Date of Graduation

8-2015

Document Type

Thesis (MS)

Program Affiliation

Medical Physics

Degree Name

Masters of Science (MS)

Advisor/Committee Chair

Kyle Jones

Committee Member

Dianna Cody

Committee Member

Cheenu Kappadath

Committee Member

Tinsu Pan

Committee Member

Shouhao Zhou

Abstract

Multiphase renal CT exams are a commonly used imaging technique for the diagnosis of renal masses. The pre-contrast, or true unenhanced (TUE), image provides a baseline for enhancement measurements which is an important criteria used to characterize renal lesions, consequently it is crucial that CT numbers measured in TUE images be accurate. The purpose of this work is to assess the feasibility of replacing TUE with virtual unenhanced (VUE) images derived from DECT data in renal CT exams. Eliminating TUE image acquisition would reduce patient dose and increase patient throughput, improving clinical efficiency.

A retrospective study was conducted for 60 consecutively selected patient exams. VUE and TUE images were compared qualitatively and the differences were tested using a Bayesian Hierarchical model. VUE images were found to be inferior to TUE images for visualization of major vessels and depiction of liver parenchyma. CT numbers were measured in the liver, spleen, spine, aorta, cystic lesions, subcutaneous fat, renal cortex and medulla, and the differences were tested with a Student’s paired t-test. There were significant differences between TUE and VUE measurements ( p-value > 0.05) in the spleen, spine, aorta, renal cortex, subcutaneous fat, and inferior vena cava. However, evaluation of the clinical relevance based on grayscale perceptibly indicated that the difference for the spleen and subcutaneous fat are not clinically meaningful.

The rapid kVp-switching GE CT750HD scanner was used to assess enhancement accuracy when using VUE compare to TUE images as the baseline for enhancement calculations across a wide range of clinical scenarios simulated in a phantom study, and the results were analyzed using Bayesian Hierarchical models. For simulation of angiomyolipoma and benign cystic lesions, enhancement values were not significantly different. However, for simulation of Bosniak category II-IV lesions, differences in measured enhancement were found to be significant. Additionally, the effect of ASIR level used in image reconstruction was assessed, and found not to affect measured CT number using a mixed effects model.

Differences in measured enhancement values for simulated borderline enhancing renal lesions demonstrate that replacement of TUE with VUE images is not feasible with the current iteration of the algorithm.

Share

COinS