Multimodality image registration for cardiac interventions

Although minimally invasive intracardiac surgery (MIICS) performed inside beating heart is less invasive and has the potential to significantly reduce patient's trauma and complications in comparison with conventional open heart surgery, one major limitation is the loss of direct vision of the target by the surgeon. Since different imaging modalities can provide complementary information about cardiac anatomy and function, this thesis focuses on the integration of intra-operative ultrasound (US) with pre-operative high quality dynamic 3D magnetic resonance (MR) or computed tomography (CT) images to assist surgical navigation in MIICS.;To facilitate the registration of pre-operative images to the patient, a unique 3D US-based technique is proposed to register a CT-derived ribcage model to the patient to provide an initial transformation for intra-operative cardiac image registration. We show that the target registration error (TRE) between registered US and CT images is less than 1.5 mm in a phantom study.;To accelerate intra-operative registration for intracardiac beating heart surgery, a two-step approach is proposed to dynamically register intra-operative real-time 3D US with pre-operative dynamic 3D MR cardiac images. A preprocessing method is also presented to ensure that the correct registration occurs at the MI maximum of the overlaid images. Validation using human MR+US volumes demonstrates a TRE of 1.76 mm.;2D ultrasound is the standard of care for monitoring cardiac interventions, but is difficult to interpret without being observed within the appropriate context. We demonstrate the means to register and fuse intra-operative 2D US and pre-operative dynamic 3D CT images for use during MCC Spatial registration is performed by maximization of MI and temporal synchronization is performed using electrocardiograph signals and a latency compensation method. Validation results show that TRE is 1.7 +/- 0.4 mm for a beating heart phantom and 2.0 +/- 0.4 mm in an in-vivo animal study. The multi-modality dynamic image registration methods presented in this thesis have the potential to substantially improve interactive image guidance for minimally invasive intracardiac beating heart surgery, and also to offer improved diagnostic capabilities for cardiac disease.;Keywords: Multimodality image registration; ultrasound; intracardiac surgery: beating heart surgery; minimally invasive surgery; image-guided surgery; surgical navigation; temporal synchronization, dynamic image registration...