A fast catheter segmentation and tracking from echocardiographic sequences based on corresponding X-ray fluoroscopic image segmentation and hierarchical GRAPH modelling

Abstract

3D soft tissue information, which X-ray images cannot provide but 3D echocardiographic imaging can, may be required in cardiac catheter-based interventions. In this paper, we propose a real-time catheter tracking strategy in echocardiographic sequences based on catheter segmentation in 2D X-ray images and registration between these two modalities. Firstly the segmentation from X-ray and the registration between X-ray and ultrasound is computed. The results from these steps reduce the search space in the ultrasound volume to a limited space surrounding a curved surface. This space is straightened and 2D SURF is calculated on the sampled cross-sections. All features are organized as a two level hierarchical graph. The longest path on the top-level graph and shortest paths on the bottom level graphs are solved. This combined path is considered as the potential catheter after B-Spline modelling and growing. The experiments on clinical data (2000 pairs of frames) show a better performance than a previous method and some dominant vesselness filters.