Background: T2-weighted cardiac MRI, performed peri-operatively during an electrophysiology procedure, can, within the acute-injury phase after RFA delivery, distinguish scar tissue from edema [1]. Although ablated lesions are observed on MR images, the visualized lesion positions may not be well aligned with the pre-operative electrophysiological (EP) roadmap due to tissue deformation. We propose to register the lesion targets onto the EP roadmap for more accurate ablation with an MRI-compatible active catheter tracking system. Objectives: To map the ablated lesion target accurately onto the EP roadmap based on rapid intra-operative MR images. Methods: An intensity-based 3D image registration approach, Demons [2], co-registered (a) intra-operative low-resolution MR images with (b) pre-operative high-resolution MR images that construct the EP roadmap. Both can have different dimension and scanning directions obtained by a 3T-MRI scanner (Fig.A1). The computation was accelerated using graphic processing units (GPUs). A left atrial (LA) model constructed of soft silicone was used, of which the LA morphology can be varied (Fig.A2). Gadolinium contrast agents could be injected into pre-defined regions inside the LA wall (Fig. B2), simulating perioperative physiological changes at ablation landmarks ( 3 ) were used to navigate to the ablation landmarks and validate the registration (Fig.B3). Results: The GPU-based co-registration was conducted within 2sec. It re-aligned the ablation target onto the EP roadmap with an average distance error of Conclusions: The proposed platform tracked EP catheters, and co-registered cardiac images in real-time, allowing for improved targeting of lesions. Ref: [1] Ranjan, Circ., 2012; [2] Vercauteren, NeuroImage , 2009; [3] Schmidt, Circ. , 2008.