Abstract: We introduce Guide2Heart, a unique dataset comprising labelled ultrasound images for echocardiographic guidance. It targets the Parasternal Long-Axis (PLAX) view, a clinically key cardiac imaging plane. The labelled data consists of four categories—Standard PLAX, Almost PLAX, Nearby PLAX, and Away from PLAX. Our dataset facilitates the development of real-time view-proximity-based guidance systems that assist echocardiographers in efficiently acquiring the standard PLAX view. Specifically, Guide2Heart includes 3322 ultrasound images from 67 studies conducted on 23 subjects. To validate the effectiveness of our dataset, we conduct multiple experiments using various CNN models and a ViT network, reporting K-fold cross-validation results. Our experimental findings highlight the potential for automated guidance in acquiring standard echocardiographic views, ultimately improving the efficiency and standardization of echocardiographic examinations.
External IDs:doi:10.1007/978-3-032-06329-8_18
Loading