A combined surface and volumetric approach for registration of patient specific models into left atrial cardiac ablation procedures

Accurate visualization and targeting of specific anatomic locations are key components for successful treatment in catheter based cardiac ablation therapy. In order to augment visualization tools such as bi-plane fluoroscopy and real-time ultrasound, recent work has focussed on registering detailed, patient-specific models into the procedure. Surface based techniques are a popular approach, however, challenges include sensitivity to both the number and as well as locations of selected surface points, thereby making registration accuracy dependent on operator experience as well as patient anatomy. In this work, we propose a new technique for registering preoperative models into patient space by combining a surface based method with a volumetric based approach. The volumetric approach utilizes points from the continuously tracked catheter which is sampled at a rate of five times per second. The technique is tested using data from a canine experiment with five ground truth fiducial markers. The combined registration technique has smaller mean errors for all fiducial markers as compared to using a surface based technique alone. The combined surface and volumetric approach is also more robust to surface point selection, thereby decreasing the sensitivity of point selection on the registration result.