Radiation dose reduction strategies for the dose-image optimisation in abdominal CT and a demonstration of validity of VGA

Myeongsoo Kim

Research output: ThesisDoctoral Thesis

499 Downloads (Pure)


The purpose of this thesis was to investigate radiation dose reduction regarding dose image quality optimisation in multi-detector CT abdominal examinations and to demonstrate the validity of a visual grading method that has been used widely by researchers in the literature. In this thesis, two major studies were performed based on quantitative and qualitative assessments. Many previous studies have accomplished dose reduction, technological developments and improved scan techniques, and these have achieved a significant dose saving compared to that in the initial development stage of multi-detector CT scanning. Knowledge of the dose level corresponding to optimal image quality is vague because the dose image quality level or standard depends on the observer’s subjective preference for diagnosis. For instance, diagnostic reference levels (DRLs) are determined at the 75th percentile point of collected survey CTDIvol data. However, these dose data are distributed with an acceptable image quality level between a minimal dose and maximum dose. Therefore, use of the 75th percentile point is questionable due to a broad dose range, and so it motivates exploration of a lower decision point,specifically the 50th. Also, an image quality analysis using a visual grading method for a subjective assessment is another question to achieve the dose optimisation. This thesis employed two hypotheses. The first hypothesis is demonstrated by two studies. One is a comparison of physical parameters of CTDIvol, noise (HU), CNR (HU), and VGA-score corresponding to the 75th and 50th percentile points. An anthropomorphic phantom known as PH-5 was employed to produce simulated CT abdominal images using a broad range of effective mAs under three different tube potentials (80, 100, and 120 kVp). 39 simulated images were obtained, and noise (HU) and CNR (HU) were measured. The second study involved a retrospective study with 107 clinical abdominal CT images that were randomly retrieved to compare the statistical significance of BMI, body weight, and cross-sectional area (CSA) corresponding to radiation dose; CTDIvol, to investigate a correlation between DRLs and patient size; a small, average and large group. The second hypothesis was explored by using the results of the VGA study to examine the validity of the VGA metric combined with logistic psychometric function compared with VGC, and VGR. 156 images of the PH-5 phantom used for this study. Two different reconstruction methods obtained the images: a filtered back projection (FBP) and adaptive statistical iterative reconstruction (ASiR). In the first study, the 75th and 50th percentile levels of DRLs with CTDIvol yielded a substantial difference with a 17.1 mGy and 12.59 mGy of GE DiscoveryTM 750HD (P =.096) and conversely a 15.02 mGy and 11.43 mGy of Philips Ingenuity (P =. 086). However, the VGAs of image quality demonstrated no statistically significant difference (p<.01). The difference in image noise and CNR values between the 75th and 50th percentiles were 15% and 14.5% for the GE DiscoveryTM 750HD (p= .012), and 13% and 12.9% of Philips Ingenuity (p<.01), respectively. In the second study, the minimum and maximum values of the CSA were 335cm2 and 921cm2 and demonstrated a 0.912 of the correlation coefficient, whereas, the correlation coefficients of BMI and weight were 0.791 and 0.854 (p < 0.01). A reference dose at the 75% level of the dose distribution classified according to the CSA and body weight were 11.2 and 11.64mGy in an average body habitus group, respectively. The achievable doses at the 50% level were 9.46 and 9.79mGy. The dose variation from 75% to 50% level in the CSA showed a CTDIvol decreasing with body habitus of 21% (small habitus) and 4% (large habitus). A threshold dose Dt is defined as the CTDIvol at a level just acceptable to produce diagnostic level image quality and a figure of merit (FOM) is defined as the slope of the normalised logistic function. The Dt and FOM were found to be 5.4, 8.1 and 9.1 mGy and 0.47, 0.51 and 0.38 of tube voltages of 80, 100 and 120 kVp, respectively, from the images reconstructed by the filtered back projection (FBP) reconstruction. The VGC results were compared to both scanners, taking all images into account. The area under the curve was determined to be AUCVGC=0.575. The Philips scanner performed better than the GE scanner. The AUCVGC=0.5, referred to an equal performance. The physical and image quality variation between the 75% and 50% points is not significantly different in a diagnostic interpretation. Consequently, the achievable doses (ADs); 50 % are acceptable to be a newer reference level of DRLs for achieving optimisation. If the possible doses align with the reference doses (75%) in current reference level, the ADs would be an effective guide level for potential dose reduction. The classification of the human habitus based on the CSA of the individual patient should be the method of choice for selecting the optimal exposure parameters for abdominal MDCT. The limitation of this study is that it investigates a single population; the results should be validated by future research on both a larger sample of Korean patients and on international communities. The VGA combined with psychometric function (VGAPF) can be used to find the optimal X-ray tube voltage for CT imaging and to evaluate the performance of image reconstruction techniques. The optimal x-ray peak voltage for the imaging of the PH-5 abdomen phantom by the Toshiba Aquilion OneTM MDCT system and Philips Ingenuity, are found to be at 100 kVp. GE Discovery is 120 kVp. The images reconstructed by the iterative reconstruction technique have lower noise, a higher contrast to noise ratio (CNR) as well as a lower dose to patients in comparison with images reconstructed by the FBP technique. However, the FBP reconstruction is still producing images with appropriate diagnostic accuracy. The VGC and VGR have a statistically significant correlation but not with the VGAPF. However, the VGAPF showed a right consistency in the study results compared to the VGC and VGR. Therefore, the accuracy of the VGAPF is better than the other methods.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Charles Sturt University
  • Zheng, Xiaoming, Co-Supervisor
  • Spuur, Kelly, Co-Supervisor
Award date20 Mar 2016
Place of PublicationAustralia
Publication statusPublished - 2016

Fingerprint Dive into the research topics of 'Radiation dose reduction strategies for the dose-image optimisation in abdominal CT and a demonstration of validity of VGA'. Together they form a unique fingerprint.

Cite this