Patient-Image Registration using A-mode ultrasound localization of features
Bass, Wayne Andrew
The objective of this dissertation is to investigate the accuracy of point and surface based image space to physical space registration performed using a spatially tracked A-mode ultrasound transducer to localize features and to determine the applicability of these techniques for use in interactive, image-guided surgery. The accuracy of subcutaneous marker localization has been demonstrated using a phantom. An spatially tracked A-mode ultrasonic localization system was constructed. The system was used to examine the accuracy of transcutaneous localization of bone implanted fiducial marker analogs. The relationship between the number of signals used to localize the fiducial markers and localization accuracy was determined. Validation was performed by comparison to an optical system. The accuracy of surface registrations based on matching the outer surface of the skull as identified by ultrasound and in CT images has been estimated in a phantom. The ultrasonic localization system was modified for use in localizing the outer surface of the skull. The effect of changes in the image model parameters and image slice thickness on registration error were examined. The effect of variations in the speed of sound was also examined. The surface registration results were validated by comparison to a fiducial marker registration. A preliminary study on the accuracy of surface registrations in twelve human patients has been performed. The ultrasonic localization system was enhanced to synchronize the acquisition of ultrasonic and optical information. Patient motion in the CT images was compensated for using the Nbar system of the CRW stereotactic frame. The surface registration results were evaluated for three different speed of sound values corresponding to the speed of sound in the tissue components of human scalp. The correlation between the number of ultrasonic points used in the surface registration algorithm and the surface registration error was evaluated. Fiducial markers attached to the CRW stereotactic frame were used to validate the surface registration results. These experiments have demonstrated a spatially tracked A-mode ultrasound transducer capable of localizing both point and surface features that can be used in registration processes for interactive, image-guided procedures.