Deep Learning for Early Detection of Cardiovascular Diseases From Medical Imaging.

Background: Cardiovascular diseases (CVDs) remain a leading cause of mortality worldwide, making early detection vital for reducing morbidity and death rates. Echocardiography is a widely used, noninvasive imaging tool for diagnosing CVDs, but manual interpretation can be time-consuming and subject to variability. This study aims to evaluate the performance of a deep learning model using echocardiogram videos for the early detection of CVDs.

Methods: We applied a convolutional neural network (CNN), based on the ResNet-50 architecture, to the EchoNet-Dynamic data set, which includes echocardiogram videos. Preprocessing involved resizing frames and applying augmentation techniques to enhance model robustness. The data set was split into training (80%) and testing (20%) subsets. The model was trained to classify patients based on the presence or absence of CVD using temporal video features.

Results: The CNN model achieved strong performance metrics, with an overall accuracy of 92.3%, a precision of 91.5%, a recall of 92.7%, and an F1-score of 92.1%. The area under the receiver operating characteristic curve (AUC-ROC) was 0.95, indicating excellent discriminatory ability. These results highlight the model's capability to detect CVDs accurately from dynamic echocardiographic imaging.

Conclusion: This study demonstrates the potential of deep learning, particularly CNN-based models, for automating the early detection of CVDs using echocardiogram videos. The high performance of the model suggests it could contribute to faster, more accurate, and cost-effective diagnosis in clinical practice. Future research should focus on improving model generalizability across diverse populations and enhancing interpretability for integration into clinical workflows.