Data registration and fusion for cardiac applications

Abstract

The registration and fusion of information from multiple cardiac image modalities such as magnetic resonance imaging (MRI), X-ray computed tomography (CT), positron emission tomography (PET) and single photon emission computed tomography (SPECT) has been of increasing interest to the medical community as tools for furthering physiological understanding and for diagnostic of ischemic heart diseases. Ischemic heart diseases and their consequence, myocardial infarct, are the leading cause of mortality in industrial countries. In cardiac image registration and data fusion, the combination of structural information from MR images and functional information from PET and SPECT is of special interest in the estimation of myocardial function and viability. Cardiac image registration is a more complex problem than brain image registration. The non-rigid motion of the heart and the thorax structures introduce additional difficulties in registration.

In this thesis the goal was develop methods for cardiac data registration and fusion. A rigid registration method was developed to register cardiac MR and PET images. The method was based on the registration of the segmented thorax structures from MR and PET transmission images. The thorax structures were segmented from images using deformable models. A MR short axis registration with PET emission image was also derived. The rigid registration method was evaluated using simulated images and clinical MR and PET images from ten patients with multivessel coronary artery diseases. Also an elastic registration method was developed to register intra-patient cardiac MR and PET images and inter-patient head MR images. In the elastic registration method, a combination of mutual information, gradient information and smoothness of transformation was used to guide the deformation of one image towards another image.

An approach for the creation of 3-D functional maps of the heart was also developed. An individualized anatomical heart model was extracted from the MR images. A rigid registration of anatomical MR images and PET metabolic images was carried out using surface based registration, and the registration of MR images with magnetocardiography (MCG) data using external markers. The method resulted in a 3-D anatomical and functional model of the heart that included structural information from the MRI and functional information from the PET and MCG. Different error sources in the registration method of the MR images and MCG data was also evaluated in this thesis. The results of the rigid MR-PET registration method were also used in the comparison of multimodality MR imaging methods to PET.