Discriminating Parathyroid Adenoma from Local Mimics by Using Inherent Tissue Attenuation and Vascular Information Obtained with Four-Dimensional CT: Formulation of a Multinomial Logistic Regression Model

Published Online:https://doi.org/10.1148/radiol.13122851

Application of a three-covariate, multinomial logistic regression model could potentially improve the confidence and accuracy of localization of abnormal parathyroid tissue and thereby facilitate deployment of appropriate surgery.


To identify a set of parameters, which are based on tissue enhancement and native iodine content obtained from a standardized triple-phase four-dimensional (4D) computed tomographic (CT) scan, that define a multinomial logistic regression model that discriminates between parathyroid adenoma (PTA) and thyroid nodules or lymph nodes.

Materials and Methods

Informed consent was waived by the institutional review board for this retrospective HIPAA-compliant study. Electronic medical records were reviewed for 102 patients with hyperparathyroidism who underwent triple-phase 4D CT and parathyroid surgery resulting in pathologically proved removal of adenoma from July 2010 through December 2011. Hounsfield units were measured in PTA, thyroid, lymph nodes, and aorta and were used to determine seven parameters characterizing tissue contrast enhancement. These were used as covariates in 10 multinomial logistic regression models. Three models with one covariate, four models with two covariates, and three models with three covariates were investigated. Receiver operating characteristic (ROC) analysis was performed to determine how well each model discriminated between adenoma and nonadenomatous tissues. Statistical differences between the areas under the ROC curves (AUCs) for each model pair were calculated, as well as sensitivity, specificity, accuracy, negative predictive value, and positive predictive value.


A total of 120 lesions were found; 112 (93.3%) lesions were weighed, and mean and median weights were 589 and 335 mg, respectively. The three-covariate models were significantly identical (P > .65), with largest AUC of 0.9913 ± 0.0037 (standard error), accuracy of 96.9%, and sensitivity, specificity, negative predictive value, and positive predictive value of 94.3%, 98.3%, 97.1%, and 96.7%, respectively. The one- and two-covariate models were significantly less accurate (P < .043).


A three-covariate multinomial logistic model derived from a triple-phase 4D CT scan can accurately provide the probability that tissue is PTA and performs significantly better than models using one or two covariates.

© RSNA, 2013

Online supplemental material is available for this article.


  • 1. Hunter GJ, Schellingerhout D, Vu TH, Perrier ND, Hamberg LM. Accuracy of four-dimensional CT for the localization of abnormal parathyroid glands in patients with primary hyperparathyroidism. Radiology 2012;264(3):789–795.
  • 2. Lubitz CC, Hunter GJ, Hamberg LM, et al. Accuracy of 4-dimensional computed tomography in poorly localized patients with primary hyperparathyroidism. Surgery 2010;148(6):1129–1137; discussion 1137–1138.
  • 3. Cheung K, Wang TS, Farrokhyar F, Roman SA, Sosa JA. A meta-analysis of preoperative localization techniques for patients with primary hyperparathyroidism. Ann Surg Oncol 2012;19(2):577–583.
  • 4. Starker LF, Mahajan A, Björklund P, Sze G, Udelsman R, Carling T. 4D parathyroid CT as the initial localization study for patients with de novo primary hyperparathyroidism. Ann Surg Oncol 2011;18(6):1723–1728.
  • 5. Rodgers SE, Hunter GJ, Hamberg LM, et al. Improved preoperative planning for directed parathyroidectomy with 4-dimensional computed tomography. Surgery 2006;140(6):932–940; discussion 940–941.
  • 6. Kunstman JW, Udelsman R. Superiority of minimally invasive parathyroidectomy. Adv Surg 2012;46(1):171–189.
  • 7. Stucken EZ, Kutler DI, Moquete R, Kazam E, Kuhel WI. Localization of small parathyroid adenomas using modified 4-dimensional computed tomography/ultrasound. Otolaryngol Head Neck Surg 2012;146(1):33–39.
  • 8. Kleinbaum DG, Klein M. Logistic regression. 3rd ed. New York, NY: Springer, 2010; 429–462.
  • 9. Obuchowski NA. ROC analysis. AJR Am J Roentgenol 2005;184(2):364–372.
  • 10. Obuchowski NA. Receiver operating characteristic curves and their use in radiology. Radiology 2003;229(1):3–8.
  • 11. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44(3):837–845.
  • 12. Gafton AR, Glastonbury CM, Eastwood JD, Hoang JK. Parathyroid lesions: characterization with dual-phase arterial and venous enhanced CT of the neck. AJNR Am J Neuroradiol 2012;33(5):949–952.
  • 13. Welling RD, Olson JA Jr, Kranz PG, Eastwood JD, Hoang JK. Bilateral retropharyngeal parathyroid hyperplasia detected with 4D multidetector row CT. AJNR Am J Neuroradiol 2011;32(5):E80–E82.
  • 14. Kutler DI, Moquete R, Kazam E, Kuhel WI. Parathyroid localization with modified 4D-computed tomography and ultrasonography for patients with primary hyperparathyroidism. Laryngoscope 2011;121(6):1219–1224.
  • 15. Sillery JC, DeLone DR, Welker KM. Cystic parathyroid adenomas on dynamic CT. AJNR Am J Neuroradiol 2011;32(6):E107–E109.

Article History

Received December 23, 2012; revision requested February 6, 2013; final revision received March 29; accepted April 16; final version accepted June 27.
Published online: Jan 2014
Published in print: Jan 2014