European heart journal. Digital health最新文献

Aims: Atrial fibrillation (AF) is the most common arrhythmia, increasing stroke risk. Detecting AF is challenging due to its asymptomatic and paroxysmal nature. This study combines photoplethysmography (PPG) with automated techniques to detect AF, assess AF burden, and monitor rhythm changes from AF to sinus rhythm (SR).

Methods and results: Ninety patients with recent-onset (duration <48 h) AF, scheduled for cardioversion, were monitored using a three-channel PPG armband on the upper arm. An ambulatory three-lead electrocardiogram (ECG) served as the gold standard. PPG recordings were segmented into 10-, 20-, 30-, and 60-min detection windows. Automated detection identified SR and AF episodes, rhythm changes, and AF burden. Sensitivities, specificities, positive predictive values (PPVs), and negative predictive values (NPVs) for rhythm detection were calculated, and the intraclass correlation coefficients (ICCs) for PPG-based AF burden were compared to the gold standard. Monitoring time ranged from 1.0 to 8.2 h per patient. Sensitivities, specificities, PPVs, and NPVs for AF detection were 93.9-94.6, 99.5-99.8, 99.4-99.7, and 93.7-95.0%, respectively. The ICC (0.97-0.98) indicated excellent agreement between PPG and the gold standard in estimating AF burden, with differences of -6.3 to -8.3 min (5.5-6.8%). Rhythm changes from AF to SR were detected in all patients (sensitivity 100%), with detection delays of 4.1 ± 1.4, 8.7 ± 2.8, 13.7 ± 3.9, and 27.8 ± 7.1 min depending on the detection window.

Conclusion: Photoplethysmography with automated analysis shows promise in detecting AF, AF burden, and rhythm changes, indicating its potential in AF screening.

Clinical trial registration: NCT04917653.

In healthcare, scarcity of high-quality human-adjudicated labelled data may limit the potential of deep neural networks (DNNs). Foundation models provide an efficient starting point for deep learning that can facilitate effective DNN training with fewer labelled training examples. In this study, we leveraged cardiologist-confirmed labels from a large dataset of 1.6 million electrocardiograms (ECGs) acquired as part of routine clinical care at UCSF between 1986 and 2019 to pre-train a convolutional DNN to predict 68 common ECG diagnoses. To our knowledge, this model is one of the most comprehensive ECG DNN models to date, demonstrating high performance with a median area under the receiver operating curve (AUC) of 0.978, median sensitivity of 0.937, and median specificity of 0.923. We then demonstrate the model's utility as a foundation model by additionally training (fine-tuning) the DNN to detect three novel ECG diagnoses with relatively small datasets: carcinoid syndrome, pericardial constriction, and rheumatic doming of the mitral valve. Fine-tuning training of the foundation model achieved an AUC of 0.772 (95% CI 0.723-0.816) for carcinoid syndrome, 0.883 (0.863-0.906) for pericardial constriction, and 0.826 (95% CI 0.802-0.854) for rheumatic doming, compared to 0.492 (95% CI 0.434-0.558), 0.689 (95% CI 0.656-0.720), and 0.701 (95% CI 0.657-0.745), respectively, for DNNs trained from scratch on the same small datasets. Our results demonstrate that the ECG foundation model learned a flexible representation of ECG waveforms and can improve performance of fine-tuned downstream models, particularly in data-limited settings.

Aims: Pathological left ventricular (LV) remodelling following aortic stenosis (AS) confers high risk for heart failure and significantly decreases survival. This study aims to introduce a new wearable acoustic cardiography (ACG) device measuring electromechanical activation time (EMAT) to identify the regression of cardiac remodelling in AS patients undergoing transcatheter aortic valve replacement (TAVR).

Methods and results: This prospective cohort study consecutively enrolled patients with severe symptomatic AS who underwent successful TAVR. The parameters EMAT and EMAT% (EMAT divided by R-R interval, expressed as a percentage) derived from ACG as well as echocardiography data were collected. Pearson correlation analysis was performed to evaluate the correlation between EMAT% and left ventricular mass index (LVMi). Receiver operating characteristic (ROC) curves were used to assess the diagnostic performance of EMAT% in predicting left ventricular hypertrophy (LVH). A total of 159 patients (mean age 72.0 years) were enrolled in the study. At baseline, 55% of patients demonstrated severe LV remodelling. Scatter plots and Pearson correlation analysis revealed a significant association between EMAT% and LVMi. The ROC curve analysis showed strong diagnostic performance of EMAT% in predicting LVH, with an area under the curve consistently exceeding 80% at baseline and during follow-up. Both EMAT% and echocardiographic parameters indicated that LV remodelling progressively improved between 1 and 6 months after TAVR, with stabilization observed at 12 months.

Conclusion: The EMAT can be considered as an effective tool to assist in the evaluation of LV remodelling after TAVR. Further studies are required to confirm its utility as a valuable non-invasive diagnostic and monitoring tool.

Aims: We explored whether artificial intelligence (AI)-enabled electrocardiographic (ECG) sex discrepancy would predict atrial fibrillation (AF) recurrence after catheter ablation for paroxysmal AF.

Methods and results: The AI-ECG sex prediction model was developed from the MIMIC-IV and externally validated on CODE-15% (AUC 0.89) and UK Biobank (AUC 0.92) cohorts. After validation, we estimated AI-ECG sex from pre-procedural sinus rhythm ECGs among paroxysmal AF patients scheduled for catheter ablation using data from a pooled AF ablation cohort (n = 4385) in South Korea. ECG sex discrepancy was defined as ECG sex probability of more than 50% for the opposite sex. During a median follow-up of 24 months, 1094 recurrences developed [median age 60 (52-67) years; women 29.0%]. ECG sex discrepant patients were older, had more heart failure, and had elevated diastolic filling pressure compared with ECG sex non-discrepant patients. The odds ratio (OR) for left atrial enlargement was significantly higher among ECG sex discrepant women [adjusted OR 1.67, 95% confidence interval (CI) 1.14-2.44, P = 0.008] but not among men (adjusted OR 0.88, 95% CI 0.66-1.17, P = 0.368). The 5-year cumulative event rate of AF recurrence was significantly higher among ECG sex discrepant women (log rank, P = 0.015) but not among men (log rank, P = 0.871). The 5-year risk of AF recurrence was significantly higher among ECG sex discrepant women [hazard ratio (HR) 1.42, 95% CI 1.10-1.83, P = 0.007] but not among men (HR 1.01, 95% CI 0.76-1.34, P = 0.940).

Conclusion: Pre-procedural ECG sex discrepancy has a prognostic value for AF recurrence after catheter ablation for paroxysmal AF in women.

Artificial intelligence (AI) tools hold great promise for the rapid and accurate diagnosis of coronary artery disease (CAD) from intravascular optical coherent tomography (IVOCT) images. Numerous papers have been published describing AI-based models for different diagnostic tasks, yet it remains unclear, which models have potential clinical utility and have been properly validated. This systematic review considered published literature between January 2015 and December 2024 describing AI-based diagnosis of CAD using IVOCT. Our search identified 8600 studies, with 629 included after initial screening and 39 studies included in the final systematic review after quality screening. Our findings indicate that most of the identified models are not currently suitable for clinical use, primarily due to methodological flaws and underlying biases. To address these issues, we provide recommendations to improve model quality and research practices to enhance the development of clinically useful AI products.

Aims: The COVID-19 pandemic has raised patient awareness of their health and highlighted the importance of remote care. Smartphones and wearable devices are now becoming essential for managing cardiovascular disease. However, low digital health readiness among cardiology patients poses a significant challenge to the effective use of these technologies. This study evaluates digital health readiness and learning ability of Japanese cardiology patients using the Digital Health Readiness Questionnaire (DHRQ), while also assessing its reliability and validity.

Methods and results: This multicentre observational study evaluated digital health readiness among patients with cardiovascular risk factors at St. Marianna University Hospital and Kawasaki Municipal Tama Hospital. The DHRQ was employed, and confirmatory factor analysis was conducted to validate the measurement model. A total of 210 questionnaires were distributed, with 208 included in the analysis. Internal consistency, measured by Cronbach's alpha, exceeded 0.7 across all factors. Model fit was evaluated with standardised root mean square residual = 0.038, root mean square error of approximation = 0.071, comparative fit index = 0.962, and Tucker-Lewis index = 0.955. Age, education, and smartphone/smartwatch ownership significantly predicted higher DHRQ scores. Older age correlated with lower scores (P < 0.001), while higher education, smartphone (P < 0.001), and smartwatch ownership (P = 0.006) correlated with higher scores. Gender and income were not significant.

Conclusion: The DHRQ proved to be valid in Japan, with education level significantly affecting scores. Improved digital health readiness is suggested to enhance patients' management of health information and communication with healthcare providers, and is expected to be linked to future healthcare systems.

Aims: Rapid myocardial revascularization in patients with acute myocardial infarction (AMI) is essential to improve clinical outcomes. There is still room for improvement in the timely diagnosis of AMI. This study aimed to develop an artificial intelligence (AI) model using electrocardiograms (ECGs) for detecting AMI needing revascularization.

Methods and results: A total of 723 389 ECGs from 300 627 patients in the derivation cohort at a single centre between 2013 and 2020, including 5872 patients with AMI (1.95%) who underwent revascularization, were used for model training and internal testing. A transformer-based deep learning model, initially trained on about one million unlabelled ECGs through self-supervised learning, was fine-tuned for AMI detection. The model's final performance was evaluated in the internal test and the external validation set. The external validation was conducted at an independent centre between 2002 and 2020 using 261 429 ECGs from 259 454 patients, including 1095 patients with AMI (0.42%). By integrating self-supervised learning to train the AI model, we enhanced the AMI detection performance, as demonstrated by an increase in the area under the receiver operating characteristic curve (AUROC) from 0.910 (95% CI, 0.904-0.915) to 0.968 (95% CI, 0.965-0.971) in the external validation set. For ST-elevation myocardial infarction and non-ST-elevation myocardial infarction detection, the AUROCs were 0.991 (95% CI, 0.989-0.993) and 0.947 (95% CI, 0.942-0.952) in the external validation set, respectively.

Conclusion: This novel ECG-based AI model may be beneficial for the timely identification of patients with AMI needing revascularization.

Aims: Deep learning methods have shown impressive performance in detecting a range of diseases from electrocardiogram (ECG) waveforms, but the breadth of diseases that can be detected with high accuracy remains unknown, and in many cases the changes to the ECG allowing these classifications are also opaque. In this study, we aim to determine the full set of cardiac and non-cardiac conditions detectable from the ECG and to understand which ECG features contribute to the disease classification.

Methods and results: Using large datasets of ECGs and connected electronic health records from two separate medical centres, we independently trained PheWASNet, a multi-task deep learning model, to detect 1243 different disease phenotypes from the raw ECG waveform. We confirmed that the ECG can be used to detect chronic kidney disease (AUC = 0.80), cirrhosis (AUC = 0.80), and sepsis (AUC = 0.84), as well as a range of cardiac diseases, and also found new detectable conditions, including respiratory failure (AUC = 0.86), neutropenia (AUC = 0.83), and menstrual disorders (AUC = 0.84). We found that of the 37 non-cardiac strongly detectable conditions, 35 were detectable by the model output for just four diseases, suggesting that they have similar effects on the ECG. We found that high performance in some conditions including neutropenia, respiratory failure, and sepsis can be explained by linear models based on conventional measurements taken from the ECG.

Conclusion: Our study uncovers a range of diseases detectable in the ECG, including many previously unknown phenotypes, and makes progress towards understanding ECG features that allow this detection.

Aims: Whole blood RNA expression is modulated in response to signals from tissues, including the vessel wall. The primary objective of this study was to explore the ability of whole blood transcriptomes, analysed using artificial intelligence (AI), to predict coronary artery calcifications (CAC).

Methods and results: A total of 196 subjects [men aged 40-70 years and women aged 50-70 years without known cardiovascular disease (CVD)] were non-consecutively enrolled for CAC assessment via chest computed tomography. Whole blood RNA was isolated and sequenced. Different AI models were trained using clinical and transcriptomic variables as distinctive features to identify the presence of CAC (Agatston score >0). Finally, we compared the predictive performance of these models. The prevalence of CAC was 43.9%. The combined AI model, incorporating transcriptome data along with age, sex, body mass index, smoking status, diabetes, and hypercholesterolaemia, achieved an area under the curve (AUC) of 0.92 (95% CI, 0.88-0.95) for predicting the presence of CAC, with a sensitivity of 92%, specificity of 80%, positive predictive value of 81%, negative predictive value of 91%, and an overall accuracy of 86%. The combined AI model demonstrated significantly improved discrimination compared with the transcriptomic model (AUC 0.79; P = 0.009), the clinical variables model (AUC 0.72; P < 0.001), and the CVD risk model (AUC 0.68; P < 0.001).

Conclusion: In this pilot study, an AI model integrating whole blood transcriptome data with clinical risk factors demonstrated the ability to predict CAC, providing incremental value over clinical models. Further studies are needed to achieve more robust validation.