Artificial intelligence applications in early detection of valvular heart disease using echocardiography

Early detection of valvular heart disease (VHD) by echocardiography plays a significant role in reducing mortality and improving clinical outcomes. In this study, conducted in Uzbekistan, the applications of artificial intelligence (AI) in the early detection of VHD by echocardiographic image interpretation were studied. Data for the study included 1500 echocardiographic images of 500 patients (280 females and 220 males, 58.4 ± 12.7 years old on average), 35% of whom had varying degrees of established valvular abnormalities. A deep learning model with ResNet-50 convolutional neural network architecture was developed and trained on 80% of the data. The results showed that the proposed model had the ability to detect valvular abnormalities, including aortic and mitral valve stenosis and regurgitation, with 94.2% accuracy (92.5% sensitivity and 95.8% specificity). The system also reduced image analysis time by 70% (from 8.5 to 2.6 minutes per patient) compared to cardiologists. The findings suggest that AI integration into the echocardiography interpretation process can be utilized as an accurate and rapid adjunct in screening populations at high risk, especially in areas with limited specialist access.