Thomas Hangartner (Advisor), David Short (Committee Member), Julie Skipper (Committee Member)
Master of Science in Engineering (MSEgr)
In computed tomography (CT) systems, many different artifacts may be present in the reconstructed image. These artifacts can greatly reduce image quality. For our laboratory prototype CT system, a fan-beam/cone-beam focal high-resolution computed tomography (fHRCT) scanner, the major artifacts that affect image quality are distortions due to errors in the reconstruction algorithm's geometric parameters, ring artifacts caused by uncalibrated detectors, cupping and streaking created by beam hardening, and patient-based motion artifacts. Optimization of the system was required to reduce the effects of the first three artifact types, and an algorithm for correction of translational motion was developed for the last. System optimization of the system occurred in three parts. First, a multi-step process was developed to determine the geometric parameters of the scanner. The ability of the source-detector gantry to translate allowed a precise method to be created for calculating these parameters. Second, a general flat-field correction was used to linearize the detectors and reduce the ring artifacts. Lastly, beam hardening artifacts were decreased by a preprocessing technique. This technique assumes linear proportionality between the thickness of the calibration material, aluminum, and the experimental measurement of ln(No/N), where No is the total number of photons entering the material and N is the number of photons exiting the material. In addition to system optimization to minimize artifacts, an algorithm for correction of translational motion was developed and implemented. In this method, the integral mass and center of mass at each projection angle was seen to follow a sinusoidal or sinusoidal-like curve. Fits were used on the motion-encoded sinograms to determine both of these curves and, consequently, the amount and direction of motion that occurred. Each projection was individually adjusted to compensate for this motion by widening or narrowing the projection based on the ratio of the actual and calculated ideal projection integrals and shifting the projection to match the actual centroid to the calculated ideal location. A custom imaging phantom with an outer diameter of approximately 16 mm was used to test the motion-correction algorithm in both simulated and experimental cases. A baseline of the error measured, taken as a fraction, was established as 0.16 for motion-free images measured on the scanner. Various motion patterns were tested. These included the distance of motion, the angle at which the motion occurred, and the ratio of the sinograms that was corrupted by motion. Experimental testing showed a maximum error increase of 2.7% from the baseline error for the motion-corrected images at 4 mm motion. The overall optimization provided acceptable results for the reconstructed image and good-quality projections for use in the motion-correction algorithm. Distortion and ring artifacts were almost completely removed, and the beam hardening artifacts were greatly reduced. The motion-correction algorithm implemented in this thesis helps minimize the amount of error due to translational motion and provides a foundation for future corrections of more complex motions.
Department or Program
Department of Biomedical, Industrial & Human Factors Engineering
Year Degree Awarded
Copyright 2012, all rights reserved. This open access ETD is published by Wright State University and OhioLINK.