only analyze when the user is at rest, which limits the ability to perform continuous monitoring or to quantify AF. OBJECTIVES: This study aimed to compare 2 methods of continuous monitoring for AF in free-living patients: a well-validated signal processing (SP) heuristic and a convolutional deep neural network (DNN) trained on raw signal. METHODS: We collected 4 weeks of continuous PPG and electrocardiography signals in 204 free-living patients. Both SP and DNN models were developed and validated both on holdout patients and an external validation set. RESULTS: The results show that the SP model demonstrated receiver-operating characteristic area under the curve (AUC) of 0.972 (sensitivity 99.6%, specificity: 94.4%), which was similar to the DNN receiver-operating characteristic AUC of 0.973 (sensitivity 92.2, specificity: 95.5%); however, the DNN classified significantly more data (95% vs 62%), revealing its superior tolerance of tracings prone to motion artifact. Explainability analysis revealed that the DNN automatically suppresses motion artifacts, evaluates irregularity, and learns natural AF interbeat variability. The DNN performed better and analyzed more signal in the external validation cohort using a different population and PPG sensor (AUC, 0.994; 97% analyzed vs AUC, 0.989; 88% analyzed). CONCLUSIONS: DNNs perform at least as well as SP models, classify more data, and thus may be better for continuous PPG monitoring.

Patient-Level Artificial Intelligence-Enhanced Electrocardiography in Hypertrophic Cardiomyopathy: Longitudinal Treatment and Clinical Biomarker Correlations.

BACKGROUND: Artificial intelligence (AI) applied to 12-lead electrocardiographs (ECGs) can detect hypertrophic cardiomyopathy (HCM). OBJECTIVES: The purpose of this study was to determine if AI-enhanced ECG (AI-ECG) can track longitudinal therapeutic response and changes in cardiac structure, function, or hemodynamics in obstructive HCM during mavacamten treatment. METHODS: We applied 2 independently developed AI-ECG algorithms (University of California-San Francisco and Mayo Clinic) to serial ECGs (n = 216) from the phase 2 PIONEER-OLE trial of mavacamten for symptomatic obstructive HCM (n = 13 patients, mean age 57.8 years, 69.2% male). Control ECGs from 2,600 age- and sex-matched individuals without HCM were obtained. AI-ECG output was correlated longitudinally to echocardiographic and laboratory metrics of mavacamten treatment response. RESULTS: In the validation cohorts, both algorithms exhibited similar performance for HCM diagnosis, and exhibited mean HCM score decreases during mavacamten treatment: patient-level score reduction ranged from approximately 0.80 to 0.45 for Mayo and 0.70 to 0.35 for USCF algorithms; 11 of 13 patients demonstrated absolute score reduction from start to end of follow-up for both algorithms. HCM scores were significantly associated with other HCM-relevant parameters, including left ventricular outflow tract gradient at rest, postexercise, and with Valsalva, and NT-proBNP level, independent of age and sex (all P < 0.01). For both algorithms, the strongest longitudinal correlation was between AI-ECG HCM score and left ventricular outflow tract gradient postexercise (slope estimate: University of California-San Francisco 0.70 [95% CI: 0.45-0.96], P < 0.0001; Mayo 0.40 [95% CI: 0.11-0.68], P = 0.007). CONCLUSIONS: AI-ECG analysis longitudinally correlated with changes in echocardiographic and laboratory markers during mavacamten treatment in obstructive HCM. These results provide early evidence for a potential paradigm for monitoring HCM therapeutic response.

Identifying Mitral Valve Prolapse at Risk for Arrhythmias and Fibrosis From Electrocardiograms Using Deep Learning.

BACKGROUND: Mitral valve prolapse (MVP) is a common valvulopathy, with a subset developing sudden cardiac death or cardiac arrest. Complex ventricular ectopy (ComVE) is a marker of arrhythmic risk associated with myocardial fibrosis and increased mortality in MVP. OBJECTIVES: The authors sought to evaluate whether electrocardiogram (ECG)-based machine learning can identify MVP at risk for ComVE, death and/or myocardial fibrosis on cardiac magnetic resonance (CMR) imaging. METHODS: A deep convolutional neural network (CNN) was trained to detect ComVE using 6,916 12-lead ECGs from 569 MVP patients from the University of California-San Francisco between 2012 and 2020. A separate CNN was trained to detect late gadolinium enhancement (LGE) using 1,369 ECGs from 87 MVP patients with contrast CMR. RESULTS: The prevalence of ComVE was 28% (160/569). The area under the receiver operating characteristic curve (AUC) of the CNN to detect ComVE was 0.80 (95% CI: 0.77-0.83) and remained high after excluding patients with moderate-severe mitral regurgitation [0.80 (95% CI: 0.77-0.83)] or bileaflet MVP [0.81 (95% CI: 0.76-0.85)]. AUC to detect all-cause mortality was 0.82 (95% CI: 0.77-0.87). ECG segments relevant to ComVE prediction were related to ventricular depolarization/repolarization (early-mid ST-segment and QRS from V1, V3, and III). LGE in the papillary muscles or basal inferolateral wall was present in 24% patients with available CMR; AUC for detection of LGE was 0.75 (95% CI: 0.68-0.82). CONCLUSIONS: CNN-analyzed 12-lead ECGs can detect MVP at risk for ventricular arrhythmias, death and/or fibrosis and can identify novel ECG correlates of arrhythmic risk. ECG-based CNNs may help select those MVP patients requiring closer follow-up and/or a CMR.

Identification of supraventricular tachycardia mechanisms with surface electrocardiograms using a convolutional neural network.

Background: It remains difficult to definitively distinguish supraventricular tachycardia (SVT) mechanisms using a 12-lead electrocardiogram (ECG) alone. Machine learning may identify visually imperceptible changes on 12-lead ECGs and may improve ability to determine SVT mechanisms. Objective: We sought to develop a convolutional neural network (CNN) that identifies the SVT mechanism according to the gold standard of SVT ablation and to compare CNN performance against experienced electrophysiologists among patients with atrioventricular nodal re-entrant tachycardia (AVNRT), atrioventricular reciprocating tachycardia (AVRT), and atrial tachycardia (AT). Methods: All patients with 12-lead surface ECG during sinus rhythm and SVT and had successful SVT ablation from 2013 to 2020 were included. A CNN was trained using data from 1505 surface ECGs that were split into 1287 training and 218 test ECG datasets. We compared the CNN performance against independent adjudication by 2 experienced cardiac electrophysiologists on the test dataset. Results: Our dataset comprised 1505 ECGs (368 AVNRT, 304 AVRT, 95 AT, and 738 sinus rhythm) from 725 patients. The CNN areas under the receiver-operating characteristic curve for AVNRT, AVRT, and AT were 0.909, 0.867, and 0.817, respectively. When fixing the specificity of the CNN to the electrophysiologist adjudicators' specificity, the CNN identified all SVT classes with higher sensitivity: (1) AVNRT (91.7% vs 65.9%), (2) AVRT (78.4% vs 63.6%), and (3) AT (61.5% vs 50.0%). Conclusion: A CNN can be trained to differentiate SVT mechanisms from surface 12-lead ECGs with high overall performance, achieving similar performance to experienced electrophysiologists at fixed specificities.

CathAI: fully automated coronary angiography interpretation and stenosis estimation.

Coronary angiography is the primary procedure for diagnosis and management decisions in coronary artery disease (CAD), but ad-hoc visual assessment of angiograms has high variability. Here we report a fully automated approach to interpret angiographic coronary artery stenosis from standard coronary angiograms. Using 13,843 angiographic studies from 11,972 adult patients at University of California, San Francisco (UCSF), between April 1, 2008 and December 31, 2019, we train neural networks to accomplish four sequential necessary tasks for automatic coronary artery stenosis localization and estimation. Algorithms are internally validated against criterion-standard labels for each task in hold-out test datasets. Algorithms are then externally validated in real-world angiograms from the University of Ottawa Heart Institute (UOHI) and also retrained using quantitative coronary angiography (QCA) data from the Montreal Heart Institute (MHI) core lab. The CathAI system achieves state-of-the-art performance across all tasks on unselected, real-world angiograms. Positive predictive value, sensitivity and F1 score are all ≥90% to identify projection angle and ≥93% for left/right coronary artery angiogram detection. To predict obstructive CAD stenosis (≥70%), CathAI exhibits an AUC of 0.862 (95% CI: 0.843-0.880). In UOHI external validation, CathAI achieves AUC 0.869 (95% CI: 0.830-0.907) to predict obstructive CAD. In the MHI QCA dataset, CathAI achieves an AUC of 0.775 (95%. CI: 0.594-0.955) after retraining. In conclusion, multiple purpose-built neural networks can function in sequence to accomplish automated analysis of real-world angiograms, which could increase standardization and reproducibility in angiographic coronary stenosis assessment.

Interindividual Variability in Self-Monitoring of Blood Pressure Using Consumer-Purchased Wireless Devices.

BACKGROUND: Engagement with self-monitoring of blood pressure (BP) declines, on average, over time but may vary substantially by individual. OBJECTIVES: We aimed to describe different 1-year patterns (groups) of self-monitoring of BP behaviors, identify predictors of those groups, and examine the association of self-monitoring of BP groups with BP levels over time. METHODS: We analyzed device-recorded BP measurements collected by the Health eHeart Study-an ongoing prospective eCohort study-from participants with a wireless consumer-purchased device that transmitted date- and time-stamped BP data to the study through a full 12 months of observation starting from the first day they used the device. Participants received no instruction on device use. We applied clustering analysis to identify 1-year self-monitoring, of BP patterns. RESULTS: Participants had a mean age of 52 years and were male and White. Using clustering algorithms, we found that a model with three groups fit the data well: persistent daily use (9.1% of participants), persistent weekly use (21.2%), and sporadic use only (69.7%). Persistent daily use was more common among older participants who had higher Week 1 self-monitoring of BP frequency and was associated with lower BP levels than the persistent weekly use or sporadic use groups throughout the year. CONCLUSION: We identified three distinct self-monitoring of BP groups, with nearly 10% sustaining a daily use pattern associated with lower BP levels.

Automated Assessment of Cardiac Systolic Function From Coronary Angiograms With Video-Based Artificial Intelligence Algorithms.

Importance: Understanding left ventricular ejection fraction (LVEF) during coronary angiography can assist in disease management. Objective: To develop an automated approach to predict LVEF from left coronary angiograms. Design, Setting, and Participants: This was a cross-sectional study with external validation using patient data from December 12, 2012, to December 31, 2019, from the University of California, San Francisco (UCSF). Data were randomly split into training, development, and test data sets. External validation data were obtained from the University of Ottawa Heart Institute. Included in the analysis were all patients 18 years or older who received a coronary angiogram and transthoracic echocardiogram (TTE) within 3 months before or 1 month after the angiogram. Exposure: A video-based deep neural network (DNN) called CathEF was used to discriminate (binary) reduced LVEF (≤40%) and to predict (continuous) LVEF percentage from standard angiogram videos of the left coronary artery. Guided class-discriminative gradient class activation mapping (GradCAM) was applied to visualize pixels in angiograms that contributed most to DNN LVEF prediction. Results: A total of 4042 adult angiograms with corresponding TTE LVEF from 3679 UCSF patients were included in the analysis. Mean (SD) patient age was 64.3 (13.3) years, and 2212 patients were male (65%). In the UCSF test data set (n = 813), the video-based DNN discriminated (binary) reduced LVEF (≤40%) with an area under the receiver operating characteristic curve (AUROC) of 0.911 (95% CI, 0.887-0.934); diagnostic odds ratio for reduced LVEF was 22.7 (95% CI, 14.0-37.0). DNN-predicted continuous LVEF had a mean absolute error (MAE) of 8.5% (95% CI, 8.1%-9.0%) compared with TTE LVEF. Although DNN-predicted continuous LVEF differed 5% or less compared with TTE LVEF in 38.0% (309 of 813) of test data set studies, differences greater than 15% were observed in 15.2% (124 of 813). In external validation (n = 776), video-based DNN discriminated (binary) reduced LVEF (≤40%) with an AUROC of 0.906 (95% CI, 0.881-0.931), and DNN-predicted continuous LVEF had an MAE of 7.0% (95% CI, 6.6%-7.4%). Video-based DNN tended to overestimate low LVEFs and underestimate high LVEFs. Video-based DNN performance was consistent across sex, body mass index, low estimated glomerular filtration rate (≤45), presence of acute coronary syndromes, obstructive coronary artery disease, and left ventricular hypertrophy. Conclusion and relevance: This cross-sectional study represents an early demonstration of estimating LVEF from standard angiogram videos of the left coronary artery using video-based DNNs. Further research can improve accuracy and reduce the variability of DNNs to maximize their clinical utility.

Deep learning augmented ECG analysis to identify biomarker-defined myocardial injury.

Chest pain is a common clinical complaint for which myocardial injury is the primary concern and is associated with significant morbidity and mortality. To aid providers' decision-making, we aimed to analyze the electrocardiogram (ECG) using a deep convolutional neural network (CNN) to predict serum troponin I (TnI) from ECGs. We developed a CNN using 64,728 ECGs from 32,479 patients who underwent ECG within 2 h prior to a serum TnI laboratory result at the University of California, San Francisco (UCSF). In our primary analysis, we classified patients into groups of TnI < 0.02 or ≥ 0.02 µg/L using 12-lead ECGs. This was repeated with an alternative threshold of 1.0 µg/L and with single-lead ECG inputs. We also performed multiclass prediction for a set of serum troponin ranges. Finally, we tested the CNN in a cohort of patients selected for coronary angiography, including 3038 ECGs from 672 patients. Cohort patients were 49.0% female, 42.8% white, and 59.3% (19,283) never had a positive TnI value (≥ 0.02 µg/L). CNNs accurately predicted elevated TnI, both at a threshold of 0.02 µg/L (AUC = 0.783, 95% CI 0.780-0.786) and at a threshold of 1.0 µg/L (AUC = 0.802, 0.795-0.809). Models using single-lead ECG data achieved significantly lower accuracy, with AUCs ranging from 0.740 to 0.773 with variation by lead. Accuracy of the multi-class model was lower for intermediate TnI value-ranges. Our models performed similarly on the cohort of patients who underwent coronary angiography. Biomarker-defined myocardial injury can be predicted by CNNs from 12-lead and single-lead ECGs.

Electrocardiogram Detection of Pulmonary Hypertension Using Deep Learning.

BACKGROUND: Pulmonary hypertension (PH) is life-threatening, and often diagnosed late in its course. We aimed to evaluate if a deep learning approach using electrocardiogram (ECG) data alone can detect PH and clinically important subtypes. We asked: does an automated deep learning approach to ECG interpretation detect PH and its clinically important subtypes? METHODS AND RESULTS: Adults with right heart catheterization or an echocardiogram within 90 days of an ECG at the University of California, San Francisco (2012-2019) were retrospectively identified as PH or non-PH. A deep convolutional neural network was trained on patients' 12-lead ECG voltage data. Patients were divided into training, development, and test sets in a ratio of 7:1:2. Overall, 5016 PH and 19,454 patients without PH were used in the study. The mean age at the time of ECG was 62.29 ± 17.58 years and 49.88% were female. The mean interval between ECG and right heart catheterization or echocardiogram was 3.66 and 2.23 days for patients with PH and patients without PH, respectively. In the test dataset, the model achieved an area under the receiver operating characteristic curve, sensitivity, and specificity, respectively of 0.89, 0.79, and 0.84 to detect PH; 0.91, 0.83, and 0.84 to detect precapillary PH; 0.88, 0.81, and 0.81 to detect pulmonary arterial hypertension, and 0.80, 0.73, and 0.76 to detect group 3 PH. We additionally applied the trained model on ECGs from participants in the test dataset that were obtained from up to 2 years before diagnosis of PH; the area under the receiver operating characteristic curve was 0.79 or greater. CONCLUSIONS: A deep learning ECG algorithm can detect PH and PH subtypes around the time of diagnosis and can detect PH using ECGs that were done up to 2 years before right heart catheterization/echocardiogram diagnosis. This approach has the potential to decrease diagnostic delays in PH.

A generalizable electrocardiogram-based artificial intelligence model for 10-year heart failure risk prediction.

Background: Heart failure (HF) is a progressive condition with high global incidence. HF has two main subtypes: HF with preserved ejection fraction (HFpEF) and HF with reduced ejection fraction (HFrEF). There is an inherent need for simple yet effective electrocardiogram (ECG)-based artificial intelligence (AI; ECG-AI) models that can predict HF risk early to allow for risk modification. Objective: The main objectives were to validate HF risk prediction models using Multi-Ethnic Study of Atherosclerosis (MESA) data and assess performance on HFpEF and HFrEF classification. Methods: There were six models in comparision derived using ARIC data. 1) The ECG-AI model predicting HF risk was developed using raw 12-lead ECGs with a convolutional neural network. The clinical models from 2) ARIC (ARIC-HF) and 3) Framingham Heart Study (FHS-HF) used 9 and 8 variables, respectively. 4) Cox proportional hazards (CPH) model developed using the clinical risk factors in ARIC-HF or FHS-HF. 5) CPH model using the outcome of ECG-AI and the clinical risk factors used in CPH model (ECG-AI-Cox) and 6) A Light Gradient Boosting Machine model using 288 ECG Characteristics (ECG-Chars). All the models were validated on MESA. The performances of these models were evaluated using the area under the receiver operating characteristic curve (AUC) and compared using the DeLong test. Results: ECG-AI, ECG-Chars, and ECG-AI-Cox resulted in validation AUCs of 0.77, 0.73, and 0.84, respectively. ARIC-HF and FHS-HF yielded AUCs of 0.76 and 0.74, respectively, and CPH resulted in AUC = 0.78. ECG-AI-Cox outperformed all other models. ECG-AI-Cox provided an AUC of 0.85 for HFrEF and 0.83 for HFpEF. Conclusion: ECG-AI using ECGs provides better-validated predictions when compared to HF risk calculators, and the ECG feature model and also works well with HFpEF and HFrEF classification.

Advancing cardiovascular medicine with machine learning: Progress, potential, and perspective.

Recent advances in machine learning (ML) have made it possible to analyze high-dimensional and complex data-such as free text, images, waveforms, videos, and sound-in an automated manner by successfully learning complex associations within these data. Cardiovascular medicine is particularly well poised to take advantage of these ML advances, due to the widespread digitization of medical data and the large number of diagnostic tests used to evaluate cardiovascular disease. Various ML approaches have successfully been applied to cardiovascular tests and diseases to automate interpretation, accurately perform measurements, and, in some cases, predict novel diagnoses from less invasive tests, effectively expanding the utility of more widely accessible diagnostic tests. Here, we present examples of some impactful advances in cardiovascular medicine using ML across a variety of modalities, with a focus on deep learning applications.

Deep Learning Electrocardiographic Analysis for Detection of Left-Sided Valvular Heart Disease.

BACKGROUND: Valvular heart disease is an important contributor to cardiovascular morbidity and mortality and remains underdiagnosed. Deep learning analysis of electrocardiography (ECG) may be useful in detecting aortic stenosis (AS), aortic regurgitation (AR), and mitral regurgitation (MR). OBJECTIVES: This study aimed to develop ECG deep learning algorithms to identify moderate or severe AS, AR, and MR alone and in combination. METHODS: A total of 77,163 patients undergoing ECG within 1 year before echocardiography from 2005-2021 were identified and split into train (n = 43,165), validation (n = 12,950), and test sets (n = 21,048; 7.8% with any of AS, AR, or MR). Model performance was assessed using area under the receiver-operating characteristic (AU-ROC) and precision-recall curves. Outside validation was conducted on an independent data set. Test accuracy was modeled using different disease prevalence levels to simulate screening efficacy using the deep learning model. RESULTS: The deep learning algorithm model accuracy was as follows: AS (AU-ROC: 0.88), AR (AU-ROC: 0.77), MR (AU-ROC: 0.83), and any of AS, AR, or MR (AU-ROC: 0.84; sensitivity 78%, specificity 73%) with similar accuracy in external validation. In screening program modeling, test characteristics were dependent on underlying prevalence and selected sensitivity levels. At a prevalence of 7.8%, the positive and negative predictive values were 20% and 97.6%, respectively. CONCLUSIONS: Deep learning analysis of the ECG can accurately detect AS, AR, and MR in this multicenter cohort and may serve as the basis for the development of a valvular heart disease screening program.

Using machine learning to uncover heterogeneity of beta blocker response in heart failure.

A recent study by Karwath et al.1 in The Lancet applied machine learning-based cluster analysis to pooled data from nine double-blind, randomized controlled trials of beta blockers, identifying subgroups of efficacy in patients with sinus rhythm and atrial fibrillation.

Predictors of incident SARS-CoV-2 infections in an international prospective cohort study.

OBJECTIVE: Until effective treatments and vaccines are made readily and widely available, preventative behavioural health measures will be central to the SARS-CoV-2 public health response. While current recommendations are grounded in general infectious disease prevention practices, it is still not entirely understood which particular behaviours or exposures meaningfully affect one's own risk of incident SARS-CoV-2 infection. Our objective is to identify individual-level factors associated with one's personal risk of contracting SARS-CoV-2. DESIGN: Prospective cohort study of adult participants from 26 March 2020 to 8 October 2020. SETTING: The COVID-19 Citizen Science Study, an international, community and mobile-based study collecting daily, weekly and monthly surveys in a prospective and time-updated manner. PARTICIPANTS: All adult participants over the age of 18 years were eligible for enrolment. PRIMARY OUTCOME MEASURE: The primary outcome was incident SARS-CoV-2 infection confirmed via PCR or antigen testing. RESULTS: 28 575 unique participants contributed 2 479 149 participant-days of data across 99 different countries. Of these participants without a history of SARS-CoV-2 infection at the time of enrolment, 112 developed an incident infection. Pooled logistic regression models showed that increased age was associated with lower risk (OR 0.98 per year, 95% CI 0.97 to 1.00, p=0.019), whereas increased number of non-household contacts (OR 1.10 per 10 contacts, 95% CI 1.01 to 1.20, p=0.024), attending events of at least 10 people (OR 1.26 per 10 events, 95% CI 1.07 to 1.50, p=0.007) and restaurant visits (OR 1.95 per 10 visits, 95% CI 1.42 to 2.68, p<0.001) were associated with significantly higher risk of incident SARS-CoV-2 infection. CONCLUSIONS: Our study identified three modifiable health behaviours, namely the number of non-household contacts, attending large gatherings and restaurant visits, which may meaningfully influence individual-level risk of contracting SARS-CoV-2.

Predicting Incident Heart Failure in Women With Machine Learning: The Women's Health Initiative Cohort.

BACKGROUND: Heart failure (HF) is a leading cause of cardiac morbidity among women, whose risk factors differ from those in men. We used machine-learning approaches to develop risk- prediction models for incident HF in a cohort of postmenopausal women from the Women's Health Initiative (WHI). METHODS: We used 2 machine-learning methods-Least Absolute Shrinkage and Selection Operator (LASSO) and Classification and Regression Trees (CART)-to perform variable selection on 1227 baseline WHI variables for the primary outcome of incident HF. These variables were then used to construct separate Cox proportional hazard models, and we compared these results, using receiver-operating characteristic (ROC) curve analysis, against a comparator model built using variables from the Atherosclerosis Risk in Communities (ARIC) HF prediction model. We analyzed 43,709 women who had 2222 incident HF events; median follow-up was 14.3 years. RESULTS: LASSO selected 10 predictors, and CART selected 11 predictors. The highest correlation between selected variables was 0.46. In addition to selecting well-established predictors such as age, myocardial infarction, and smoking, novel predictors included physical function, number of pregnancies, number of previous live births and age at menopause. In ROC analysis, the CART-derived model had the highest C-statistic of 0.83 (95% confidence interval [CI], 0.81-0.85), followed by LASSO 0.82 (95% CI, 0.81-0.84) and ARIC 0.73 (95% CI, 0.70-0.76). CONCLUSIONS: Machine-learning approaches can be used to develop HF risk-prediction models that can have better discrimination compared with an established HF risk model and may provide a basis for investigating novel HF predictors.

Performance of a Convolutional Neural Network and Explainability Technique for 12-Lead Electrocardiogram Interpretation.

Importance: Millions of clinicians rely daily on automated preliminary electrocardiogram (ECG) interpretation. Critical comparisons of machine learning-based automated analysis against clinically accepted standards of care are lacking. Objective: To use readily available 12-lead ECG data to train and apply an explainability technique to a convolutional neural network (CNN) that achieves high performance against clinical standards of care. Design, Setting, and Participants: This cross-sectional study was conducted using data from January 1, 2003, to December 31, 2018. Data were obtained in a commonly available 12-lead ECG format from a single-center tertiary care institution. All patients aged 18 years or older who received ECGs at the University of California, San Francisco, were included, yielding a total of 365 009 patients. Data were analyzed from January 1, 2019, to March 2, 2021. Exposures: A CNN was trained to predict the presence of 38 diagnostic classes in 5 categories from 12-lead ECG data. A CNN explainability technique called LIME (Linear Interpretable Model-Agnostic Explanations) was used to visualize ECG segments contributing to CNN diagnoses. Main Outcomes and Measures: Area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were calculated for the CNN in the holdout test data set against cardiologist clinical diagnoses. For a second validation, 3 electrophysiologists provided consensus committee diagnoses against which the CNN, cardiologist clinical diagnosis, and MUSE (GE Healthcare) automated analysis performance was compared using the F1 score; AUC, sensitivity, and specificity were also calculated for the CNN against the consensus committee. Results: A total of 992 748 ECGs from 365 009 adult patients (mean [SD] age, 56.2 [17.6] years; 183 600 women [50.3%]; and 175 277 White patients [48.0%]) were included in the analysis. In 91 440 test data set ECGs, the CNN demonstrated an AUC of at least 0.960 for 32 of 38 classes (84.2%). Against the consensus committee diagnoses, the CNN had higher frequency-weighted mean F1 scores than both cardiologists and MUSE in all 5 categories (CNN frequency-weighted F1 score for rhythm, 0.812; conduction, 0.729; chamber diagnosis, 0.598; infarct, 0.674; and other diagnosis, 0.875). For 32 of 38 classes (84.2%), the CNN had AUCs of at least 0.910 and demonstrated comparable F1 scores and higher sensitivity than cardiologists, except for atrial fibrillation (CNN F1 score, 0.847 vs cardiologist F1 score, 0.881), junctional rhythm (0.526 vs 0.727), premature ventricular complex (0.786 vs 0.800), and Wolff-Parkinson-White (0.800 vs 0.842). Compared with MUSE, the CNN had higher F1 scores for all classes except supraventricular tachycardia (CNN F1 score, 0.696 vs MUSE F1 score, 0.714). The LIME technique highlighted physiologically relevant ECG segments. Conclusions and Relevance: The results of this cross-sectional study suggest that readily available ECG data can be used to train a CNN algorithm to achieve comparable performance to clinical cardiologists and exceed the performance of MUSE automated analysis for most diagnoses, with some exceptions. The LIME explainability technique applied to CNNs highlights physiologically relevant ECG segments that contribute to the CNN's diagnoses.

Predictors of incident viral symptoms ascertained in the era of COVID-19.

BACKGROUND: In the absence of universal testing, effective therapies, or vaccines, identifying risk factors for viral infection, particularly readily modifiable exposures and behaviors, is required to identify effective strategies against viral infection and transmission. METHODS: We conducted a world-wide mobile application-based prospective cohort study available to English speaking adults with a smartphone. We collected self-reported characteristics, exposures, and behaviors, as well as smartphone-based geolocation data. Our main outcome was incident symptoms of viral infection, defined as fevers and chills plus one other symptom previously shown to occur with SARS-CoV-2 infection, determined by daily surveys. FINDINGS: Among 14, 335 participants residing in all 50 US states and 93 different countries followed for a median 21 days (IQR 10-26 days), 424 (3%) developed incident viral symptoms. In pooled multivariable logistic regression models, female biological sex (odds ratio [OR] 1.75, 95% CI 1.39-2.20, p<0.001), anemia (OR 1.45, 95% CI 1.16-1.81, p = 0.001), hypertension (OR 1.35, 95% CI 1.08-1.68, p = 0.007), cigarette smoking in the last 30 days (OR 1.86, 95% CI 1.35-2.55, p<0.001), any viral symptoms among household members 6-12 days prior (OR 2.06, 95% CI 1.67-2.55, p<0.001), and the maximum number of individuals the participant interacted with within 6 feet in the past 6-12 days (OR 1.15, 95% CI 1.06-1.25, p<0.001) were each associated with a higher risk of developing viral symptoms. Conversely, a higher subjective social status (OR 0.87, 95% CI 0.83-0.93, p<0.001), at least weekly exercise (OR 0.57, 95% CI 0.47-0.70, p<0.001), and sanitizing one's phone (OR 0.79, 95% CI 0.63-0.99, p = 0.037) were each associated with a lower risk of developing viral symptoms. INTERPRETATION: While several immutable characteristics were associated with the risk of developing viral symptoms, multiple immediately modifiable exposures and habits that influence risk were also observed, potentially identifying readily accessible strategies to mitigate risk in the COVID-19 era.

Machine learning prediction of blood alcohol concentration: a digital signature of smart-breathalyzer behavior.

Excess alcohol use is an important determinant of death and disability. Machine learning (ML)-driven interventions leveraging smart-breathalyzer data may help reduce these harms. We developed a digital phenotype of long-term smart-breathalyzer behavior to predict individuals' breath alcohol concentration (BrAC) levels trained on data from a smart breathalyzer. We analyzed roughly one million datapoints from 33,452 users of a commercial smart-breathalyzer device, collected between 2013 and 2017. For validation, we analyzed the associations between state-level observed smart-breathalyzer BrAC levels and impaired-driving motor vehicle death rates. Behavioral, geolocation-based, and time-series-derived features were fed to an ML algorithm using training (70% of the cohort), development (10% of the cohort), and test (20% of the cohort) sets to predict the likelihood of a BrAC exceeding the legal driving limit (0.08 g/dL). States with higher average BrAC levels had significantly higher alcohol-related driving death rates, adjusted for the number of users per state B (SE) = 91.38 (15.16), p < 0.01. In the independent test set, the ML algorithm predicted the likelihood of a given user-initiated BrAC sample exceeding BrAC ≥ 0.08 g/dL, with an area under the curve (AUC) of 85%. Highly predictive features included users' prior BrAC trends, subjective estimation of their BrAC (or AUC = 82% without the self-estimate), engagement and self-monitoring, time since the last measure, and hour of the day. In conclusion, an ML algorithm successfully quantified a digital phenotype of behavior, predicting naturalistic BrAC levels exceeding 0.08 g/dL (a threshold associated with alcohol-related harm) with good discrimination capability. This result establishes a foundation for future research on precision behavioral medicine digital health interventions using smart breathalyzers and passive monitoring approaches.