Feature Based Registration of Ultrasound and CT Data of a Scaphoid
Loading...
Date
2010-05-28T18:30:45Z
Authors
Koslowski, Brian
Keyword
Feature-Based Registration , Ultrasound , CT , Scaphoid
Abstract
Computer assisted surgery uses a collection of different techniques including but not limited to: CT-guided, fluoroscopy-guided, and ultrasound-guided imaging which allows medical staff to view bony anatomy of a patient in relation to surgical tools on a computer screen. By providing this visual data to surgeons less invasive surgeries can be performed on a patient's fractured scaphoid. The data required for a surgeon to perform a minimally invasive surgery while looking only at a computer screen, and not directly at a patient's anatomy, will be provided by CT and ultrasound data. We will discuss how ultrasound and CT data can be used together to allow a minimally invasive surgery of the scaphoid to be performed.
In this thesis we will explore two techniques of registering segmented ultrasound images to CT data; an Iterative Closest Point (ICP) approach, and an Unscented Kalman Filter-based Registration (UKF). We use two different ultrasound segmentation methods; a semi-automatic segmentation, and a Bayesian segmentation technique. The segmented ultrasound data is then registered to a CT volume. The success or failure of the
registrations is measured by the error calculated in mapping the corresponding land-
marks to one another and calculating the target registration error. The results show that the Unscented Kalman Filter-based registration using the Bayesian segmentation of ultrasound images has the least registration error, and has the most robustness to error in initial alignment of the two data sets.