The electrocardiogram (ECG) serves as a valuable diagnostic tool, providing crucial information about life-threatening cardiac conditions such as Atrial Fibrillation and Myocardial Infarction. A prompt and efficient assessment of ECG exams in environments like emergency rooms (ERs) can significantly improve the chances of survival for high-risk patients. In this study, we have developed an artificial intelligence-driven screening system specifically designed to analyze 12-lead ECG images. Our proposed method has been trained on an extensive dataset comprising 99,746 12-lead ECG exams collected from the ambulatory section of a tertiary hospital. The primary objective was to accurately classify the exams into three classes: Normal (N), Atrial Fibrillation (AFib), and Other (O). The evaluation of our method resulted in AUROC scores of 95.3%, 99.1%, and 93.3% for N, AFib, and O, respectively. To further validate our approach, we conducted evaluations using the Chinese Physiological Signal Challenge database. In this evaluation, we achieved AUROC scores of 91.8%, 97.5%, and 70.4% for the classes N, AFib, and O, respectively. Additionally, we assessed our method using 1,074 exams acquired in the ER, and achieved AUROC values of 98.3%, 98.0%, and 97.7% for the classes N, AFib, and O, respectively. Finally, we developed and deployed a system with a trained model within the ER of a tertiary hospital for research purposes. The system automatically retrieves newly captured ECG chart images from the Picture Archiving and Communication System (PACS) within the ER. These images undergo necessary preprocessing steps and serve as input for our proposed classification method. This comprehensive approach has resulted in the establishment of an efficient and versatile end-to-end framework for ECG classification. The results of our study highlight the potential of leveraging artificial intelligence in the screening of ECG exams, offering a promising solution for the rapid assessment and prioritization of patients in the ER.

The resting 12-lead electrocardiogram (ECG) is a widely used diagnostic tool in modern medicine, providing crucial insights into various heart conditions. Recently, the application of Artificial Intelligence (AI) to infer health-related information from the 12-lead ECG has gained significant interest. In this study, we propose a deep-learning approach to predict sex and age from both 12-lead and reduced-lead ECGs (12L, 6L, 4L, 3L, 2L, and 1L), and analyze their implications for predicting patient mortality. We employed a ResNeXt-based architecture and trained our model using the CODE15 dataset. Our best sex prediction model achieved an F1-score of 0.800 ± 0.007, Sensitivity (Se) of 0.807 ± 0.016, Positive Predictive Value (PPV) of 0.793 ± 0.022, and an Area Under the Curve (AUC) of 0.910 ± 0.007 when using the 12-lead ECG configuration. Similarly, our best age estimation model achieved a mean absolute error (MAE) of 8.961 ± 0.180, a Pearson Correlation (ρ) of 0.810 ± 0.004, and a coefficient of determination (R2) of 0.637 ± 0.014 when using the 4-lead ECG configuration. Through the evaluation of different lead-set configurations, we demonstrated that even with a reduced number of leads, our models achieved comparable performance to those ob- tained using the conventional 12-lead ECG setup. Moreover, we found that the mortality risk, assessed by the hazard ratio (HR), increased when our age model predicted an age higher than the actual age by a certain threshold for all lead sets (12L: 2.49, 6L: 2.17, 4L: 2.53, 3L: 2.54, 2L: 2.76, 1L: 2.65). Likewise, when our model misclassified the patient’s actual sex, the mortality risk also increased (12L: 1.36, 6L: 1.26, 4L: 1.38, 3L: 1.33, 2L: 1.20, 1L: 1.12). Additionally, we observed a decrease in mortality risk when our method predicted an age lower than the actual age by a certain threshold (12L: 0.71, 6L: 0.71, 4L: 0.74, 3L: 0.71, 2L: 0.75, 1L: 0.64). Overall, our research shows the efficacy of reduced lead ECGs in predicting age, determining sex, and providing valuable insights into patient mortality.