Differentiation of High Lipid Content From Low Lipid Content Adrenal Lesions Using Single-Source Rapid Kilovolt (Peak)-Switching Dual-Energy Multidetector CT : Journal of Computer Assisted Tomography

Secondary Logo

Journal Logo

Abdominal CT Techniques and Dose Optimization

Differentiation of High Lipid Content From Low Lipid Content Adrenal Lesions Using Single-Source Rapid Kilovolt (Peak)-Switching Dual-Energy Multidetector CT

Morgan, Desiree E. MD; Weber, Adam C. MD; Lockhart, Mark E. MD, MPH; Weber, Therese M. MD; Fineberg, Naomi S. PhD; Berland, Lincoln L. MD

Author Information
Journal of Computer Assisted Tomography 37(6):p 937-943, November/December 2013. | DOI: 10.1097/RCT.0b013e3182aaf996
  • Free

Abstract

An incidental adrenal lesion is an unsuspected and asymptomatic mass, usually detected on abdominal computed tomography (CT) obtained for other purposes1 and is a common finding in patients undergoing abdominal imaging, with a prevalence of more than 4% on imaging and 8.7% on autopsy.2 When discovered on imaging tests, radiologists guide evaluation of these lesions, whether benign or malignant.1,3,4 Most incidental adrenal lesions are benign adenomas.5,6 Current CT attenuation diagnostic criteria for an adrenal adenoma are based on the lesion having high intracellular fat content that can be accurately identified on an unenhanced CT scan.7 If an adrenal nodule demonstrates an attenuation value below 10 HU, the lesion is considered benign with 71% sensitivity and 98% specificity.8 When an abdominal CT scan is obtained in a routine postcontrast manner in patients who undergo abdominal CT for indications other than adrenal disease,1,9,10 it can be difficult to determine the diagnosis for an incidental adrenal lesion because several types of lesions have similar appearances. In addition, up to 40% of adrenal adenomas have low fat content11 and cannot be distinguished from potentially malignant lesions on either unenhanced or enhanced CT, unless a washout series is obtained at 15 minutes.11,12 Although Song et al13 showed that most incidentally discovered adrenal lesions are benign even when indeterminate on the initial CT scan, in clinical practice, a patient with a lesion that does not satisfy CT criteria for benign adenoma usually undergoes additional imaging with CT or magnetic resonance imaging, with or without laboratory testing, and may eventually go on to have a biopsy.1,9,10 Therefore, it would be useful to have the ability to characterize a lesion as a benign adenoma on the initial postcontrast examination.

Dual-energy CT provides a possible method to accurately characterize adrenal nodules, using only a single postcontrast series on the initial imaging encounter. By using low and high energies in a single scan, the constituents of an imaged volume can be deduced through material decomposition based on the difference in absorption characteristics for different elements between the 2 energies.14 With this method, the water or fat density material decomposition image contains no image voxels with attenuation similar to iodine and hence may serve as a “virtual unenhanced” image, obviating the need for a dedicated noncontrast CT for adrenal lesion characterization and reducing the patient’s risk from additional radiation exposure. Previous literature has shown the efficacy of dual-source dual-energy CT in the diagnosis of AI compared to regular CT.15–17 Unlike dual-source dual-energy CT, virtual unenhanced material decomposition images acquired using single-source rapid kilovolt (peak)-switching dual-energy (RSDE) CT provide densities in units of milligrams per milliliter rather than Hounsfield units. This represents a fundamental difference between the 2 existing commercially available dual-energy CT scanners. Therefore, it is expected that diagnostic criteria between DSDE and RSDE would differ. Unlike dual-source dual-energy CT, in which virtual unenhanced images are generated from blended image data (image space processing) and density is measured in familiar Hounsfield units, the virtual unenhanced material density images on single-source rapid kilovolt (peak)-switching dual-energy CT are generated from detector data (projection or data space processing) and density is not measured in Hounsfield units but in milligrams per milliliter of the material, that is, water in the case of water(-iodine) basis pair images or fat in the case of fat(-iodine) pair. If any AI could be correctly characterized as benign with these images, additional imaging or clinical investigations could be avoided. The goal of this intrapatient comparison study is to assess the efficacy of RSDE parameters, including characterizing adrenal lesions as high lipid content (HLC) or low lipid content (LLC) and providing guidance for using rapid-switching single-source dual-energy CT in the diagnosis of AI, based on receiver operating characteristic (ROC)-derived thresholds for adenoma characterization.

METHODS

This retrospective study protocol was reviewed and approved by the University of Alabama at Birmingham institutional review board, with a waiver for written informed consent. Strict HIPAA compliance was observed.

Subject Selection

Subjects were identified from a consecutive set of all outpatients who underwent multiphasic abdominal CT examinations on a single-source dual-energy scanner to evaluate known or suspected hepatic or pancreatic disease during a 6-month period. A search of the reports of these examinations from the radiology information system identified adrenal abnormalities including nodules, masses, and hyperplasia, which constituted the study group. Subjects for whom the dual-energy data were not properly stored were excluded. Lesions were diagnosed using the previously published guidelines for characterization based on Hounsfield unit attenuation measurements on unenhanced CT and washout on delayed contrasted CT images, presented in Table 1.1,8,12,18 In addition, “high lipid content” lesions were defined as those which demonstrated an attenuation value below 10 HU on the conventional unenhanced (CU) series; “low lipid content” lesions had Hounsfield unit greater than or equal to 10. An investigator not performing the dual-energy lesion measurements used all available prior scans combined with clinical information to determine the diagnosis. Hyperplasia was defined as maximum width perpendicular to long axis of the gland being over 2 standard deviations above the mean (10.1 mm for right and 12.1 mm for left).19 None of the patients included in the population had previous histologic evaluation performed on adrenal tissue. Chart review was performed to assess stability of nodules in follow-up.

T1-14
TABLE 1:
Clinical Diagnosis of Incidental Adrenal Lesions

MDCT Data Acquisition and Postprocessing

All subjects were scanned using a standardized abdominal (liver or pancreas) multiphasic protocol (Table 2) on a Discovery CT750 HD scanner (General Electric Healthcare, Waukesha, WI). Pancreatic parenchymal phase or hepatic late arterial phase images were acquired using the dual-energy technique, and standard unenhanced and portal venous phase images were obtained using conventional polychromatic beam MDCT technique. All patients received 720-mL oral contrast, and weight-based IV contrast dosage [eg, a standard 82-kg individual received 42 g I = 115-mL Isovue 370 (Bracco, Inc, Milan, Italy)] injected at 2.8 to 4.0 mL/s to achieve a 30-second fixed injection duration. Hepatic late arterial phase images were acquired at 30 seconds (for hepatic indications), pancreatic parenchymal phase images were acquired at 35 seconds after the initiation of injection (for pancreatic indications), and portal venous phase images at 60 seconds after completion of the injection. All images were acquired at 0.625 mm, reconstructed at 2.5 and 5 mm, respectively, for routine Picture Archiving and Communication System (PACS) viewing. Dual-energy images were evaluated on an independent workstation for the dual-energy scanner, the Gemstone Spectral Image (GSI) Viewer, ADW 4.5 (General Electric Healthcare, Milwaukee, WI). A single reader made all measurements on the PACS and the GSI viewer. Radiation exposure information was collected from the scanner as a single cumulative dose length product (DLP) for both the conventional and dual-energy components of the examinations (Figs. 1 and 2).

T2-14
TABLE 2:
Rapid-Switching Dual-Energy Multiphasic Abdominal CT Examination and Postprocessing Technical Parameters
F1-14
FIGURE 1:
Adrenal adenoma, typical lipid rich or HLC, in a 61-year-old man with history of pancreatic cancer. A, CU image ROI value is − 3.47 HU; (B) RSDE 70 keV postcontrast “PACS equivalent viewing energy” density is 48 HU; (C) RSDE 140 keV density is 3 “HU”; (D) RSDE water value is 991 mg/mL water; (E) RSDE fat value is 984 mg/mL fat. Note that images B-E are all derived from single arterial phase postcontrast image. [“PACS equivalent” refers to standard monoenergetic RSDE viewing energy that is similar in appearance to polychromatic beam CT appearance at 120 kV(p)].
F2-14
FIGURE 2:
Adrenal metastasis in a 58-year-old man with hepatocellular carcinoma. A, CU image ROI value is 51 HU; (B) RSDE 70 keV postcontrast “PACS equivalent viewing energy” density is 106 HU; (C) RSDE 140 keV density is 49 “HU”; (D) RSDE water value is 1034 mg/mL water; (E) RSDE fat value is 1026 mg/mL fat. Note that images B-E are all derived from single arterial phase postcontrast image. [“PACS equivalent” refers to standard monoenergetic RSDE viewing energy that is similar in appearance to polychromatic beam CT appearance at 120 kV(p)].

Image Data Collection

Lesions were measured on PACS using CU 5.0-mm images to obtain values for the single largest dimension for discrete lesions and maximum width perpendicular to the long axis of the adrenal body for hyperplasia diagnoses. Attenuation measurements were made at a level corresponding to the largest cross-sectional area of the adrenal lesion, placing the region of interest (ROI) in the largest area of the lesion while avoiding calcifications, necrosis, and margins. Virtual unenhanced arterial phase images were then analyzed on the independent workstation to obtain identical anatomic location density values for water minus iodine and fat minus iodine material decomposition basis-pair images, and for Hounsfield unit on the simulated monoenergetic 140-keV image. At a viewing energy of 140 keV, the effects of iodine on the image are minimized, as this energy is farthest on the RSDE spectrum from the k edge of iodine. Just as for PACS measurements, ROIs on the GSI viewer were selected to include maximum area and homogeneity. This method was applied to 2.5- to 5-mm axial reconstructions of the source (0.625 mm) dual-energy images. For all ROIs, mean value and standard deviation were recorded. All RSDE measurements were made by a single investigator.

Statistical Analysis

The different lesion diagnoses were compared using analysis of variance (ANOVA) to determine statistical significance of differences in mean reference values based on CU Hounsfield unit. Multiple RSDE variables were compared for all lesions types using ANOVA and Tukey HSD. The RSDE lesion density values were correlated with unenhanced Hounsfield unit using Pearson coefficient. Myelolipomas, which present no diagnostic challenge, were excluded from ROC curve analysis. The ROC analysis was used to determine a clinically relevant threshold for single-source rapid-switching dual-energy density values equivalent to 10 HU on unenhanced conventional MDCT. Significance was defined as P < 0.05.

RESULTS

Forty-seven adrenal lesions were identified in 18 women and 22 men, with a mean age of 66.5 years (range, 34–81 years), average weight 192 lb (range, 120–323 lb). There were 29 left- and 18 right-sided lesions; mean size, 2.5 cm (range, 0.7–15 cm). Lesions were bilateral in 7 subjects. There were 29 HLC and 18 LLC lesions identified during the 6-month interval: 12 adenomas; 4 lipid-poor adenomas; 8 malignant nodules (1 adrenal cortical carcinoma and 7 metastases); 7 myelolipomas; 6 hyperplasia; and 10 adenomatous hyperplasia. Statistically significant differences of the mean lesion Hounsfield unit on CU CT were present (P < 0.001, ANOVA), and on Tukey HSD comparison: myelolipoma attenuation less than adenoma equals adenomatous hyperplasia less than hyperplasia equals lipid-poor adenoma less than malignant nodule. Comparison of mean lesion RSDE material density across lesion groups also demonstrated statistically significant differences among different lesion types (Table 3).

T3-14
TABLE 3:
Adrenal Lesion RSDE CT Variables [Mean (SD)]

All single-source rapid-switching dual-energy measured variables had a significant correlation with CU CT Hounsfield unit attenuation values, with Pearson coefficient, r, ranging from 0.90 to 0.92 (P < 0.001) (Table 4). The attenuation in Hounsfield unit at 140 keV had the greatest correlation (r = 0.946). Virtual unenhanced milligrams per milliliter measurements had r = 0.933 for fat(-iodine) and 0.929 for water(-iodine) material decomposition basis-pair. Tukey analysis of all RSDE measurements showed significant discrimination of all lesions except for differentiation of lipid-poor adenomas from malignant lesions. Lipid-poor adenomas and malignant lesions had significantly greater density values than other lesion types (Table 3) (P < 0.001) but the dual-energy variables did not discriminate between these 2 lesion types. Receiver operating characteristic curve areas for Hounsfield unit 140-keV images, fat(-iodine), and water(-iodine) were 0.929 (0.039), 0.917 (0.046), and 0.912 (0.048), respectively (P < 0.001). When values were selected to obtain 94.4% specificity, 64% of HLC lesions had Hounsfield unit 140 keV values less than 9.5 HU, 59% had fat(-iodine) values less than 987 mg/mL, and 50% had water(-iodine) values less than 994 mg/mL.

The mean DLP for the total multiphasic abdominal CT examination (some of which also included coverage of the chest) was 1661 mGy-cm (range, 868–3427 mGy-cm) and for the dual-energy component DLP was 662 mGy-cm (range, 492–895 mGy-cm).

T4-14
TABLE 4:
Pearson Correlation With Unenhanced Conventional CT HU Reference Standard

No imaging or clinical follow-up was available for 9 lesions, including 2 metastases and 3 myelolipomas with typical macroscopic fat imaging characteristics. Two patients with metastases had histologic confirmation after the RSDE CT. In the 33 benign incidental adrenal lesions for which imaging or clinical follow-up was available at a mean imaging follow-up of 15.6 months (range, 5–24 months), no significant interval growth was found and no metastases developed.

DISCUSSION

Incidental adrenal lesions are commonly found on contrast-enhanced CT examinations performed for other indications.1 If unenhanced images are present, there are defined attenuation criteria to differentiate types of adrenal lesions. However, even if CU images are not acquired, it may still be possible to characterize a lesion if the study was performed using a dual-energy technique, as has been shown using modern dual-source dual-energy scanners.16,17 Our objective was to determine if clinically relevant material rapid-switching dual-energy CT density thresholds in milligrams per milliliter water could substitute for the 10-HU threshold on conventional MDCT. If so, this might obviate the need for additional imaging tests, avoiding extra costs in health care expenditure and radiation exposure to patients. Although the 94% specificity for each of the RSDE variables measured in our study is less than the 98% specificity for the widely used CU Hounsfield unit threshold of 10,8 any successful characterization of a lesion as benign represents an incremental gain in avoidance of further workup. On the basis of a threshold specificity of 94% in this study, sensitivity for confirmation of a benign lesion based on HLC using the postcontrast RSDE variable of 140 keV HU, fat(-iodine) mg/mL and water(-iodine) mg/mL is 64%, 59%, and 50%, respectively.

There are substantial technical differences between the “virtual” unenhanced images that are generated with dual-source versus single-source dual-energy CT scanners. Both represent post–IV-contrast injection images where the iodine has been “subtracted” through the material decomposition process. On the dual-source system, because virtual unenhanced images are created in “image space” the measurement of density is in Hounsfield units. However, because the single-source rapid-switching system creates virtual unenhanced images in “projection space,” the density is measured in terms of milligrams per milliliter of the material being evaluated. This presents a problem, a limitation of RSDE compared to DSDE for characterizing incidentally discovered adrenal lesions, because the validated 10-HU threshold to characterize a lesion as benign cannot currently be applied to RSDE virtual unenhanced images. Although Hounsfield unit measured on the 140-keV images had the greatest linear correlation with CU Hounsfield unit, it should be stressed that these “pseudounenhanced” images do still contain minimal effects iodine. Hence, the Hounsfield unit measured at 140 keV are different from CU Hounsfield unit values, and cannot be compared in the method of Ho et al.17 A new multimaterial postprocessing software for RSDE, termed GSIVUE by the manufacturer (General Electric Healthcare, Waukesha, WI) and available after March 2013 may overcome this hurdle but needs further evaluation before being applied to clinical practice. This software, applied to postcontrast images, uses a multimaterial decomposition to quantify then replace iodine fractional values with blood, creating a monoenergetic image on which Hounsfield units are measurable. Although several groups (including ours) are currently evaluating the use of this software, there are no published data yet on this recent development. It should also be appreciated that the software that provides the quantitation of material content is also in evolution. This may lead to changes in material content values as the software changes. Therefore, although the values we have derived may change slightly as new software becomes validated, the principle of the use of RSDE for this application is not affected.

The few published reports of incidental adrenal lesion characterization using the derived virtual unenhanced images have used the dual-source dual-energy technique. Gnannt et al16 showed good accuracy compared to traditional unenhanced CT in a multireader study evaluating a population of 42 patients with 51 lesions. Using the patient’s CU CT as a reference standard, the sensitivity, specificity, and accuracy for virtual unenhanced images for classifying a lesion as probably benign were 76%, 82%, and 78% for reader 1 and 79%, 95%, and 86% for reader 2 when lesions of all size were considered. For lesions 1 cm or greater, these values improved to 95%, 100%, 97%, and 91%, 100%, 95%, respectively. Importantly, there was a high correlation between the virtual unenhanced Hounsfield unit and CU Hounsfield unit, with no significant difference in the mean Hounsfield unit values on virtual unenhanced compared to CU images, and excellent interobserver agreement with the measurements. Similar findings were confirmed by Ho et al17 in a smaller group of 19 patients with 23 incidentally discovered nodules of multiple etiologies, also using dual-source dual-energy techniques. In that study, the authors reported no statistically significant difference in the mean Hounsfield unit measurements of adenomas [10.3 (13.1) on virtual unenhanced vs 8.9 (10.4) on CU images], or metastases [35.7 (6.0) on virtual unenhanced vs 32.6 (6.1) on CU images]. However, on average for the 3 readers, the virtual unenhanced images were 1.8 (1.7) HU higher than CU images, and virtual unenhanced images classified adrenal adenomas as greater than 10 HU, whereas CU images classified the same nodule as less than or equal to 10 HU in 13% (3/23), 4% (1/23), and 9% (2/23) for readers 1, 2, and 3, respectively. Importantly, no malignant nodules were misclassified.

There is one study evaluating dual-source dual-energy CT in a population of patients with known adrenal lesions. Compared to the intrapatient reference standard of washout assessment, Kim et al20 found diminished sensitivity for detection of lipid-rich adenomas on both the “early” (1-minute postcontrast) and “late” (15-minute postcontrast) virtual unenhanced images, due to insufficient iodine subtraction using the 3-point material decomposition algorithm on dual-source dual-energy CT. The authors suggested that CU acquisitions could not be eliminated in efforts to reduce overall radiation exposure for their dedicated multiphasic adrenal dual-energy assessment. However, in the context of an incidentally discovered adrenal lesion, had there been no CU CT, the authors would have been able to identify 7 of 18 lipid-rich adenomas on the early and 11 of 18 lipid-rich adenomas on the delayed images. In other words, these patients would have needed no additional imaging to confirm a benign lipid-rich adenoma, saving both radiation and health care costs.

Our study shows strong correlation between each of the material density variables measured on the single-source rapid-switching dual-energy images with traditional unenhanced Hounsfield unit. In addition, each RSDE variable revealed consistently lower means for myelolipoma less than adenoma less than malignant nodule equals lipid-poor adenoma on Tukey HSD test. However, it is easier to conceptualize the correlation between Hounsfield unit on virtual unenhanced dual-source dual-energy images and Hounsfield unit on CU images. Although direct comparison of virtual unenhanced water(-iodine) and fat(-iodine) basis pairs in milligrams per milliliter values in terms of Hounsfield unit is not possible, we believe our findings support using material decomposition with single-source rapid-switching dual-energy as a substitute for CU images for characterizing incidentally discovered adrenal lesions as benign on postcontrast images. With this method, further imaging can be avoided, and the use of this method to measure milligrams per milliliter fat or water should fill a diagnostic gap for now. In the future, Hounsfield unit values may be calculable on the GSIVUE images available on RSDE, and that method will need to be validated independently.

Limitations of this study include the retrospective observational methods, but consecutive cases of patients with any adrenal lesion during the 6-month period were included to minimize selection bias. We included instances of adrenal hyperplasias, as our objective was to identify high and low fat content within adrenal lesions to correlate the quantitative material density information with conventional Hounsfield unit values. The prevalence of malignant adrenal lesions was higher than expected for incidental lesions, reflecting the clinical practice at our institution to place patients requiring multiphasic liver or pancreas CT examinations on the dual-energy unit, and the pretest probabilities of metastases were likely greater in these 2 populations than in the general population. There was no uniform pathological proof; however, similar imaging and clinical parameters to classify adrenal nodules have been used by other authors. Our results reflect one method of applying dual-energy CT to abdominal scanning, and are not directly applicable to other methods of dual-energy CT, as mentioned earlier in the discussion. The ROC analysis and thresholds are based on a small population, and will need to be verified in a larger population, preferably in a multi-institutional environment.

In conclusion, there is a strong correlation between single-source rapid-switching dual-energy CT material density values in milligrams per milliliter, 140-keV simulated monoenergetic Hounsfield unit, and CU attenuation values in Hounsfield unit for HLC and LLC adrenal lesions, with clinically relevant thresholds sufficient to differentiate between the 2 classes of lesions. The ability to characterize most of the adenomatous lesions on postcontrast material decomposition single-source rapid-switching dual-energy CT images suggests that adrenal nodules in some patients may be satisfactorily characterized on postcontrast RSDE examinations, eliminating the need in such patients for further evaluation with CU or adrenal protocol CT or magnetic resonance imaging.

REFERENCES

1. , , , ACR Appropriateness Criteria on Incidentally Discovered Adrenal Mass. [http://guidelines.gov/content.aspx?id=37940]. Reston, VA: American College of Radiology (ACR); 2012.
2. , . Adrenal incidentaloma: evaluation and management. J Clin Pathol. 2008; 61: 1168–1173.
3. , , , Adrenal imaging: a comprehensive review. Radiol Clin North Am. 2012; 50: 219–243.
4. , , , Incidental adrenal lesions: principles, techniques, and algorithms for imaging characterization. Radiology. 2008; 249: 756–775.
5. , , , The clinical significance of adrenal incidentalomas. Eur J Clin Invest. 2011; 41: 552–560.
6. , , , A survey on adrenal incidentaloma in Italy. J Clin Endocrinol Metab. 2000; 85: 637–644.
7. , , , Benign and malignant adrenal masses: CT distinction with attenuation coefficients, size, and observer analysis. Radiology. 1991; 179: 415–418.
8. , , , Characterization of adrenal masses using unenhanced CT: an analysis of the CT literature. AJR Am J Roentgenol. 1998; 171: 201–204.
9. . Approach to the patient with an adrenal incidentaloma. J Clin Endocrinol Metab. 2010; 95: 4106–4113.
10. , , , Clinical utility of noncontrast computed tomography attenuation value (Hounsfield units) to differentiate adrenal adenomas/hyperplasias from nonadenomas: Cleveland Clinic experience. J Clin Endocrinol Metab. 2005; 90: 871–877.
11. , , , Characterization of indeterminate (lipid-poor) adrenal masses: use of washout characteristics at contrast-enhanced CT. Radiology. 2000; 217: 798–802.
12. , , , CT time-attenuation washout curves of adrenal adenomas and nonadenomas. AJR Am J Roentgenol. 1998; 170: 747–752.
13. , , . The incidental indeterminate adrenal mass on CT (> 10 H) in patients without cancer: is further imaging necessary? Follow-up of 321 consecutive indeterminate adrenal masses. AJR Am J Roentgenol. 2007; 189: 1119–1123.
14. . Gemstone Spectral Imaging. CT and X-ray Systems Lab, GE Global Research; 2008. GE Healthcare.
15. , , , Dual-energy CT for characterization of adrenal nodules: initial experience. AJR Am J Roentgenol. 2010; 194: 1479–1483.
16. , , , Dual-energy CT for characterization of the incidental adrenal mass: preliminary observations. AJR Am J Roentgenol. 2012; 198: 138–144.
17. , , , Characterization of adrenal nodules with dual-energy CT: can virtual unenhanced attenuation values replace true unenhanced attenuation values? AJR Am J Roentgenol. 2012; 198: 840–845.
18. , , , State-of-the-art adrenal imaging. Radiographics. 2001; 21: 995–1012.
19. , , , Clin Radiol. 1994; 49: 453–455.
20. , , , Adenoma characterization: adrenal protocol with dual-energy CT. Radiology. 2013; 267: 155–163.
Keywords:

dual-energy CT; adrenal; adenoma; incidental; rapid kilovolt switching; material density

Copyright © 2013 Wolters Kluwer Health, Inc. All rights reserved.