UBC Theses and Dissertations
Development of a prototype scanner for pulsed ultrasound computed tomography McFarland, Sheila J.
A prototype scanner for pulsed ultrasound computed tomography (UCT) has been built and tested with the aim of developing new imaging techniques that hold potential for the improved early detection of breast cancer. Two reconstruction algorithms were tested for their ability to produce quantitative 2D cross-sections of γ[sub k](r) and reflectivity, R(r), where [formulas not included] k(r) and p(r) are the compressibility and density throughout the image space, while k[sub o] and p[sub o] describe water. γ[sub p] is subsequently calculated from γ[sub k] and R. A Direct Fourier Method (DFM) that ignores attenuation was tested both as is and with the addition of an approximate attenuation correction based on the average distance ultrasound travels through tissue and water. An Algebraic Reconstruction Technique (ART) was also developed, which iteratively solves for γ[sub k] and R subject to rigorous attenuation correction. In computer simulation, the DFM produced valid γ[sub k] and γ[sub p] cross-sections only in the absence of attenuation and when |r| < 0.05|r₃|, where r and r₃ are the position vectors of any scatter point and the source or detector, respectively. Points separately by 0.07 mm were resolved in simulation, and point amplitudes were reconstructed quantitatively to within 97 ± 3% and 95 ± 4% of their actual values for γ[sub k] and γ[sub p], respectively. Beyond |r| < 0.05|r₃|, Gaussian point-spread-functions (PSF's) were distorted into low amplitude halos. The approximate attenuation correction failed in simulation. Experimentally, the DFM reconstructed images of a cylindrical tissue phantom with negative γ[sub k] and R and a 3 mm diameter. The resulting γ[sub k] and R cross-sections were primarily negative with an object diameter of 2.94 ± 0.06 mm and 1.94 ± 0.04 mm for γ[sub k] and R, respectively. A valid γ[sub p] could not be calculated due to the narrow R function. Experimentally, point-spread-functions were severely distorted by errors of only 1-2% in the measurement of |r₃|. In simulation, valid γ[sub k] and γ[sub p] cross-sections were reconstructed with the ART for |r| < 0.3|r[sub j]| and in the presence of strong attenuation. Points separated by 0.05 mm were resolved in simulation, and point amplitudes were reconstructed quantitatively to within 100 ± 14% and 99 ± 4% of their actual values for γ[sub k] and γ[sub p], respectively. Strong attenuation was corrected to within an error of 0.2%. Convergence was not possible for images of more than 20 x 20 pixels due to ill-conditioning of the system. These theoretical results indicate, though, that the method holds promise given the application of conditioning algorithms.
Item Citations and Data