CT Radiation Dose and Image Quality Optimization Using a Porcine ...

Peer Review

CT Radiation Dose and Image Quality Optimization Using a Porcine Model Francis Zarb, PhD Mark F McEntee, PhD

Louise Rainford, PhD

Purpose To evaluate potential radiation dose savings and resultant image quality effects with regard to optimization of commonly performed computed tomography (CT) studies derived from imaging a porcine (pig) model. Methods Imaging protocols for 4 clinical CT suites were developed based on the lowest milliamperage and kilovoltage, the highest pitch that could be set from current imaging protocol parameters, or both. This occurred before significant changes in noise, contrast, and spatial resolution were measured objectively on images produced from a quality assurance CT phantom. The current and derived phantom protocols were then applied to scan a porcine model for head, abdomen, and chest CT studies. Further optimized protocols were developed based on the same methodology as in the phantom study. The optimization achieved with respect to radiation dose and image quality was evaluated following data collection of radiation dose recordings and image quality review. Relative visual grading analysis of image quality criteria adapted from the European guidelines on radiology quality criteria for CT were used for studies completed with both the phantom-based or porcine-derived imaging protocols. Results In 5 out of 16 experimental combinations, the current clinical protocol was maintained. In 2 instances, the phantom protocol reduced radiation dose by 19% to 38%. In the remaining 9 instances, the optimization based on the porcine model further reduced radiation dose by 17% to 38%. Conclusion The porcine model closely reflects anatomical structures in humans, allowing the grading of anatomical criteria as part of image quality review without radiation risks to human subjects. This study demonstrates that using a porcine model to evaluate CT optimization resulted in more radiation dose reduction than when imaging protocols were tested solely on quality assurance phantoms.

C

omputed tomography (CT) imaging as a source of ionizing radiation is a growing concern.1,2 Developments in CT technology provide a wide range of clinical applications, increasing its use. Although the benefits of justified CT examinations outweigh potential risks, several international research groups are investigating methods of limiting radiation dose while maintaining diagnostic efficacy.1,3-5

Literature Review

Radiologists and radiologic technologists can manipulate scan parameters to reduce or increase the radiation dose to patients. Dose reduction techniques can result in images with a higher level of noise. Finding an optimal balance between radiation dose and image quality benefits

patients because radiologists are more confident with the diagnostic process. Thus, radiologists’ involvement in optimization research is essential.5-10 Many radiation dose reduction strategies have been introduced in CT. These strategies are based mainly on the assessment of image quality and on modifications of CT scan parameters using psychophysical measures on quality assurance (QA) phantoms.11-15 However, results of QA phantom images might not reflect the effects seen in clinical human images. Therefore, a more realistic method of image quality assessment is required, 6 such as an animal model resembling human anatomical structures. The use of animals in biomedical research has been common practice for many years.16 Researchers typically

RADIOLOGIC TECHNOLOGY, November/December 2013, Volume 85, Number 2

127

Peer Review CT Radiation Dose and Image Quality Optimization Using a Porcine Model

Figure 1. The Catphan 600 low-contrast module.

opt for animals whose genetic makeup and anatomy closely resemble human physiology. One such animal species is the Sus scrofa domesticus, or farm pig. The porcine (pig) cardiovascular system, skin, and digestive tract are considered effective models for human transplants. Because of these similarities, the toxic effects of chemicals and drugs on porcine models resemble the substances’ effects on humans more closely than other laboratory animals.17-19 Porcine models also commonly are used for training in different aspects of medicine such as anaesthesia and surgery.17-22 However, no literature reports the use of porcine models in radiology as part of radiation dose optimization.

Methods

To compare optimization of radiation dose and image quality, 3 studies (head, abdomen, and chest) were performed using both a CT Catphan 600 phantom (Phantom Laboratory, Salem, New York) and a porcine model. The CT images were produced by scanning the porcine model using the current protocol, the QA phantom-based optimized protocol, and the further optimized porcine model protocol prior to using the final optimized protocols in a clinical setting. The evaluation of the porcine images was based on the European guidelines on quality criteria for CT.23 128

Approval for the study was obtained from the University of Malta Research and Ethics Committee (UREC Ref No: 001/2009), the Animal Welfare Committee (Licence RP 03/2010), and the participating hospitals. The study was conducted in Malta in 4 CT suites involving a SOMATOM Emotion Duo (Siemens, Munich, Germany); a Brilliance CT 64-slice scanner (Philips, Andover, Massachusetts); a BrightSpeed 16-slice scanner (GE Healthcare, Wauwatosa, Wisconsin); and a GE HiSpeed NX/I dual CT scanner. A 2010 CT radiation dose survey in Malta identified the 3 most commonly requested CT studies: head (43%), abdomen (33%), and chest (9%).24 Scan parameters for each of these studies from all 4 CT suites were collected using a survey questionnaire based on the NRPB-W67 developed by the National Radiation Protection Board during the 2003 radiation dose survey in the United Kingdom.25 Regression analysis performed on the data identified milliamperage (mA), kilovoltage (kV), and pitch as the 3 major contributors to radiation dose in these CT studies.24 Phantom-optimized protocols were developed by scanning a CT phantom, starting with the current protocol parameters and systematically lowering the mA incrementally until significant changes (P < .05) were measured in 3 image quality characteristics: contrast resolution, noise, and spatial resolution. The limit of mA reduction was identified as the one previous to the setting producing significant changes to all 3 image quality characteristics. This mA setting was then applied with similar reductions in kV. Once the kV limit was identified, the mA and kV settings were applied, increasing the pitch. Image quality was objectively evaluated for contrast resolution by measuring the number of contrast target discs from the low contrast resolution module within the phantom (see Figure 1). Noise was measured by recording the mean standard deviation from circular regions of interest taken at the center and 4 peripheral locations within the image uniformity module (see Figure 2). Spatial resolution was gauged by the number of distinctly discernible line pairs within the high spatial resolution module (see Figure 3).26 The threshold limit for radiation dose optimization was established as the one that resulted in changes in noise and contrast resolution without affecting spatial resolution.27 RADIOLOGIC TECHNOLOGY, November/December 2013, Volume 85, Number 2

Peer Review Zarb, McEntee, Rainford

Figure 2. The Catphan 600 image uniformity module with regions of

interest measurement locations.

Figure 3. The Catphan 600 high spatial resolution module.

The phantom protocols were then applied to scan a porcine model. A humanely euthanized 5-month-old female farm pig was transported to the CT suites and imaged over the course of one day. The porcine model weighed 66 kg, with an abdominal anteroposterior

diameter of 30 cm and transverse diameter of 28 cm, reflecting similar attenuation characteristics of an average-sized human adult (weight 70 kg +/- 5 kg; abdomen: anteroposterior diameter 28 cm and transverse diameter 34.5 cm).23,28 Further optimization was achieved during the porcine scanning as images were objectively quantified and rescanned using lower scan parameters. The same method used in the phantom studies was followed on all participating CT suites with the threshold of optimization identified as the one before significant changes were recorded in noise levels. Noise was measured by recording the mean standard deviation from 4 circular regions of interest at the cerebrum for the head, the liver and spleen (level of the upper pole of the right kidney) for the abdomen, and the heart (level of the xiphoid process) for the chest. The sets of porcine studies across the CT suites were compiled for final image quality evaluation and comparison. Porcine images using the current protocol (P0) were used as a reference for comparison to porcine images obtained using the Catphan 600 protocol (P1) and further porcine-optimized protocol (P2). Relative visual grading analysis was conducted to quantify subjective opinions about image quality. The visibility of anatomical structures was compared and graded against the same structures in a reference scan. The list of image quality criteria used in this study was based on the list of CT quality criteria issued in the European Guidelines on Quality Criteria for Computed Tomography (EUR 16262).23 The underlying assumption in the use of these criteria is that if normal anatomy is reproduced accurately, then pathology will be demonstrated. 6 The European criteria were adapted to account for the anatomical differences between humans and pigs and the differences between scanning live and dead models. The adaptations were made on the advice of a consultant radiologist and a veterinary surgeon after they reviewed the porcine images. Radiation dose was recorded in terms of CT dose descriptors, volume CT dose index, and dose-length product from the displayed measurements on the respective CT units. The scan length was kept constant as the same porcine model was used for the same anatomical regions for the different protocols. The evaluation panel consisted of 9 observers: 4 radiologists and 5 radiologic technologists, all with


129


more than 2 years of CT imaging experience. The only training given to the observers was one familiarization session in porcine anatomy. The observers graded the visibility of the structures on a 5-point scale:  0 – visibility inferior to the reference image and no diagnostic value.  1 – visibility inferior to the reference image but with diagnostic value.  2 – visibility equal to the reference image.  3 – visibility superior to the reference image.  4 – visibility far superior to the reference image. A visual grading analysis score calculated from these results allowed statistical analysis of the differences between the protocols in terms of image quality.29 The Statistical Package for Social Sciences (SPSS) version 19 (IBM, Armonk, New York) was used to compute statistics where differences were considered significant outside the 95% confidence interval. The Kruskal-Wallis test was computed to compare the rating scores of all the quality criteria for multiple scan protocols. When no significant difference (P > .05) was found among the protocols in terms of image quality, the protocol that gave the least radiation dose was identified as the optimized protocol to apply in a future clinical trial. When there was a significant difference (P < .05) between the protocols, the reference protocol (P0) was analyzed against each of the other protocols separately using the Mann-Whitney U test so the protocol with no significant difference in image quality could be identified. The current protocol (P0) was retained when a significant difference between all the protocols was identified.30,31 In terms of radiation dose ranking, the current protocol (P0) gave the highest radiation dose, while the further optimized porcine protocol (P2) gave the lowest among the 3 scan protocols. Thus, the phantom protocol (P1) produced a lower radiation dose than the current protocol (P0) but a higher radiation dose than the further optimized porcine protocol (P2). Therefore, P0 > P1 > P2.

Results

Interobserver reliability was measured using Cohen Kappa (κ), which measures the agreement among raters who classify items into categories, and ranged 130

between -0.12 and 0.39, indicating a poor to fair agreement among the observers. The κ range among the radiologists was -0.11 and -0.01, while that among the radiographers was -0.08 and 0.08. The Mann-Whitney U test showed no significant difference between the κ agreements of the 2 disciplines (P ≤ .12). It should be noted that variations in observer agreement are common. Observers disagree with one another and with their own findings for 2 reasons: the subjective nature of image interpretation related to image quality or factors influencing the ability of observers to interpret the image, such as knowledge, experience, and competence of the observer.32 Table 1 demonstrates the selected protocols and scan parameter changes performed in reducing radiation dose. In 5 instances further optimization from the current protocol (P0) was not viable, and the current protocol was retained. This was the case, for example, in suite 1 for the cerebrum and abdomen, suite 3 for the abdomen and chest, and suite 4 for the chest studies. The phantom protocol based on the Catphan 600 tests (P1) and the results of the porcine study evaluation were in agreement in 2 cases: the phantom results (P1) were considered sufficient for optimization in suite 1 for the chest and suite 4 for the abdomen studies. Radiation dose reduction using the Catphan 600 protocols ranged between 19% and 38%. In the other 9 scenarios, further optimization beyond the phantom results was possible, and the radiation dose further reduced without compromising image quality and diagnostic efficacy. These scenarios were:  Suite 1 – posterior fossa.  Suite 2 – posterior fossa, cerebrum, abdomen, and chest.  Suite 3 – posterior fossa and cerebrum.  Suite 4 – posterior fossa and cerebrum. Radiation dose reduction using further optimized protocols during the porcine study ranged between 17% and 38%.

Discussion

Adaptations to the image quality criteria required for each anatomical region of the porcine model replicating human anatomy were considered, with the head criteria in particular need of revision (see Table 2). All mammals have the same brain structures, but the human brain is larger than that of most other animals. RADIOLOGIC TECHNOLOGY, November/December 2013, Volume 85, Number 2


Table 1 Summary of Scan Parameter Changes and Their Effect on Patient Dose Posterior Fossa

Cerebrum

a

Abdomen

Chest

Parameter Change

% Dose Reduction

Parameter Change

% Dose Reduction

Parameter Change

% Dose Reduction

Parameter Change

% Dose Reduction

S1

P2 -20% mA (75-60 mA)

P = .05 (-20%)

P0 no change

P = .002 no change

P0 no change

P = .001 no change

P1 -20 kV (130-110 kV)

P = .18 (-38%)

S2

P2 -30% mA (350-250 mA)

P = 1.0 (-30%)

P2 -30% mA (350-250 mA)

P = .1 (-30%)

P2 + pitch (0.89-1.14)

P = .42 (-17%)

P2 -40% mA (250-150 mA max)

P = .21 (-38%)

S3

P2 -20 kV (120-100 kV)

P0 no change

P < .001 no change

P0 no change

P < .001 no change

P1 -19% mA (160-130 mA)

P = .7 (-19% )

P0 no change

P < .001 no change

P = .13 (-36%)

+ noise index (3.8-3.99) S4

P2 -38% mA (130-80 mA)

P2 + max mA (200-440 mA)

P = .69 (-24%)

+ noise index (3.8-9.5) P = .53 (-38%)

P2 -38% mA (130-80 mA)

P = .69 (-38%)

Abbreviations: S, suite; P0, current protocol; P1, phantom protocol; P2, further optimized porcine protocol. a All dose reductions had 95% confidence intervals.

Two significant differences in humans are a larger viscera-motor system, which controls laryngeal and respiratory muscles essential for speech, and a larger prefrontal cortex with more extensive and widespread connections crucial for symbolic learning and cognitive functions. 33 Hence, it was difficult to identify the basal ganglia, ventricular system, and mesencephalon in the porcine CT images of the head because of their small size (see Figure 4). Criterion 6 also required revision, as no contrast media was administered to the porcine model, thus hindering visualization of this criterion. However, a further criterion was included (criterion 7) to enable evaluation of bone structures. Abdominal and chest criteria required fewer amendments because the liver, gallbladder, digestive system, and urinary system are similar between the 2 species.34 Abdominal and thoracic structures reflected similar attenuation in terms of fat distribution and size (there is less intraperitoneal fat demarcating abdominal viscera in the porcine model), requiring fewer adaptations to the criteria of the abdomen and chest. However, the

absence of contrast media in the porcine images also reduced delineation of structures (see Tables 3 and 4; Figures 5 and 6). A further complication was that the optimization of protocols had to be individualized to the different manufacturers because of differences in image acquisition mechanisms related to the specific CT equipment. Reductions of specific parameters on one unit did not yield the same effects on a different unit in terms of radiation dose and image quality. In most cases, optimization was achieved by a reduction in mA, the major contributor to radiation dose in CT. Also, the reduction in mA produced a similar reduction in radiation dose in terms of volume CT dose index and dose-length product and, therefore, radiation dose overall.35 The use of low mA values has been investigated, especially in the examination of specific anatomical regions such as the head, abdomen, and chest. Studies have shown that it is possible to reduce considerably the mA and still obtain images adequate for diagnosis.13,36-40


131


dose because radiation dose is inveresly proportional to pitch. 35,42 These parameter changes produced further optimization over the phantom studies in 9 instances and a further dose reduction ranging between 17% and 38% during the porcine study. In suite 2, scan protocols for all areas could be further optimized and the radiation dose reduced. This research group is currently extending optimization to pediatric and obese patients. Implementing the final set of selected optimized protocols in the clinical setting enabled a more detailed evaluation of the protocols. Visual grading characteristic analysis with radiologists rating the quality of anatomical structures seen on clinical images determined whether the selected protocols were appropriate for clinical use. These optimized protocols have been implemented in clinical use since November 2011 across the 4 clinical centers that participated, including the national clinical teaching hospital of Malta.

Table 2 Adapted Critical Reproduction Criteria for Human and Porcine Head Human Brain

Porcine Brain

1. Visually sharp reproduction of the border between white and gray matter

Same

2. Visually sharp reproduction of the basal ganglia

Not applicable

3. Visually sharp reproduction of the ventricular system

Not applicable

4. Visually sharp reproduction of the cerebrospinal fluid space around the mesencephalon

Not applicable

5. Visually sharp reproduction of the cerebrospinal fluid space over the brain

Same

6. Visually sharp reproduction of the great vessels and the choroid plexus after intravenous contrast media

Not applicable

7. Not applicable

Visually sharp discrimination between cortical and spongy bone

1 5 A

B

Figure 4. Computed tomography (CT) cross-sectional anatomy for

the human head (A) and porcine head (B) analyzed in Table 2. The numbers in B correspond to criteria in Table 2.

Changes in kV also reduce radiation dose but increase noise in the CT image. These effects depend on the type of CT equipment and software installed on the unit.37,41 Increasing the pitch decreases the radiation dose because the time any point spends in the x-ray beam also decreases. That point receives a reduced radiation 132

Limitations Observer variances could be attributed to the inexperience and limited training in identifying porcine anatomy. To improve this, additional training or labeling the anatomical structures on the reference set of images during the evaluation is recommended. In addition, methods such as receiver operator characteristic analysis or visual grading characteristic analyses require further consideration, as these also can be used to reduce observer variances.43,44 The ability to predict image quality based on physical parameters and using QA phantoms has been questioned by researchers, as changes in image quality on phantom images might differ from that on anatomical images of patients. 6,45 Radiologic technologists need to see how changes in technical parameters affect visualization of anatomical structures in patient images. The inability to correlate changes in technical imaging parameters to measures of image quality could be related to difficulties encountered while looking at the combination of scan parameters and obtaining anatomical images in a controlled way. The use of a porcine model helped overcome these difficulties; however, other issues of concern existed. First, hospital authorities required that the porcine specimen be sealed adequately and that departmental RADIOLOGIC TECHNOLOGY, November/December 2013, Volume 85, Number 2


Table 3

14

Adapted Critical Reproduction Criteria for Human and Porcine Abdomen Human Abdomen

Porcine Abdomen

1. Visually sharp reproduction of the liver parenchyma and intrahepatic vessels

Same

2. Visually sharp reproduction of the splenic parenchyma

Same

3. Visually sharp reproduction of the intestine

Not applicable

4. Visually sharp reproduction of the perivascular retroperitoneal space

Not applicable

5. Visually sharp reproduction of the pancreatic contours

Same

6. Visually sharp reproduction of the duodenum

Same

7. Visually sharp reproduction of the kidneys and proximal ureters

Visually sharp reproduction of the kidneys

8. Visually sharp reproduction of the aorta

Visually sharp discrimination between the aorta and inferior vena cava

9. Visually sharp reproduction of the aortic bifurcation and common iliac arteries

Not applicable

10. Reproduction of lymph nodes < 15 mm in diameter

Not applicable

11. Reproduction of branches of the abdominal aorta

Not applicable

12. Visually sharp reproduction of the vena cava

Not applicable

13. Reproduction of tributaries to the vena cava, in particular the renal veins

Visually sharp reproduction of the renal veins

14. Not applicable

Visually sharp reproduction of the gallbladder

15. Not applicable

6 2 1

A

B

13 7 15

Visually sharp reproduction of the thecal sac and nerve roots in the lumbar region

C

D

Figure 5. CT cross-sectional anatomy for the human (A, C) and

porcine abdomen (B, D) analyzed in Table 3. The numbers in B and D correspond to the criteria in Table 3.

disinfection protocols were followed during the imaging process for infection control. Infection control was the reason a euthanized porcine model was preferred over an anesthetized pig. Imaging a euthanized model, however, could result in differences in attenuation from imaging a live model. Performing all the imaging on the same day and in the shortest time possible limited the effects of decomposition on tissue attenuation. Other differences in attenuation are the result of the diffusion of water through soft tissue, such as accumulation of fluid in the paranasal sinuses, air in the cerebral arterial system, air-fluid levels in the bowels, free peritoneal fluid, pleural effusions, bilateral airspace filling, fluid in the trachea and throughout the respiratory tree, pericardial effusions, and air-fluid levels in the heart. 46,47 Therefore, imaging a euthanized animal did not hinder the evaluation of the image quality criteria except for criterion 1, the loss of differentiation between white and gray matter in the brain.

Conclusion

This study found that the results of objective phantom studies do not always present the best possible outcomes for radiation dose and image quality optimization. Significant radiation dose optimization was possible with the porcine model, resulting in significant


133


A

Table 4

B

Adapted Critical Reproduction Criteria for Human and Porcine Chest

11

Human Chest

Porcine Chest

3

1. Visually sharp reproduction of the thoracic aorta

Not applicable

8

2. Visually sharp reproduction of the anterior mediastinal structures, including thymic residue (if present)

Not applicable

3. Visually sharp reproduction of the trachea and main bronchi

Same

4. Visually sharp reproduction of the paratracheal tissue

Not applicable

5. Visually sharp reproduction of the carina and lymph node area

Not applicable

6. Visually sharp reproduction of the oesophagus

Same

7. Visually sharp reproduction of the pleuromediastinal border

Not applicable

8. Visually sharp reproduction of large and medium-sized pulmonary vessels

Same

9. Visually sharp reproduction of segmental bronchi

Visually sharp reproduction of peripheral small bronchi

10. Visually sharp reproduction of the lung parenchyma

Same

11. Visually sharp reproduction of the border between the pleura and the thoracic wall

Visually sharp reproduction of pleural reflections (fissures)

12. Not applicable

Visually sharp reproduction of the thecal sac and nerve roots in the thoracic region

Figure 6. CT cross-sectional anatomy for the human (A) and porcine

(B) chest analyzed in Table 4. The numbers in B correspond to criteria in Table 4.

radiation dose savings. Without the CT images containing anatomical structures, as in this porcine model study, optimization cannot be maximized. Further work is required to address specific issues with using the porcine model; however, the findings of this work suggest that researchers should consider animal 134

experiments with regard to diagnostic image quality analysis when undertaking optimization research. Perhaps current QA phantoms need to be improved so they more closely simulate clinical conditions and clinically relevant image information. Francis Zarb, PhD, is a lecturer in the Department of Radiography, Faculty of Health Sciences, at the University of Malta. He can be reached at [email protected]. Mark F McEntee, PhD, is a senior lecturer in the Discipline of Medical Radiation Sciences, Brain and Mind Research Institute, Faculty of Health Sciences, at the University of Sydney in Australia. He can be reached at [email protected]. Louise Rainford, PhD, is a senior lecturer and head of Diagnostic Imaging and Biological Imaging Research in the School of Medicine & Medical Science at University College Dublin in Ireland. She can be reached at louise.rainford @ucd.ie. Reprint requests may be mailed to the American Society of Radiologic Technologists, Communications Department, at 15000 Central Ave SE, Albuquerque, NM 87123-3909, or e-mailed to [email protected]. © 2013 American Society of Radiologic Technologists

References

1. Council Directive 97/43/EURATOM of 30 June 1997 on health protection of individuals against the dangers of ionizing radiation to medical exposure. European Union Web site. http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri =CELEX:31997L0043:EN:HTML. Accessed September 13, 2013. RADIOLOGIC TECHNOLOGY, November/December 2013, Volume 85, Number 2


2. McCollough CH, Bruesewitz MR, Kofler JM Jr. CT dose reduction and dose management tools: overview of available options. Radiographics. 2006;26(2):503-512. 3. A critical review of the draft 2005 ICRP recommendations. Paper presented at: European Commission scientific seminar; November 4, 2004; Luxembourg. http://ec.europa.eu /energy/nuclear/radiation_protection/doc/publication /150.pdf. Accessed August 14, 2013. 4. Fazel R, Krumholz H, Wang Y, et al. Exposure to low-dose ionizing radiation from medical imaging procedures. N Eng J Med. 2009;361(9):849-857. doi:10.1056/NEJMoa0901249. 5. McCollough C, Guimaraes L, Fletcher J. In defence of body CT. AJR Am J Roentgenol. 2009;193(1):28-39. doi:10.2214 /AJR.09.2754. 6. Sandborg M, Tingberg A, Dance DR, et al. Demonstration of correlations between clinical and physical image quality measures in chest and lumbar spine screen-film radiography. Br J Radiol. 2001;74(882):520-528. 7. Wilting J, Zwartkruis A, van Leeuwen M, Timmer J, Kamphuis AG, Feldberg M. A rational approach to dose reduction in CT: individualized scan protocols. Eur Radiol. 2001;11(12):2627-2632. 8. Golding S, Shrimpton P. Radiation doses in CT: are we meeting the challenge? Br J Radiol. 2002;75(889):1-4. 9. McCollough C. Automatic exposure control in CT: are we done yet? Radiology. 2005;237(3):755-756. doi:10.1148 /radiol.2373051151. 10. Freiherr G. Dose-saving technologies proliferate throughout CT. Diagnostic Imaging Web site. http://www.diagnostic imaging.com/articles/dose-saving-technologies-proliferate -throughout-ct. Published April 29, 2010. Accessed September 13, 2013. 11. Haaga J, Miraldi F, Macintyre W, LiPuma JP, Bryan PJ, Wiesen E. The effect of mAs variation upon computed tomography image quality as evaluated by in vivo and in vitro studies. Radiology. 1981;138(2):449-454. 12. Rogers LF. Radiation exposure in CT: why so high? AJR Am J Roentgenol. 2001;177(2):277. 13. Ravenel J, Scalezzi E, Huda W, Garrisi W. Radiation exposure and image quality in chest CT examinations. AJR Am J Roentgenol. 2001;177(2):279-284. 14. Haaga J. Radiation dose management: weighing risk versus benefit. AJR Am J Roentgenol. 2001;177(2):289-291. 15. Francone M, Catalano C, Vasselli F. Finding dose/quality balance presents CT imaging challenge. Diagnostic Imaging Web site. http://www.diagnosticimaging.com/articles /finding-dosequality-balance-presents-ct-imaging-challenge /page/0/4. Published October 2, 2008. Accessed September 13, 2013.

16. Swindle M. The use of animals in surgical research. J Invest Surg. 1988;1(1):3-4. 17. Swindle M, Smith A. Information Resources on Swine in Biomedical Research: 1990-2000. United States Department of Agriculture Web site. http://www.nal.usda.gov/awic/pubs /swine/. Accessed August 14, 2013. 18. Tanaka H, Kobayashi E. Education and research using experimental pigs in a medical school. J Artif Organs. 2006;9(6):136-143. 19. Swindle M, Makin A, Herron A, Clubb F, Frazier K. Swine as models of biomedical research and toxicology testing. Veterinary Pathology. 2012;49(2):344-356. doi:10.1177/0300985811402846. 20. Clarkson WA, Restrepo CS, Bauch TD, Rubal BJ. Using multidetector computed tomography in a swine model to assess the effects of sublingual nitroglycerine and intravenous adenosine on epicardial coronary arteries. Comp Med. 2009;59(6):573-579. 21. Gurr A, Kevenhörster KS. The common pig: a possible model for teaching ear surgery. Eur Arch Otorhinolaryngol. 2010;267(2):213-217. doi:10.1007/s00405-009-1040-6. 22. Aigner B, Renner S, Kessler B, et al. Transgenic pigs as models for translational biomedical research. J Mol Med. 2010;88(7):653-664. doi:10.1007/s00109-010-0610-9. 23. European guidelines on quality criteria for computed tomography. European Commission; 1999. http://www.drs .dk/guidelines/ct/quality/. Accessed August 14, 2013. 24. Zarb F, Rainford L, McEntee M. Image quality assessment tools for optimization of CT images. Radiography. 2010;16(2):147-153. 25. Shrimpton P, Hillier M, Lewis M, Dunn M. Doses from computed tomography (CT) examinations in the UK-2003 review. National Radiologic Protection Board. NRPB W67. 2005. http://www.biophysicssite.com/Documents/NRPB _W67/NRPB_W67.pdf. Accessed August 14, 2013. 26. The phantom laboratory. Catphan 500 and 600 manual. 2012. http://www.phantomlab.com/library/pdf/catphan 500-600manual.pdf. Accessed August 14, 2013. 27. Zarb F, Rainford L, McEntee M. Developing optimized CT scan protocols: phantom measurements of image quality. Radiography. 2011;17(2):109-114. 28. Kalra M, Maher M, Toth T, et al. Strategies for CT radiation dose optimization. Radiology. 2004;230(3):619-628. 29. Månsson L. Methods for the evaluation of image quality: a review. Radiat Protection Dosimetry. 2000;90(1-2):89-99. 30. Daly L, Bourke G, McGilvray J. Confidence intervals and hypothesis tests: two-group comparisons. In: Interpretation and Uses of Medical Statistics. 5th ed. Oxford: Wiley-Blackwell Science; 2000:218-221.


135


31. Kirkwood B, Sterne J. Relaxing model assumptions. In: Essential Medical Statistics. 2nd ed. Hoboken, NJ: WileyBlackwell Science; 2006:342-350. 32. Manning D, Gale A, Krupinski E. Perception research in medical imaging. Br J Roentgenol. 2005;78(932):683-685. doi:10.1259/bjr/72087985. 33. Deacon T. What makes the human brain different? Ann Rev Anthropol. 1997;26:337-357. doi:10.1146/annurev .anthro.26.1.337. 34. Swindle M. Comparative anatomy of the pig [technical bulletin]. Columbia, MO: Sinclair Research; 2002; June 15, 2012. 35. McNitt-Gray M. Tradeoffs in CT image quality and dose. Med Phys. 2006;33(6):2154-2162. 36. Rusinek H, Naidich D, McGuinness G, et al. Pulmonary nodule detection: low dose versus conventional CT. Radiology. 1998;209(1):243-249. 37. Cohnen M, Fischer H, Hamacher J, Lins E, Kötter R, Mödder U. CT of the head by use of reduced current and kilovoltage: relationship between image quality and dose reduction. Am J Neuroradiol. 2000;21(9):1654-1660. 38. Kalra MK, Prasad S, Saini S, et al. Clinical comparison of standard-dose and 50% reduced-dose abdominal CT: effect on image quality. Am J Radiol. 2002;179(5):1101-1106. 39. Mullins M, Lev M, Bove P, et al. Comparison of image quality between conventional and low dose enhanced head CT. Am J Neuroradiol. 2004;25(4):533-538. 40. Dose reduction in CT while maintaining diagnostic confidence: a feasibility/demonstration study. International Atomic Energy Agency Web site. http://www-pub.iaea .org/books/IAEABooks/8198/Dose-Reduction-in-CT -while-Maintaining-Diagnostic-Confidence-A-Feasibility -Demonstration-Study. Published 2009. 41. Sigal-Cinqualbre AB, Hennequin R, Abada HT, Chen X, Paul JF. Low-kilovoltage multi-detector row chest CT in adults: feasibility and effect of image quality and iodine dose. Radiology. 2004;231(1):169-174. 42. Primak AN, McCollough CH, Bruesewitz MR, Zhang J, Fletcher JG. Relationship between noise, dose, and pitch in cardiac multidetector row CT. Radiographics. 2006;26(6):1785-1794. 43. Swensson RG, King JL, Good WF, Gur D. Observer variation and the performance accuracy gained by averaging ratings of abnormality. Med Phys. 2000;27(8):1920-1933. 44. Receiver operator characteristic analysis in medical imaging: Report 79. J ICRU. 2008;8(1). doi:10.1093/jicru/ndn003. 45. Leitz WK, Månsson LG, Hedberg-Vikstrom BR, Kheddache S. In search of optimum chest radiography techniques. Br J Roentgenol. 1993;66(784):314-321. 46. Shiotani S, Kohno M, Ohashi N, et al. Non-traumatic postmortem computed tomographic (PMCT) findings of the lung. Forensic Sci Int. 2004;139(1):39-48. 136

47. Chew FS, Releya-Chew A, Ochoa ER Jr. Postmortem computed tomography of cadavers embalmed for use in teaching gross anatomy. J Comput Assist Tomogr. 2006;30(6):949-954.
