6DoF catheter detection, application to intracardiac echocardiography

Abstract

Hybrid imaging systems, consisting of fluoroscopy and echocardiography, are increasingly selected for intra-operative support of minimally invasive cardiac interventions. Intracardiac echocardiograpy (ICE) is an emerging modality with the promise of removing sedation or general anesthesia associated with transesophageal echocardiography (TEE). We introduce a novel 6 degrees of freedom (DoF) pose estimation approach for catheters (equipped with radiopaque ball markers) in single X-Ray fluoroscopy projection and investigate the method's application to a prototype ICE catheter. Machine learning based catheter detection is implemented in a Bayesian hypothesis fusion framework, followed by refinement of ball marker locations through template matching. Marker correspondence and 3D pose estimation are solved through iterative optimization. The method registers the ICE volume to the C-arm coordinate system. Experiments are performed on synthetic and porcine in-vivo data. Target registration error (TRE), defined in the echo cone, is the basis of our preliminary evaluation. The method reached 8.06 ± 7.2 mm TRE on 703 cases. Potential uses of our hybrid system include structural heart disease interventions and electrophysiologycal mapping or catheter ablation procedures.