Three-dimensional ultrasound probe pose estimation from single-perspective x-rays for image-guided interventions

Abstract

Registration and fusion of real-time trans-esophageal (TEE) and live fluoroscopy images have potential to provide improved visualization during minimally invasive aortic valve replacement and other cardiac interventions. We present an approach for achieving this registration using 3D TEE pose estimation from single-perspective x-ray images, and assess the performance of both point-based and intensity-based tracking techniques. Simulated and experimental accuracy studies demonstrated tracking errors of under 0.58mm and 0.32°, and phantom trials resulted in a 3D US-to-fluoroscopy RMS registration error of 3.5mm. © 2010 Springer-Verlag.