Publications by authors named "Alvaro E Ulloa Cerna"
Background: Several large trials have employed age or clinical features to select patients for atrial fibrillation (AF) screening to reduce strokes. We hypothesized that a machine learning (ML) model trained to predict AF risk from 12‑lead electrocardiogram (ECG) would be more efficient than criteria based on clinical variables in indicating a population for AF screening to potentially prevent AF-related stroke.
Methods: We retrospectively included all patients with clinical encounters in Geisinger without a prior history of AF.
Article Synopsis
- The study focuses on improving the diagnosis of structural heart diseases through a new ECG-based machine learning model that predicts various conditions, potentially increasing patient outcomes.
- By analyzing 2.2 million ECGs linked to health records, researchers tested their model on seven echocardiography-confirmed diseases, ultimately achieving a high predictive accuracy (0.91 AUC) with a 42% positive predictive value.
- The composite model outperformed individual disease predictions and showed consistent results across different datasets, emphasizing the effectiveness of incorporating diverse patient data for better heart disease detection.
Use of machine learning (ML) for automated annotation of heart structures from echocardiographic videos is an active research area, but understanding of comparative, generalizable performance among models is lacking. This study aimed to (1) assess the generalizability of five state-of-the-art ML-based echocardiography segmentation models within a large Geisinger clinical dataset, and (2) test the hypothesis that a quality control (QC) method based on segmentation uncertainty can further improve segmentation results. Five models were applied to 47,431 echocardiography studies that were independent from any training samples.
View Article and Find Full Text PDFArticle Synopsis
- Atrial fibrillation (AF) can lead to serious health issues if not detected early, and the study aims to use a deep neural network to predict new-onset AF from resting 12-lead ECGs in patients without a previous AF history.
- Researchers analyzed 1.6 million ECG traces from 430,000 patients, achieving good predictive performance with an area under the receiver operating characteristic curve of 0.85, indicating an effective ability to identify those at risk for AF within a year.
- The model also demonstrated that it could indicate a high risk for AF-related strokes, successfully identifying 62% of patients who experienced such strokes within three years, thereby highlighting the potential for targeted screening strategies.
Machine learning promises to assist physicians with predictions of mortality and of other future clinical events by learning complex patterns from historical data, such as longitudinal electronic health records. Here we show that a convolutional neural network trained on raw pixel data in 812,278 echocardiographic videos from 34,362 individuals provides superior predictions of one-year all-cause mortality. The model's predictions outperformed the widely used pooled cohort equations, the Seattle Heart Failure score (measured in an independent dataset of 2,404 patients with heart failure who underwent 3,384 echocardiograms), and a machine learning model involving 58 human-derived variables from echocardiograms and 100 clinical variables derived from electronic health records.
View Article and Find Full Text PDFArticle Synopsis
- The study investigates using a deep neural network (DNN) to predict 1-year all-cause mortality from electrocardiogram (ECG) voltage-time data collected over 34 years.
- The DNN was trained on over 1.1 million ECGs from nearly 253,400 patients, achieving a high accuracy (AUC of 0.88) in predicting mortality, even among patients whose ECGs were deemed 'normal' by doctors.
- Results indicate that the DNN can uncover significant prognostic insights that may not be apparent to cardiologists, with a notable hazard ratio of 9.5 for predicting 1-year mortality.
Article Synopsis
- Heart failure is a common and expensive condition, prompting the need for better management strategies that utilize population health data and machine learning.
- By analyzing electronic health record data from Geisinger, the study aimed to predict 1-year mortality for patients with heart failure by examining various clinical metrics and care gaps.
- The study found that a machine learning model (XGBoost) effectively identified 2,844 patients at high risk of death, suggesting that addressing identified care gaps could potentially save 231 lives within a year.