Detecting cardiac abnormalities using 12-Lead ECG and deep learning

The standard 12-lead electrocardiogram (ECG) is widely used by cardiologists in diagnosing cardiac abnormalities. However, manual interpretation of ECG signals can be time consuming and dependent on the skills of the clinicians. In this work, an approach for detection of cardiac abnormalities using automatic analysis of 12-lead ECG is presented and validated on a comprehensive dataset with eight cardiac abnormalities and one normal sinus rhythm. The proposed approach uses the raw ECG signals as a direct input to a model comprised of convolutional neural network and bi-directional long short-term memory. The dataset includes subjects with multiple cardiac conditions to account for which a binary classification strategy is utilized, in particular, the one-against-all classification method. A weighted cross entropy loss function is used to compensate for the imbalance in the class sizes. While the ECG signals in the dataset are up to 60 seconds long, the proposed approach utilizes only the first 15 seconds of the signals as it was seen to produce comparable performance with lower computational costs. An average validation accuracy of 96.19%, F-score of 0.8026, and AUC of 0.9624 is achieved using the proposed method.