Data Analysis: Evaluation of nanoscale contrast agent enhanced CT scan to differentiate between benign and malignant lung cancer in mouse model

Robert C. Bell, Deevakar Rogith, [...], and Ketankumar Ghaghada

Additional article information

Abstract

Proposed is a method for statistical analysis for a small sample size, repeated measure experiment with nesting factors. In the original experiment the Student t-test was used for analysis. Using the same data, we modeled the experiment into two groups of mice with benign and malignant primary lung tumors. 4 tumor nodules were selected from each mouse (N= 36). The dependent variables are the volume, diameter, and signal attenuation measured using computed tomography (CT). The measurements are made before injecting the contrast and at 0, 72, and 168 hours after injection. The contrast agent enhances tumor nodule volume and volume differences between benign and malignant tumor nodules measured across time (p < 0.05). The signal attenuation measured across time differentiates between benign and malignant groups (p < 0.05). There is significant correlation between rate of change of volume and diameter of tumor. The advantages of this statistical method are discussed.

Introduction and Background

In the United States lung cancers have an incidence of 62 per 100,000 population and death rate of 52 per 100,000 population1. It amounts to one-fourth of all cancer deaths in the country1. Mean 5-year survival rate is 15%1. Diagnosis of lung cancer at an early stage is vital for most patients with lung cancer present at advanced stages1. There is little evidence to substantiate population screening for lung cancer using Chest X-Ray, low dose Computed Tomography (CT) and sputum cytology2–4. These tests have shown to have high false positive rate about 95% in screening population and so tissue diagnosis may be needed to confirm the diagnosis5. In comparison with plain CT, contrast enhanced CT has been found useful in detection of small cell carcinoma type of lung cancer, which usually presents in malignant spectrum and with mediastinal lymph nodes6–8

To differentiate benign and malignant lung tumors using CT scan, contrast agents are used9. To enhance the tumor tissues, a new nanoscale contrast agent (liposomal-iodinated) is being studied10. The contrast is systemically introduced. CT images are taken before injection of contrast and at 0 hours, 3 days (72 hours) and 7 days (168 hours) after injection. Depending on the uptake of the contrast, the size and vascularity of the lesion are computed. Rapid change of size and high vascularity differentiate a benign tumor from a malignant tumor11–13. The study is conducted in mouse models. Two groups of mice with benign and malignant primary lung tumor are used. The two mouse lung tumor models used in this study are Kras alone (KrasLA1) or in combination with p53 (LSL-KrasG12D;p53FL/FL) which represent the benign and malignant groups, respectively14, 15.

Hypothesis

A dynamic CT scan (at times: pre-scan, 0 days, 3 days, and 7 days) after injection of nanoscale contrast agent can be used to differentiate between benign and malignant lung nodules in a mouse model using the characteristics: nodule diameter, nodule volume, change in nodule volume over time (rate of growth, ΔV) and signal a (ΔHU) of the nodules.

Research Questions

  1. Is there a difference between volume of tumor in benign and malignant groups, measured before (pre) and at 0, 72 and 168 hours after the injection of contrast?
  2. Is there a difference between signal attenuation from the nodules in benign and malignant groups, measured before (pre) and at 0, 72 and 168 hours after the injection of contrast? Does this agree with the results from the original analysis?
  3. Is there a relationship between rate of change of volume (ΔV) and diameter in each of the two groups?

Methodology

Only the raw data from the original experiment ([9] Badea et al, 2011) was used for the analysis. At the time of analyzing the data from 9 mice were available to us - four mice from benign group and five mice from malignant group. A total of 36 tumor nodules, selected from 9 mice with 4 nodules randomly selected from each mouse to achieve a balanced design. Random selection of the nodules helps to control for variability in the tumors, but will not eliminate it. The mouse was injected with the liposomal-iodinated contrast agent under study. The diameter and volume measurements were made independently using X-ray Computed Tomography and analyzed using OsiriX (v 3.6, 64 bit). From the diameter measurement the nodules were classified into size groups. The measurements are made before (pre) and after injection (0hr) of contrast, 72hr and 168hrs after injection of dye.

Design:

Three-factor (type of tumor, mouse, and nodule size group) nested design having tumor size group crossed with type of tumor, mice nested within type of tumor, and tumor nodules as subjects. The dependent variables are the size of tumor (diameter, mm), volume of tumor (mm3) and signal attenuation detected by CT in Hounsfield units (ΔHU) measured before (pre) and after injection (0hr) of contrast, 72hr and 168hrs after injection of dye. Mouse is the nested factor. [Figure 1]

Statistical Analysis:

For research questions 1and 2: Multivariate repeated measures design with mouse as a nested random factor and type of tumor and size groups as fixed between subjects factor and measurement at times as within-subjects factor was used. For research question 3: linear regression analysis was used. All Analysis was done using IBM SPSS Statistics version 19, General Linear Model and Regression procedures.

Procedure:

Type of Tumor has 2 levels: Benign and Malignant. The Tumor size groups have 4 levels: (1) <1mm, (2) 1 - 1.5mm, (3) 1.5 - 2.5mm, (4) > 2.5mm. Within each mouse, tumor nodules are randomly selected and measurements made (size, volume, and signal attenuation). Within subject-factors are sizes of tumor measured before (pre) and at 0, 72, and 168 hours after injection of dye. The rate of change in volume is calculated.

To determine if there is any relationship between initial measurement of size of tumor and rate of change of volume (dependent variables), against type of tumor and size groups (independent variables), linear regression analysis is performed. This is possible because the diameter and volume measurement are done independently and are not simply being calculated from the other. To determine if there is a difference in signal attenuation between benign and malignant groups, the multivariate repeated measures analysis is done. The levels for the repeated measures within-subject factor will be measurement at 0, 72 and 168 hours. This will determine whether significant difference in signal attenuation differentiate between benign and malignant groups. Similarly, the multivariate repeated measures analysis for volume of tumor measured before and after injection of dye will answer if the contrast enhances the volume of the tumor for detection.

Results

Difference in volume:

After contrast injection, the volume differences between the benign and malignant tumor groups increased, exhibiting a significant (F(1,23) = 5.49, p = 0.028, power = 0.612) linear trend across time (Figure 2). Tumor volumes within the group increased, exhibiting a significant (F(1,23)= 27.80, p < 0.001, power = 0.99) linear trend across time (Figure 2). The contrast enhanced volumes of both types of tumors. The malignant group showed a greater increase in volume over time than the benign group. No other statistically significant (α = .05) differences were obtained in this step.

Difference in signal attenuation measured:

After Contrast Injection, the tumor signal attenuation differences between benign and malignant groups increased, exhibiting a significant linear trend (F(1,23) = 50.855, p < .001, power = 1.0) across time and quadratic trend (F(1,23) = 15.589, p = .001, power = .966) across time. This shows that contrast is able to differentiate between benign and malignant groups over time [Figure 3].

The signal attenuation measured within each group also exhibits a significant quadratic trend (F(1,23)= 92.55, p < .001, power = 1.0) and cubic trend ( F(1,23)= 233.89, p < .001, power = 1.0) across time [Figure 3].

Effect of tumor type and tumor size on change of volume:

From Table 1, SPSS output of Regression analysis shows that Tumor type (B = 7.883, p = .001) and Tumor size (B = 0.681, p = .013) independently predict tumor volume change.

Discussion

The contrast agent enhances tumor nodule signal attenuation and enhances volume differences between benign and malignant tumor nodules measured across time (p < 0.05). The nanoparticle contrast agent enhances the tumor volume. The rate of change of volume is significantly related to type of tumor. The change in volume in benign group is less pronounced than in the malignant group. There is significant correlation between rate of change of volume and diameter of tumor.

The difference in signal attenuation (ΔHU) measured across time differentiates between benign and malignant groups (p < 0.05). The contrast agent extravagates into the tissue and because of the slow wash-out time of contrasts from malignant tissues16, 17, the signal attenuation of benign lesion peak early compared to malignant nodules. The signal attenuation is the measure of total signal of the nodule, which includes signal from blood vessels within the nodules and tumor tissues of the nodule less the signal measures before the contrast agent was given (pre-contrast scan). Thus it is also possible to study the vasculature of the tumor tissues, by measuring the blood volumes. Further study should be done in this area. Signal attenuation also implies the specificity of the contrast for the tumor type and thus can be used as targeted drug delivery systems.

Practical limitations include small sample size of the original experiment and the possibility of exclusion bias based on random selection of lung nodules. Only type of tumor and initial tumor size were considered, while change in tumor volume is affected by a number of factors such as age, sex, tumor cell type, initial tumor stage, nodule size, shape, signal attenuation and volume measurement techniques18–23. Though size of the tumor is used clinically in classifying or staging the tumor, the effect of change in size across time is not evaluated. One possible reason could be that change in tumor volume is easily detected compared to change in tumor size24.

Hierarchical linear mixed model or multilevel modeling analyses typically require large samples even for a few predictors25. In studies where the number of levels of the random higher level random factor, and number of experimental units is small, failure of convergence or Hessian matrix inversion may not infrequently preclude analyses employing linear mixed models or multilevel modeling. Or, in those cases the best fitting covariance structure may overly specifically fit the small sample so that outcomes lose the ability to generalize. In such cases, the multivariate repeated measures approach might be appropriate. The present research design includes fixed, random and nested factors.

Conclusion

The nanoparticle contrast agent is able to enhance tumor nodule volumes, volume differences between benign and malignant tumor nodules, and the signal attenuation measured is able to differentiate between benign and malignant types. To increase statistical power and generalizability for detection of contrast agent effects, further systematic replication and extension using animal studies with larger sample sizes are needed. Ultimately, large scale clinical trials would prove the utility of this novel contrast agent.

APPENDIX

Table 1:

Relation Between Volume Change and Tumor Size

Coefficientsa
ModelUnstandardized CoefficientsStandardized CoefficientstSignificance95.0% Confidence Interval for BCorrelationsCollinearity Statistics
BStandard ErrorBetaLower BoundUpper BoundZero-orderPartialPartToleranceVariance Inflation Factor
(Constant)−1.746.421−4.150.000−2.603−.890
Tumor Diameter in cm (Average of pre and hr_0)7.9932.212.4893.613.0013.49212.495.656.532.432.7801.283
Tumor Type.681.259.3562.632.013.1551.207.586.417.314.7801.283
aDependent Variable: Change in Volume

Figure 1:

An external file that holds a picture, illustration, etc. Object name is amia_2012_symp_0027f1.jpg

Illustration of Three Factor Nested Design.

Note: The figure is an illustration of the design (nesting) and does not represent the total mouse and nodules used in the study.

Figure 2:

Effect of volume measured across time.

a) Profile Plot

An external file that holds a picture, illustration, etc. Object name is amia_2012_symp_0027f2a.jpg

b) Volume Standard Error Plot:

An external file that holds a picture, illustration, etc. Object name is amia_2012_symp_0027f2b.jpg

Figure 3:

Effect of Signal Attenuation Measured across time.

a) Profile Plot

An external file that holds a picture, illustration, etc. Object name is amia_2012_symp_0027f3a.jpg

b) Signal Attenuation Standard Error Plot:

An external file that holds a picture, illustration, etc. Object name is amia_2012_symp_0027f3b.jpg

Article information

AMIA Annu Symp Proc. 2012; 2012: 27–35.
Published online 2012 Nov 3.
PMCID: PMC3540499
PMID: 23304269
Robert C. Bell,1,2 Deevakar Rogith,1 Craig W. Johnson,1 Cristian T. Badea,3 Khannan K. Athreya,4 Gabriela Espinosa,1,2 Darin Clark,3 A. Paiman Ghafoori,5 Yifan Li,5 David G. Kirsch,5,6 Ananth Annapragada,2 and Ketankumar Ghaghada2
1School of Biomedical Informatics, The University of Texas Health Sciences Center at Houston, Houston, TX 77030
2The Edward B. Singleton Department of Pediatric Radiology, Texas Children’s Hospital, Houston, TX 77030
3Center for In Vivo Microscopy, Duke University Medical Center, Durham, NC 27710
4University of Texas Medical School at Houston, The University of Texas Health Sciences Center at Houston, Houston, TX 77030
5Department of Radiation Oncology, Duke University Medical Center, Durham, NC 27710
6Department of Pharmacology and Cancer Biology, Duke University Medical Center, Durham, NC 27710
Copyright ©2012 AMIA - All rights reserved.
This is an Open Access article: verbatim copying and redistribution of this article are permitted in all media for any purpose
Articles from AMIA Annual Symposium Proceedings are provided here courtesy of American Medical Informatics Association

REFERENCES

1. Howlader N, Noone A, Krapcho M, Neyman N, Aminou R, Waldron W, et al. SEER Cancer Statistics Review, 1975–2008. Bethesda, MD: National Cancer Institute; 2011. Available from: http://seer.cancer.gov/csr/1975_2008/ [Google Scholar]
2. Humphrey LL, Teutsch S, Johnson M. Lung Cancer Screening with Sputum Cytologic Examination, Chest Radiography, and Computed Tomography: An Update for the U.S. Preventive Services Task Force. Annals of Internal Medicine. 2004;140(9):740–53. [PubMed] [Google Scholar]
3. Doria-Rose VP, Marcus PM, Szabo E, Tockman MS, Melamed MR, Prorok PC. Randomized controlled trials of the efficacy of lung cancer screening by sputum cytology revisited. Cancer. 2009;115(21):5007–17. [PMC free article] [PubMed] [Google Scholar]
4. Mazzone P, Mekhail T. Lung cancer screening. Current Oncology Reports. 2007;9(4):265–74. [PubMed] [Google Scholar]
5. Reduced Lung-Cancer Mortality with Low-Dose Computed Tomographic Screening. New England Journal of Medicine. 2011;365(5):395–409. [PMC free article] [PubMed] [Google Scholar]
6. Pfannenberg A, Aschoff P, Brechtel K, Müller M, Bares R, Paulsen F, et al. Low dose non-enhanced CT versus standard dose contrast-enhanced CT in combined PET/CT protocols for staging and therapy planning in non-small cell lung cancer. European Journal of Nuclear Medicine and Molecular Imaging. 2007;34(1):36–44. [PubMed] [Google Scholar]
7. Hollings N, Shaw P. Diagnostic imaging of lung cancer. European Respiratory Journal. 2002 Apr 1;19(4):722–42. 2002. [PubMed] [Google Scholar]
8. Cascade PN, Gross BH, Kazerooni EA, Quint LE, Francis IR, Strawderman M, et al. Variability in the detection of enlarged mediastinal lymph nodes in staging lung cancer: a comparison of contrast-enhanced and unenhanced CT. American Journal of Roentgenology. 1998 Apr 1;170(4):927–31. 1998. [PubMed] [Google Scholar]
9. Swensen SJ, Viggiano RW, Midthun DE, Müller NL, Sherrick A, Yamashita K, et al. Lung Nodule Enhancement at CT: Multicenter Study1. Radiology. 2000 Jan 1;214(1):73–80. 2000. [PubMed] [Google Scholar]
10. Badea CT, Athreya KK, Espinosa G, Clark D, Ghafoori AP, Li Y, et al. Computed Tomography Imaging of Primary Lung Cancer in Mice Using a Liposomal-Iodinated Contrast Agent. PLoS ONE. 2012;7(4):e34496. [PMC free article] [PubMed] [Google Scholar]
11. Zhang M, Kono M. Solitary pulmonary nodules: evaluation of blood flow patterns with dynamic CT. Radiology. 1997 Nov 1;205(2):471–8. 1997. [PubMed] [Google Scholar]
12. Yankelevitz DF, Reeves AP, Kostis WJ, Zhao B, Henschke CI. Small Pulmonary Nodules: Volumetrically Determined Growth Rates Based on CT Evaluation1. Radiology. 2000 Oct 1;217(1):251–6. 2000. [PubMed] [Google Scholar]
13. Swensen SJ, Jett JR, Hartman TE, Midthun DE, Mandrekar SJ, Hillman SL, et al. CT Screening for Lung Cancer: Five-year Prospective Experience1. Radiology. 2005 Apr 1;235(1):259–65. 2005. [PubMed] [Google Scholar]
14. Meuwissen R, Linn SC, van der Valk M, Mooi WJ, Berns A. Mouse model for lung tumorigenesis through Cre/lox controlled sporadic activation of the K-Ras oncogene. Oncogene. 2001;20(45):6551–8. [PubMed] [Google Scholar]
15. Kirsch DG, Grimm J, Guimaraes AR, Wojtkiewicz GR, Perez BA, Santiago PM, et al. Imaging primary lung cancers in mice to study radiation biology. International Journal of Radiation Oncology* Biology* Physics. 2010;76(4):973–7. [PMC free article] [PubMed] [Google Scholar]
16. Littleton JT, Durizch ML, Moeller G, Herbert DE. Pulmonary masses: contrast enhancement. Radiology. 1990 Dec 1;177(3):861–71. 1990. [PubMed] [Google Scholar]
17. Miles KA. Tumour angiogenesis and its relation to contrast enhancement on computed tomography: a review. European Journal of Radiology. 1999;30(3):198–205. [PubMed] [Google Scholar]
18. Usuda K, Saito Y, Sagawa M, Sato M, Kanma K, Takahashi S, et al. Tumor doubling time and prognostic assessment of patients with primary lung cancer. Cancer. 1994;74(8):2239–44. [PubMed] [Google Scholar]
19. Lavrenkov K, Partridge M, Cook G, Brada M. Positron emission tomography for target volume definition in the treatment of non-small cell lung cancer. Radiotherapy and Oncology. 2005;77(1):1–4. [PubMed] [Google Scholar]
20. Sone S, Tsushima K, Yoshida K, Hamanaka K, Hanaoka T, Kondo R. Pulmonary Nodules: Preliminary Experience with Semiautomated Volumetric Evaluation by CT Stratum. Academic Radiology. 2010;17(7):900–11. [PubMed] [Google Scholar]
21. Gavrielides MA, Kinnard LM, Myers KJ, Petrick N. Noncalcified Lung Nodules: Volumetric Assessment with Thoracic CT1. Radiology. 2009 Apr;251(1):26–37. 2009. [PMC free article] [PubMed] [Google Scholar]
22. Ravenel JG, Leue WM, Nietert PJ, Miller JV, Taylor KK, Silvestri GA. Pulmonary Nodule Volume: Effects of Reconstruction Parameters on Automated Measurements—A Phantom Study1. Radiology. 2008 May;247(2):400–8. 2008. [PMC free article] [PubMed] [Google Scholar]
23. Reeves AP, Chan AB, Yankelevitz DF, Henschke CI, Kressler B, Kostis WJ. On measuring the change in size of pulmonary nodules. Medical Imaging, IEEE Transactions on. 2006;25(4):435–50. [PubMed] [Google Scholar]
24. Plathow C, Schoebinger M, Fink C, Hof H, Debus J, Meinzer H-P, et al. Quantification of Lung Tumor Volume and Rotation at 3D Dynamic Parallel MR Imaging with View Sharing: Preliminary Results1. Radiology. 2006 Aug;240(2):537–45. 2006. [PubMed] [Google Scholar]
25. Tabchnick BG, Fidell LS. Multilevel Linear Modeling Using Multivariate Statistics. 5 ed. Boston (MA): Allyn and Bacon (Pearson); 2007. p. 787. [Google Scholar]