Three-dimensional medical ultrasound acquisition and data registration and fusion

Multiple imaging technologies are routinely used in combination to help diagnose and treat illnesses. Medical images from one modality can be compared with subsequent images to track changes over time, or these images can be combined with data from a different modality to allow comparison of identical regions of interest. Image registration is the process by which two sets of data are spatially aligned, and image fusion describes the method of combining and displaying information from two sources of data.; Ultrasound registration is a challenge due to the two-dimensional nature of the modality. Unlike magnetic resonance imaging, it does not have an inherent coordinate system that can be used to align two scans from a patient. Before registration of ultrasound can be accomplished, the data must be transformed into a 3D coordinate system. This thesis presents two methods for generating 3D ultrasound data for the purpose of volume registration.; An interactive, three-dimensional medical image registration algorithm is presented and tested. Intrinsic landmarks, such as vessels and surfaces, are used to perform rigid registration and fusion of surrounding tissue. Phantom and in vivo data sets were acquired from MRI and US to test the algorithm. Registration error was measured in the final fusion images.; Two applications for ultrasound registration are presented. The first application is a method for comparing sonoelastography ultrasound with histology data to determine cancer detectability in excised prostate tissue. A process for creating 3D histology volumes from 2D images is presented. Prostate surfaces are segmented and registered to determine alignment error.; The second application deals with serial fusion of in vivo ultrasound data for lesion tracking. Doppler and harmonic imaging techniques are compared for vessel reconstruction accuracy. In vivo liver and thyroid data are acquired from two volunteers with known focal lesions to test the tracking capabilities of the registration algorithm. Scans are collected over a period of weeks and registered using segmented vessels and surfaces. Displacements between lesion centroids are calculated and reported.