Soft tissue discrimination ex vivo by dual energy computed tomography

European Journal of Radiology 75 (2010) e124–e128

Contents lists available at ScienceDirect

European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad

Soft tissue discrimination ex vivo by dual energy computed tomography H. Zachrisson a,b,∗ , E. Engström b , J. Engvall a,b , L. Wigström a,b , Ö. Smedby a,c , A. Persson a,c a

Center for Medical Image Science and Visualization, Linköping University, Linköping University Hospital, SE-581 85 Linköping, Sweden Clinical Physiology, Department of Medical and Health Sciences (IMH), Linköping University, Linköping University Hospital, SE-581 85 Linköping, Sweden c Radiology, Department of Medical and Health Sciences (IMH), Linköping University, Linköping University Hospital, SE-581 85 Linköping, Sweden b

a r t i c l e

i n f o

Article history: Received 28 April 2009 Received in revised form 1 February 2010 Accepted 2 February 2010 Keywords: Dual-source computed tomography Dual energy CT imaging Soft tissue characterization Plaque

a b s t r a c t Purpose: Dual Energy Computed Tomography (DECT) may provide additional information about the chemical composition of tissues compared to examination with a single X-ray energy. The aim of this in vitro study was to test whether combining two energies may significantly improve the detection of soft tissue components commonly present in arterial plaques. Methods: Tissue samples of myocardial and psoas muscle, venous and arterial thrombus as well as fat from different locations were scanned using a SOMATOM Definition Dual Source CT system (Siemens AG, Medical Solutions, Forchheim, Germany) with simultaneous tube voltages of 140 and 80 kV. The attenuation (Hounsfield units, HU) at 80 and 140 kV was measured in representative regions of interest, and the association between measured HU values and tissue types was tested with logistic regression. Results: The combination of two energy levels (80 and 140 kV) significantly improved (p < 0.001) the ability to correctly classify venous thrombus vs arterial thrombus, myocardium or psoas; arterial thrombus vs myocardium or psoas; myocardium vs psoas; as well as the differentiation between fat tissue from various locations. Single energy alone was sufficient for distinguishing fat from other tissues. Conclusion: DECT offers significantly improved in vitro differentiation between soft tissues occurring in plaques. If this corresponds to better tissue discrimination in vivo needs to be clarified in future studies. © 2010 Elsevier Ireland Ltd. All rights reserved.

1. Introduction By conventional (single energy) Computed Tomography (CT), only three tissue types, can be differentiated in atherosclerotic plaque imaging including calcium, fat and mixed tissue [1]. For clinical studies of plaques, however, it is essential to discriminate between soft tissue types such as thrombus, collagen, fat, muscle fibres as well as calcifications, in order to assess the vulnerability of plaques. Traditional CT only produces one numeric value for each voxel. Dual-source CT scanners, however, allow simultaneous scanning at two peak X-ray energies which may open new diagnostic possibilities. To distinguish different tissues from each other, some physical property that differs between the tissues needs to be measured. Conventional attenuation measurements based on a single photon energy spectrum have limited value in this respect. When the attenuation is measured at two energies, their values are not exactly proportional to each other. It turns out that the deviation from a linear relationship, which may be measured with the Dual Energy Index (DEI) [2], is closely related to the effective atomic number,

∗ Corresponding author at: Dept of Clinical Physiology, Linköping University Hospital, SE-581 85 Linköping, Sweden. Tel.: +46 13223309. E-mail address: [email protected] (H. Zachrisson). 0720-048X/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2010.02.001

(Zeff ), an entity introduced to describe the X-ray energy absorption in biological specimens [3,4]. In fact, the effective atomic number can be estimated from the DEI, as long as the atomic number does not exceed 55, i.e. for all elements commonly encountered in the body, as well as for the most common contrast agent, iodine [2]. In this experimental study, soft tissue samples commonly present in arterial plaques were collected at autopsy. The biomaterials included fat tissues from different locations, as well as psoas muscle and myocardium. In addition, venous and arterial thrombi were synthesized. Our hypothesis was that combining images obtained at two energies would significantly improve the classification of different biomaterials compared to single energy (conventional CT).

2. Methods 2.1. Soft tissue samples Postmortem tissue samples were collected at autopsy (pericolic fat, perirenal fat, pericardial fat, subcutaneous fat, greater omentum fat, as well as cardiac and psoas muscle). Psoas muscle and omental fat were chosen as examples of muscle and fat tissue, respectively, that were easy to collect in well-defined samples without a mixture of other tissues.

H. Zachrisson et al. / European Journal of Radiology 75 (2010) e124–e128

e125

The tissue samples were harvested from two healthy adults 1 day postmortem. A forensic pathologist harvested the samples by means of surgical dissection. This pathologist was blinded to all imaging procedures and was instructed to harvest macroscopically homogenous samples. In the case of muscle, adjacent tissue such as fascia was left out to achieve homogeneity. The samples all measured about 15 mm × 15 mm × 40 mm and were stored in separate but identical plastic test tubes at all times. They were kept cool and scanned within 1 h after dissection in order to maintain cellular integrity. Histopathology with haematoxylin-eosin-saffron was used as reference as a standard staining method. Histological analysis was used as a gold standard to verify intact cellular integrity of the tissue samples on a microscopic level after scanning. Micro and macroscopic analysis showed a high degree of agreement. No exclusion of the material was performed due to mixture of adjacent tissue. 2.2. Arterial and venous thrombi Arterial and venous blood samples were collected from healthy volunteers with normal hematocrit and hemoglobin levels. The blood samples were utilized to create arterial and venous thrombi models. Physiologic activation of the blood clotting factors was initiated using Tissue Factor. Tissue Factor is a protein (CD142) found in subendothelial tissue which gets exposed to blood at sites of blood vessel damage or plaque rupture. The experimental thrombi were imaged in a low-artifact phantom using DECT. For the venous clot venous blood was held in a serum clot activator test tube and shaken for 30 s, then transferred to a plastic test tube to clot for 1 h at 37 ◦ C. A plastic test tube was filled with water to serve as a control.

Fig. 1. The sample test tubes were scanned individually in the center of the phantom. Test tubes were aligned to the Z-axis and held in place by a perforated nonattenuating circular disc inside the water phantom. The water phantom measured 15 cm in diameter.

2.3. Scan sequence and settings The tissues were all scanned using a SOMATOM Definition Dual Source CT system (Siemens AG, Medical Solutions, Forchheim, Germany) in dual energy mode. The samples were placed inside a water phantom in order to reduce beam-hardening artifacts. The sample test tubes were scanned individually in the center of the phantom. Test tubes were aligned to the Z-axis and held in place by a perforated non-attenuating circular disc inside the water phantom. The water phantom measured 15 cm in diameter and provided sufficient attenuation to render water its correct radiological property of zero Hounsfield Units (Figs. 1–3). The plastic test tubes had suitable radiological properties, with an X-ray attenuation identical to that of water (manufactured of a Swedish company “Noax lab” (polystyrene plastic material, length 110 mm, volume 15 ml (Noax lab, Kungsängen, Sweden). http://www.noaxlab.se/laboratorieequipment/Provror/Index. html#8). The empty test tubes were scanned within the water phantom and the tubes measured zero Hounsfield Units. Soft tissue samples were scanned separately as well as arterial and venous thrombi. One hour after the clotting reactions were initiated, the tubes were submerged in a water phantom consisting of water at the temperature of 37 ◦ C and scanned separately (Figs. 1–3). I. Scan settings for the soft tissue samples were: (1) Single energy mode, 120 kV, 182 mAs, rotation time 1.0 s, collimation 0.3 mm, reconstruction interval 0.2 mm, kernel D30 s, image matrix 512 × 512, field of view 50 mm resulting in a voxel size of 0.1 mm × 0.1 mm × 0.4 mm. (2) Dual energy (DE) mode, 140 and 80 kV simultaneous acquisition, 95 and 403 mAs respectively. Collimation 0.3 mm,

Fig. 2. The plastic test tubes had suitable radiological properties, manufactured of a Swedish company “Noax lab” (polystyrene plastic material, length 110 mm, volume 15 ml).

Fig. 3. The accuracy of the method was verified by measuring water at zero Hounsfield units.

e126

H. Zachrisson et al. / European Journal of Radiology 75 (2010) e124–e128

Table 1 Descriptive statistics (mean ± standard deviation) for attenuation measurements with two energies. Tissue

Number of voxels

Omental fat Venous thrombus Arterial thrombus Myocardium Psoas muscle

13,677 16,520 16,819 16,352 16,170

HU 140 kV

HU 80 kV

−91.4 35.1 42.0 50.5 49.8

−120.9 41.5 48.3 56.3 57.2

± ± ± ± ±

17.3 18.7 19.7 17.0 20.5

± ± ± ± ±

HUmean 22.7 19.6 19.8 19.8 20.0

−106.2 38.3 45.1 53.4 53.5

DEI ± ± ± ± ±

−0.0166 0.0031 0.0030 0.0028 0.0035

15.8 12.6 15.1 12.9 13.7

± ± ± ± ±

0.0141 0.0139 0.0123 0.0125 0.0141

Table 2 Prediction of tissue type from original HU measurements using logistic regression. Az values for models including one or two independent variables.

Venous thrombus vs Omental fat Arterial thrombus vs Omental fat Myocardium vs Omental fat Psoas muscle vs Omental fat Arterial thrombus vs Venous thrombus Myocardium vs Venous thrombus Psoas muscle vs Venous thrombus Myocardium vs Arterial thrombus Psoas muscle vs Arterial thrombus Psoas muscle vs Myocardium

HU 140 kV

HU 80 kV

HU 140 kV and HU 80 kV

1.000 1.000 1.000 1.000 0.599 0.733 0.696 0.629 0.602 0.520

1.000 1.000 1.000 1.000 0.613 0.706 0.716 0.606 0.621 0.514

1.000 1.000 1.000 1.000 0.644 0.799 0.791 0.655 0.652 0.514

In all cases, the independent variables contributed significantly (p < 0.01) to the model.

rotation time 1 s, reconstruction interval 0.2 mm, kernel D30s, image matrix 512 × 512, field of view 50 mm resulting in a voxel size of 0.1 mm × 0.1 mm × 0.4 mm. II. Scan settings for experimental arterial and venous thrombi were: DE mode, 140 and 80 kV simultaneous acquisition, 95 and 403 mAs respectively, pitch 0.6, collimation 0.6 mm, rotation time 1 s, reconstruction interval 0.5 mm, kernel D30 s, image matrix 512 × 512 and FoV 50 mm. Regions of interest (with a minimum diameter of 5 mm) were drawn in central areas representing a single type of tissue (fat, arterial thrombus, myocardium or psoas muscle) and corresponding voxel values were extracted using the open-source software Image J 1.40f (5) (http://rsb.info.nih.goc/ij/). From the original HU measurements at 140 and 80 kV, a mean value (HUmean ) and the Dual Energy Index (DEI) [2] were computed as HUmean =

HU140 + HU80 2

and DEI =

HU80 − HU140 HU80 + HU140 + 2000

2.4. Statistical methods Descriptive statistics (mean and standard deviation) were computed for each of the four measurements (HU 140 kV, HU 80 kV, HUmean , and DEI) in each of the four tissues. The ability to predict tissue type from the measured HU values was tested with binomial logistic regression. For each pair of tissues, two models using HU values from either of the two energy levels as independent variable, and one model including both of them simultaneously, were created and evaluated with a likelihood ratio test for each independent variable as well as the area under the Receiver operating characteristic curve (ROC), Az . In addition, a corresponding analysis was performed for HUmean and DEI. All statistical computations were carried out in JMP 7.0.1 (SAS Inc., Cary, NC, USA). Approval for this study was obtained from the Regional Ethical Review Board in Linköping.

3. Results Mean and standard deviation (SD) HU values measured at the two peak energies for omental fat, venous thrombus, arterial thrombus, myocardium and psoas muscle are shown in Table 1. Results of the prediction of tissue type from HU values for the same tissue types are given in Table 2. By combining two energy levels (80 and 140 kV) significantly improved the classification of venous vs arterial thrombus; myocardium or psoas muscle vs arterial or venous thrombus; and myocardium vs psoas muscle. The accuracy for distinguishing between myocardium and arterial thrombus, measured as Az , increased from 0.606 (80 kV) to 0.655 (80 and 140 kV), whereas Az for myocardium vs venous thrombus improved from 0.706 to 0.799 for the same energies. For distinguishing fat tissue from the other tissues, single energy alone was sufficient for completely correct classification. The prediction of the same tissue types from HUmean and DEI is shown in Table 3. In general, Az showed higher values for HUmean than for DEI (range 0.503–1.0 vs 0.479–0.847). The addition of DEI to a logistic regression model including HUmean was in several cases significant, but in general increased the Az values only marginally. Descriptive statistics for measurements in different types of fat tissue (colon fat, omental fat, pericardial fat, renal fat, and subcutaneous fat) are shown in Table 4. Logistic regression including Table 3 Prediction of tissue type from HUmean and Dual energy index using logistic regression. Az values for models including one or two independent variables.

Venous thrombus vs Omental fat Arterial thrombus vs Omental fat Myocardium vs Omental fat Psoas muscle vs Omental fat Arterial thrombus vs Venous thrombus Myocardium vs Venous thrombus Psoas muscle vs Venous thrombus Myocardium vs Arterial thrombus Psoas muscle vs Arterial thrombus Psoas muscle vs Myocardium

HUmean

DEI

HUmean and DEI

1.000 1.000 1.000 1.000 0.644 0.797 0.790 0.653 0.651 0.503b

0.839 0.849 0.844 0.841 0.479a 0.486 0.525 0.505a 0.509 0.515

1.000 1.000 1.000 1.000 0.644a 0.799 0.791a 0.655 0.652 0.514b

In all other cases, the independent variables contributed significantly (p < 0.05) to the model. a No significant effect of DEI. b No significant effect of HUmean .

H. Zachrisson et al. / European Journal of Radiology 75 (2010) e124–e128

e127

Table 4 Descriptive statistics (mean ± standard deviation) for attenuation measurements with two energies for different types of fatty tissue. Tissue

Number of voxels

HU 140 kV

HU 80 kV

Colon fat Omental fat Pericardial fat Renal fat Subcutaneous fat

11,130 10,201 8372 10,200 11,235

−103.4 −90.7 −91.8 −98 −107.6

−134.5 −119.2 −112.7 −122.3 −129.7

± ± ± ± ±

25.5 16.2 23.4 19.7 18.2

Table 5 Prediction of type of fatty tissue from original HU measurements using logistic regression. Az values for models including one or two independent variables.

Colon fat vs Omental fat Colon fat vs Pericardial fat Colon fat vs Renal fat Colon fat vs Subcutaneous fat Omental fat vs Pericardial fat Omental fat vs Renal fat Omental fat vs Subcutaneous fat Pericardial fat vs Renal fat Pericardial fat vs Subcutaneous fat Renal fat vs Subcutaneous fat

HU 140 kV

HU 80 kV

HU 140 kV and HU 80 kV

0.662 0.637 0.561 0.557 0.509 0.618 0.745 0.591 0.706 0.627

0.702 0.794 0.671 0.553 0.589 0.549 0.650 0.646 0.742 0.618

0.735 0.804 0.677 0.579 0.597 0.623 0.771 0.661 0.796 0.688

In all cases, the independent variables contributed statistically significantly (p < 0.001) to the model.

Table 6 Prediction of type of fatty tissue from HUmean and Dual energy index using logistic regression. Az values for models including one or two independent variables.

Colon fat vs Omental fat Colon fat vs Pericardial fat Colon fat vs Renal fat Colon fat vs Subcutaneous fat Omental fat vs Pericardial fat Omental fat vs Renal fat Omental fat vs Subcutaneous fat Pericardial fat vs Renal fat Pericardial fat vs Subcutaneous fat Renal fat vs Subcutaneous fat

HUmean

DEI

HUmean and DEI

0.729 0.765 0.640 0.455a 0.545 0.611 0.764 0.652 0.794 0.692

0.506 0.582 0.553 0.573 0.582 0.545 0.571 0.532 0.510 0.522

0.735 0.804 0.677 0.578 0.598 0.623 0.771 0.661 0.797 0.689

In all other cases, the independent variables contributed statistically significantly (p < 0.001) to the model. a No significant effect of HUmean .

HU values at 80 and 140 kV (Table 5) shows that combining two energies (80 and 140 kV) significantly (p < 0.001) improved the possibility for correct classification of different kinds of fat tissue. The Az value for omental fat vs subcutaneous fat, e.g. was improved from 0.650 (80 kV) to 0.771 (80 and 140 kV). However for distinguishing all different kinds of fat tissue from other tissues, single energy alone was adequate for correct classification (cf. Table 2). Finally, when predicting the type of fat from HUmean and DEI (Table 6), for all pairs significant improvements were noted when both variables were included, again with, in most cases, only slight increases in the Az values compared to the values for HUmean alone. 4. Discussion In the present study, the addition of a second X-ray energy level increased the predictive power of the logistic regression model to a small but significant extent. By combining two energies (80 and 140 kV) the possibility for correct classification of soft tissue significantly improved. This was the case for venous vs arterial thrombus; myocardium or psoas vs arterial thrombus; myocardium or psoas vs venous thrombus, as well as differentiation between different types of fat tissue compared to single energy (conventional CT). However,

± ± ± ± ±

HUmean 19.3 21.7 17.8 18.5 20.0

−118.9 −105 −102.2 −110.1 −118.7

DEI ± ± ± ± ±

17.2 15.2 15.4 13.5 13.7

−0.0176 −0.016 −0.0116 −0.0136 −0.0126

± ± ± ± ±

0.0164 0.0132 0.0155 0.0151 0.0152

for distinguishing fat tissue from other tissues, single energy alone was sufficient for correct classification. Correct classification of soft tissues is difficult by conventional CT since they form a closely clustered group on the linear Hounsfield scale of X-ray attenuation. Although the first experiments with dual energy CT date back to the 1970s [5], the limited spatial resolution, the unstable CT numbers with systematic errors and long scan durations hampered the success for general application of the technique [6]. Separate imaging of bone and soft tissue was proposed early [7]. However, the practical utility of this principle was severely limited by the difficulties in combining images acquired at two different points in time. It was not until the advent of scanners with dual X-ray tubes that simultaneous acquisition at two different energy levels became practically feasible without the complicated and costly set-up of a synchrotron radiation source [8]. With the two tubes and detectors mounted orthogonally [9], both spiral acquisitions run simultaneously, which largely excludes changes in contrast enhancement or patient movement between the acquisitions. Although the recently introduced dual source CT scanner was primarily developed to achieve high temporal resolution in cardiac imaging [10,11], it also resolves a main drawback of contrast material applications with dual energy differentiation. The information about effective atomic number provided by DECT can be used to discriminate materials that in single-energy images have similar HU values, such as calcium and iodine contrast [12]. Using image post-processing with three-material decomposition, bone and calcified plaque can be removed from the image with a bone removal module, which facilitates the assessment of vascular structures [13]. New dual energy applications are continuously developed such as material differentiation, plaque imaging, iodide quantification and bone removal [12,14–16]. One of the clinical applications of iodine differentiation in dual energy CT (DECT) is the possibility to display contrast-enhanced vessels without calcium-containing structures. Single energy does not differentiate between the iodine contrast in the vessel lumen and the calcium in the vessel wall. By using dual energy (80 and 140 kV) computation calcified plaques in the vessel wall are clearly seen and soft plaques are suppressed. Early studies have suggested that plaque in different stages can be identified according to their attenuation of the X-ray signal. Lipid-rich plaques have Hounsfield values (HU) less than 60, calcified plaques above 120 HU and fibrotic plaques around 100 HU [1]. DECT shows potential to add information in the issue of the vulnerability of plaques. In such plaques, iron and calcium deposits accumulate with high spatial coincidence and nearly similar size. With DECT the presence of iron in haemorrhagic lesions and calcium can be visualized [17]. Thus DECT can provide information that is presently difficult to obtain by ultrasound. MRI studies have also shown that it is possible to detect the fibrous cap and necrotic core as well as intraplaque hemorrhage of the plaque [18–20]. However a disadvantage of MRI studies is that scanning in general is considerably longer than with CT. Conventional CT has been compared to ultrasonography concerning carotid plaque morphology and histology [21]. Still, plaque classifi-

e128

H. Zachrisson et al. / European Journal of Radiology 75 (2010) e124–e128

cation remains a challenge and the important issue of which plaque structure components indicate a symptomatic lesion continues to be debated. We found that in vitro DECT significantly improved in vitro differentiation between muscle and thrombus as well as fat and thrombus which might be of importance for plaque visualization and in vascular studies. As seen in Table 2, by combining two energy levels (80 and 140 kV) significantly improved the classification of soft tissue discrimination also including venous vs arterial thrombus. Still, the capability for correct classification is limited. The Az values reveal that the ability to correctly classify voxels was only marginally improved by the addition of a second energy. The dual energy index (DEI), which was proposed to describe differences in attenuation characteristics between different elements, was of limited value in our study, as the Az values for DEI were, in general, lower than for HUmean . However, the addition of DEI in the model made a significant contribution in four pairs of tissues (Myocardium vs Venous thrombus; Myocardium vs Arterial thrombus; Psoas muscle vs Arterial thrombus; Psoas muscle vs Myocardium). One major problem with DECT in differentiating soft tissue structures such as muscle, cartilage, ligaments and tendons is the fact that these tissues mainly contain elements with low atomic numbers and thus differ only slightly in DEI. However Johnson et al. demonstrated recently that collagen could be differentiated from water and soft tissue [12]. In our study, psoas muscle and omental fat were chosen as examples of muscle and fat tissue, respectively, that were easy to collect in well-defined samples without admixture of other tissues. The important forensic problem of discrimination between clots that have been developed in a living body from postmortem clotting, by application of dual energy is a matter of further investigation. Concerning the issue of distinguishing fat tissue from different locations it was possible already with single energy to separate the tissues whereas all pair of comparisons showed significantly improved possibilities for classification using dual energy. The latter finding might be of importance for improving postmortem imaging [22], where detection of fat embolism from different locations could be critical. DECT will in the future offer new diagnostic possibilities concerning different kinds of CT investigations for soft tissue discrimination. One important field is tumour diagnosis. Preliminary results indicate that DECT may be an excellent imaging technique to identify patients with ectopic parathyroid adenoma [23]. In the vascular field plaque assessment as well as discrimination of fresh and old thrombus material will be a challenge for stroke imaging where DECT has the possibility to offer considerable clinical impact. Future studies have to be performed evaluating an optimal model from training data in new test material both ex vivo and in vivo including new better equipment available by DECT. Since the various tissue types can be differentiated by virtue of the differences in effective atomic number between them, it seems reasonable to expect similar results when the method, in future studies, will be applied in vivo. Although the addition of a second energy in most cases yielded significant results, the improvement in classification ability measured as Az was in most cases marginal (cf. Tables 2 and 5). 4.1. Limitations Our analysis of the data included logistic regression using voxel data. An alternative approach would be to apply discriminant analysis but that method requires stronger statistical assumptions and is strictly valid only for multivariate normal distributions. From Table 3 it is clear that the addition of DEI in the model made a significant contribution in four pairs of tissues (Myocardium vs Venous thrombus; Myocardium vs Arterial thrombus; Psoas muscle

vs Arterial thrombus; Psoas muscle vs Myocardium). Although the Az values are only marginally improved it is clear that that the value of DEI is limited. 5. Conclusion DECT offers significantly improved in vitro differentiation between soft tissues occurring in plaques, but the capability for correct classification is still limited. Future studies are needed to clarify the diagnostic value of these findings and whether this corresponds to better tissue discrimination in vivo. Conflict of interest No conflict of interest exists. References [1] Brodoefel H, Reimann A, Heuschmid M, et al. Characterization of coronary atherosclerosis by dual-source computed tomography and HU-based color mapping: a pilot study. Eur Radiol 2008;18(11):2466–74. [2] Schmidt B, McCollough CH. Dual-energy computed tomography. In: Gerber TC, Kantor B, Williamson EE, editors. Computed tomography of the cardiovascular system. London: Informa Health Care; 2007. p. 451–62. [3] Spiers FW. Effective atomic number and energy absorption in tissues. Br J Radiol 1946;19(218):52–63. [4] Murty RC. Effective atomic numbers of heterogeneous materials. Nature 1965;207(4995):398–9. [5] Rutherford RA, Pullan BR, Isherwood I. Measurement of effective atomic number and electron density using an EMI scanner. Neuroradiology 1976;11(1):15–21. [6] Kelcz F, Joseph PM, Hilal SK. Noise considerations in dual energy CT scanning. Med Phys 1979;6(5):418–25. [7] Hemmingsson A, Jung B, Ytterbergh C. Dual energy computed tomography: simulated monoenergetic and material-selective imaging. J Comput Assist Tomogr 1986;10(3):490–9. [8] Tsunoo T, Torikoshi M, Sasaki M, Endo M, Yagi N, Uesugi K. Distribution of electron density using dual-energy X-ray CT. IEEE Trans Nucl Sci 2003;50(5):1678–82. [9] Flohr TG, McCollough CH, Bruder H, et al. First performance evaluation of a dual-source CT (DECT) system. Eur Radiol 2006;16(2):256–68. [10] Johnson TR, Nikolaou K, Wintersperger BJ, et al. Dual-source CT cardiac imaging: initial experience. Eur Radiol 2006;16(7):1409–15. [11] Achenbach S, Ropers D, Kuettner A, et al. Contrast-enhanced coronary artery visualization by dual-source computed tomography-initial experience. Eur J Radiol 2006;57(3):331–5. [12] Johnson TR, Krauss B, Sedlmair M, et al. Material differentiation by dual energy CT: initial experience. Eur Radiol 2007;17(6):1510–7. [13] Flohr TG, Bruder H, Stierstofer K, Petersilka M, Schmidt B, McCollough CH. Image reconstruction and image quality evaluation for a dual source CT scanner. Med Phys 2008;35(12):5882–97. [14] Barreto M, Schoenhagen P, Nair A, et al. Potential of dual-energy computed tomography to characterize atherosclerotic plaque: ex vivo assessment of human coronary arteries in comparison to histology. J Cardiovasc Comput Tomogr 2008;2(4):234–42. [15] Tomandl BF, Hammen T, Klotz E, Ditt H, Stemper B, Lell M. Bonesubtraction CT angiography for the evaluation of intracranial aneurysms. AJNR 2006;27(1):55–9. [16] Lell MM, Anders K, Uder M, et al. New techniques in CT angiography. Radiographics 2006;26(Suppl 1):S45–62. [17] Langheinrich AC, Michniewicz A, Sedding DG, et al. Quantitative X-ray imaging of intraplaque hemorrhage in aortas of apoE (−/−)/LDL (−/−) double knockout mice. Invest Radiol 2007;42(5):263–73. [18] Chu B, Kampschulte A, Ferguson MS, et al. Hemorrhage in the atherosclerotic carotid plaque: a high-resolution MRI study. Stroke 2004;35(5):1079–84. [19] Yuan C, Zhang SX, Polissar NL, et al. Identification of fibrous cap rupture with magnetic resonance imaging is highly associated with recent transient ischemic attack or stroke. Circulation 2002;105(2):181–5. [20] Saam T, Hatsukami TS, Takaya N, et al. The vulnerable, or high-risk, atherosclerotic plaque: noninvasive MR imaging for characterization and assessment. Radiology 2007;244(1):64–77. [21] Gronholdt ML. B-mode ultrasound and spiral CT for the assessment of carotid atherosclerosis. Neuroimaging Clin N Am 2002;12:421–35. [22] Persson A, Jackowski C, Engström E, Zachrisson H. Advances of dual source, dual-energy imaging in postmortem CT. Eur J Radiol 2008;68(3):446–55. [23] Gimm O, Juhlin C, Morales O, Persson A. When other imaging techniques fail: dual energy computed tomography (DECT) localizes ectopic parathyroid adenoma. Department of Surgery, Department of Radiology, and Center for Medical Image Science and Visualization (CMIV), University Hospital, Linköping University, Sweden, submitted for publication. Radiology 2010.