Dual Energy on Dual Source CT in Abdominal Imaging

comparison between Virtual-Non-Contrast and True-Non-Contrast images using quantitative and qualitative analysis and evaluation of dose report


Authors
Pub.Info
Article Navigation
DOI
Abstract
Pub.Info

Submission Date: 2020-04-11
Review Date: 2020-04-24
Pubblication Date: 2020-05-07
Printed on:

Abstract

Abstract:

Establish whether virtual non-contrast images can replace real non-contrast images, avoiding the preliminary scan, thus saving the patient dose.
Forty-one patients were studied on a second-generation dual-source scanner, triphasic were applied on all patients, arterial and venous phase was acquired in dual-energy. HU values of TNC and VNC scans were compared through ROI on liver, spleen, kidneys, aorta, muscle, and fat. Qualitative analysis of the data sets was performed by four readers and the values from 1 to 4 (1. poor, 2. fair, 3. good, 4. excellent.) were assigned to define the diagnostic quality of the images. Then the absorbed dose ratios between the virtual images and the pre-contrast and portal phase images were compared.
HU values were analyzed with the t-test and the difference was statistically significant P<0.0001. Image quality was evaluated as excellent or good in 91,3% of TNC and 81,6% of VNC. At last, VNC showed a sensitive dose reduction -46% compared to conventional triphasic protocol. Overall VNC has shown a good image quality comparable to TNC. The dose reduction obtained from the extraction of images without iodine, avoiding further scans, suggests the use of the dual-energy protocol in many exams.

Introduction

Dual Energy CT (DECT) is one of the last frontiers in Computed Tomography technology, which bases its principle on the acquisition at two energy levels (keV) simultaneously (fig.1), obtaining image information from two different x-ray spectra. For this reason, DECT is also commonly defined as Spectral CT and provides images with different attenuation from two x-ray beams, to better characterize tissues.

When we talk about attenuation, we refer to the attenuation coefficient of tissue (mt) that can be described as mr that refers to the total mass attenuation coefficient for all photon/matter interactions. Particularly important interaction is the photoelectric absorption (t/r), which contributes mostly to imaging, and it is related to the atomic number (Z) of the medium and the photon energy of the radiation (E) cubed whereby:

In diagnostic imaging, the photoelectric effect is relevant for the elements with high atomic number (Z) like Iodine or Calcium, and it occurs when the photon energy is equal to, or greater than the binding energy of the electron in K-shell [1-2].

In the presence of two-photon beam energy levels across a medium, we will get more attenuation values for that medium. The difference between high and low keV values varies among elements, according to their K-edge [3-8].

Indeed, elements with low Z have minor differences, as soft tissues, while elements as Iodine and Calcium show a major difference at two energy levels (fig.2). Therefore, it is possible to decompose these materials using specific software as the Material Decomposition (MD), which can isolate and subtract the Iodine or Calcium from tissues. An application is the possibility to remove iodine material from post-contrast scans, obtaining images without contrast or “virtual non-contrast” [9-12] when they are required.

Material and Methods

Between April and June 2019, 72 abdominal CT studies were performed after intravenous contrast material acquired in dual-energy modality. 41 out of 72 underwent a preliminary un-enhanced scan and retrospectively analyzed. The patients, 25 males, and 15 females had a mean age of 68 years (minimum 30, maximum 88; average 73, median 70). Body Mass Index ranged from 20 to 29, and patients with values above 29 were excluded. Patients presenting with hepatic steatosis, splenectomy or nephrectomy were also excluded.

CT Protocol

The above-mentioned patients underwent a three-phase study protocol including un-enhanced scans followed by arterial and venous phases performed on a second-generation dual-source scanner. Dual Energy (DE) mode was applied on both post-contrast phases in 6 out of 41 patients, while in 20 in the arterial phase and in the remaining 15 in the venous phase. The scans without contrast media were acquired with kV modulator and automatic mAs about 170, the pitch was 0.6, gantry rotation of 0.5 seconds and collimation of 1.5 mm x 128 z focus. For dual-energy scans, the power of the two tubes was 100 kV (tube A) and Sn140 kV (tube B), while the mAs were dispensed automatically with a dose modulator. The pitch was 0.6 and the collimation 0.6 mm x 128 z focus, the rotation speed of the gantry was 0.5 seconds. All images were acquired with a thickness of 3 mm and reconstructed with the iterative algorithm. High iodinated contrast medium (400 mg/ml) was intravenously injected at 3 ml/sec flow rate followed by 50 ml saline solution. The time of the arterial phase was established by a ROI positioned on the thoracic descending aorta fixed at a value of 150 HU. The venous phase was acquired 40 seconds after the end of the arterial phase.

Post-processing

Dual Energy scans generated two data sets for each tube/detector system with different energy levels, afterward, the image sets were processed with material decomposition software for Iodine subtraction and un-enhanced images were generated. Virtual Non-contrast (VNC) images were used for the quantitative and qualitative comparison with the True Non-Contrast images (TNC) (fig.3).

Quantitative analysis

The quantitative analysis was performed by measuring the Hounsfield values using the ROIs on the different organs acquired in TNC and VNC, taking care to position them in the same points on the two sets of images (fig.4). ROIs were drawn with a diameter of 1cm2, for each data set three of them were placed on the liver, avoiding the portal vessels, on the spleen, on descending aorta from T11 to L1, on the kidneys bilaterally, on the para-vertebral muscles and retroperitoneal and subcutaneous fat. A total of 2160 ROIs were analyzed.

Qualitative analysis

The images were also subjected to a qualitative comparison through viewing by four readers with different experiences, from two to twenty years. Assessment concerned exclusively:

  • Evaluation of hepatic parenchyma with surrounding structures;
  • The distinction of the portal branches in the hepatic parenchyma;
  • Rating of the renal parenchyma.

The evaluation was based on a scale of four qualitative values: 4=excellent, images without artifacts and well-defined structures, 3=good, presence of minor artifacts, but still readable, 2=fair, presence of some artifacts, images not easy to read, 1=poor, the artifacts do not allow to distinguish the structures, therefore not readable images. Each value was assigned to the single organ on the single image set and the results per patient were compared. Finally, a cut-off of reportable images was established, “sufficient” for 4 and 3 values, “insufficient” for 2 and 1 values.

Statistical analysis

Attenuation values were subjected to the Shapiro-Wilks normality test, afterward for normal distribution values a paired t-test was used, while for non-normal distributions a non-parametric Wilcoxon Signed Rank test was used. The qualitative evaluation has been subjected to the Mann Whitney U test and represented on histograms and 100% stacked column charts. All statistical analyses were performed by using an appropriate software the QtiPlot software www.qtiplot.com and www.statskingdom.com.

Results

Attenuation values of TNC and VNC data sets demonstrated statistically significant differences in the arterial phase on liver (p<0.0001), aorta (p<0.0001), kidney (p=0.008), spleen (p<0.0001) and fat (p<0.0001), and in the venous phase (tab.2 and 3) on liver (p<0.0001), spleen (p<0.0001), kidney (p=0.001), muscle (p<0.0001) and fat (p<0.0001) (tab.1). Only aorta values had no statistically significant differences (p=0.1) (tab.2).

Comparative distributions of attenuation values were represented on the box-whisker diagram (fig.5), which shows higher values for VNC than TNC. A further comparison was to evaluate the distribution of the average differences between the HU values using the Bland Altman diagram, which shows a good agreement between two data sets, with the majority of the values, included within the two error bars (average ±1.96 x standard deviation) (fig.6). 

Qualitative analysis showed statistically significant differences for kidneys p<0.0001, portal branches p=0.02 and liver p<0.0001. The stacked column graph (fig.7) shows the trend of the percentage in subjective evaluation. The liver, and kidneys in TNC images differ with more assessments as “excellent” (67%) compared to VNC who has been best evaluated as “good” (54%) (fig.8). However, the cut-off has included most values suitable for reporting of both true non-contrast images (91%) and virtual non-contrast images (81%) (fig.9).

Dose reduction

Un-enhanced and dual-energy scans had the same CTDI values. However, subtracting DLP values of TNC from Dual Energy scans, the significant lower delivered dose is shown. A comparison of the dose was performed by DLP medians (mGy*cm) with and without un-enhanced scans (tab. 5). The median difference was expressed as a percentage and shown a lower dose in scans without an un-enhanced phase of 48% for arterial phase scans and 46% for venous phase scans. Thus, a significant dose reduction is demonstrated if virtual un-enhanced images were obtained from dual-energy data sets.

Discussion

TNC and VNC attenuation values were statistically different, due to a different X-ray spectrum. Measurement differences have been calculated as a percentage of values lying within a tolerance interval of ±15 HU. Percentages were expressed for any tissue (tab.4), high values have been seen in the kidney (97%) and muscle (93%), less of 80% the other tissues. Meanwhile, the overall percentage for all tissues within the tolerance interval was 82,3% (p=0.09). 

Moreover, also qualitative analysis showed differences between the two acquisition ways. So, TNC is considered to be more accurate compared to VNC with more excellent evaluations. On the other hand, VNC proved to be as adequate for reporting due to more good evaluations adding potential dose reduction because of dual-energy acquisition. It makes possible to use VNC images as an un-enhanced assessment in abdominal imaging in several applications. 

Final considerations

Although the VNCs have been demonstrated lower quality compared to TNCs, indeed they are sufficient for reporting, and they are a helpful means to obtain un-enhanced images from a single dataset with contrast medium, thus further scans are not needed. A dose reduction of the VNCs is widely demonstrated [13-16], almost halved, this is important in young patients in which DE scans can be used routinely. However several conditions can give unsure results, for example, we said that x-ray dual energy levels well distinguish Iodine materials, but other elements can be same to Iodine in attenuation measurements such as some metallic stitches, some endovascular prosthesis, gallstones[17], and some kinds of calcifications, that may be likewise subtracted in VNC reconstruction processing. Anyway, further studies should be performed to assess other materials’ attenuation differences to best understanding the behavior in Dual Energy modality.

Fig. 1Double x-ray spectra; the combination of 100 kVp and 140 kVp with the addition of 0.4 mm tin filter applied to the high energy (blue spectrum) allows to decrease the characteristic discrete peaks of tungsten and increase Bremsstrahlung effect that improves the distance between the two curves and it makes to grow spectral contrast. 
Fig. 2 – Mass attenuation coefficients for Iodine (red), bone (blue) and water (green). K-edge of Iodine is about 33 keV, while the diagnostic energy values are 55 keV for 80 kV and 73 keV for 140 kV settings. Note that Iodine has a major difference in this range of respect to bone and water. This feature makes it possible to well distinguish Iodine from the bone.
 Fig. 3 Images generation from a Dual Energy CT data set using Material Decomposition post-processing. DECT performs two image sets in unique acquisition at 100 kV and 140 kV (A, B respectively). Afterward, it generates a DECT mixed images equivalent to 120 kV conventional image (C). MD software distinguishes the contrast material and generates an iodine only image (D). A mixed image with iodine overlays map is generated, and voxels with Iodine are color-coded as orange. (E), thus it is possible to subtract iodine material generating a virtual non-contrast image (F), that can replace the true non-contrast image (G).
Fig. 4 Placement of the ROIs on TNC and VNC data sets regarding liver, aorta, spleen, kidneys, muscles and fat. First, the ROIs were drawn on TNC images, then copied and pasted on the VNC image series at the corresponded positions.
Table 1Q1MedianQ3p value
HU Liver Arterial TNC455055<0,0001
HU Liver Arterial VNC556367
HU Aorta Arterial TNC313742<0,0001
HU Aorta Arterial VNC384653
HU Kidney Arterial TNC2728300,008
HU Kidney Arterial VNC273136
HU Spleen Arterial TNC38,54347<0,0001
HU Spleen Arterial VNC49,55559
HU Fat Arterial TNC-116-109-96<0,0001
HU Fat Arterial VNC-108-101-88
Table 2Q1MedianQ3p value
HU Aorta Venous TNC3137,541,50,1
HU Aorta Venous VNC28,532,537,5
HU Kidney Venous TNC252829,50,001
HU Kidney Venous VNC263036
HU Muscle Venous TNC35,54348,5<0,0001
HU Muscle Venous VNC41,54752
HU Fat Venous TNC-113-107,5-98<0,0001
HU Fat Venous VNC-105-99,5-89
Tab. 1-2 To expect differences between attenuation values of two populations, Wilcoxon Signed Rank for non-normal distributions was performed; test shows only aorta values in venous phase have no statistical differences respect to the other organs.

Table 3DoFMeanSDp value
HU Liver Venous TNC5047,017,6<0,0001
HU Liver Venous VNC5054,298,5
HU Spleen Venous TNC4841,554,8<0,0001
HU Spleen Venous VNC4855,105,5
HU Muscle Arterial TNC6640,118,7<0,0001
HU Muscle Arterial VNC6645,928,5
Tab. 3 – The​​ normally distributed values were subjected to the t-test for paired samples, resulting in a statistically significant difference in all cases.

Fig. 5 Box-whisker plotgraphics show the difference in attenuation values between the two acquisition types. Note the trend toward higher HUs in VNC images, except the aorta in the venous phase.
Fig. 6 – Bland Altman plot analyses for attenuation differences between TNC and VNC dataset with a linear trend line. Note the majority of dots lying within the two standard deviation bars, resulting in a good agreement between the two datasets.
Table 4Tolerance HU ±15P
Liver77%0,6
Aorta79%0,5
Kidney97%0,6
Spleen74%<0,0001
Muscle93%0,1
Fat74%0,1
Tab. 4Attenuation differences were evaluated as percentages which are referred to values within tolerance interval fixed to ±15 HU.

Fig. 7 – Quality assessment represented on a 100% graph with stacked columns and divided by organs and acquisition type. TNC images show the best quality of kidney and liver compared to VNCs. The portal vessels were rate excellent lower for both acquisitions due to lower Contrast to Noise Ratio (CNR) that made the read most difficult compared to other tissues.
Fig. 8 – Overall percentage values represented on 100% stacked columns per acquisition type. TNCs exceed VNCs in “excellent” values, however, good quality in VNCs is demonstrated.
Fig. 9 Cut-off that divided quality assessment in two categories, “readable” for excellent and good values, and not readable for fair and poor values. The majority of values are included in readable column, this means that VNC images has high percentage to be considered as true non contrast images.
DLP (mGy*cm)DLP (DE only)DifferencepDose reduction (%)
TNC + DE Arterial 839434405<0,000148%
TNC + DE Venous 878473405<0,000146%
Tab. 5Comparisondose differences in biphasic protocols with and without true non-contrast phase. Numbers are expressed as median and percentage of difference.

References

  1. Donald Graham MEd TDCR, Paul Cloke MSc TDCR, Martin Vosper MSc HDCR. (2012) Principles and Applications of Radiological Physics. Churchill Livingstone. ISBN-10: 0702052159.
  2. Ph.D. Mahesh, Mahadevappa. (2009) MDCT Physics: The Basics–Technology, Image Quality and Radiation Dose. Lippincott Williams & Wilkins. ISBN-10: 9780781768115.
  3. Tobias Fält, Marcus Söderberg, Johan Wassélius, and Peter Leander (2015) Material Decomposition in Dual-Energy Computed Tomography Separates High-Z Elements from Iodine, Identifying Potential Contrast Media Tailored for Dual Contrast Medium Examinations. J Comput Assist Tomogr 00: 00–00.
  4. Thorsten R. C. Johnson (2012) Dual-Energy CT: General Principles. American Journal of Roentgenology 199(5_supplement): S3-S8.
  5. T. Barrett, D.J. Bowden, N. Shaida, E.M. Godfrey, A. Taylor, D.J. Lomas, A.S. Shaw (2010) Virtual unenhanced second generation dual-source CT of the liver: Is it time to discard the conventional unenhanced phase? European Journal of Radiology 81(7):1438-45.
  6. C. N. De Cecco, V. Buffa, S. Fedeli, M. Luzietti, A. Vallone, R. Ruopoli, V. Miele, M. Rengo, P. Paolantonio, M. Maurizi Enrici, A. Laghi, V. David (2010) Dual energy CT (DECT) of the liver: conventional versus virtual unenhanced images. European Radiology 20(12):2870-5.
  7. C.N. De Cecco, V. Buffa, S. Fedeli, A. Vallone, R. Ruopoli, M. Luzietti, V. Miele, M. Rengo, M. Maurizi Enrici, P. Fina, A. Laghi, V. David (2010) Preliminary experience with abdominal dual-energy CT (DECT): true versus virtual nonenhanced images of the liver. La radiologia medica 115(8):1258-66.
  8. Juergen Fornaro & Sebastian Leschka & Dennis Hibbeln & Anthony Butler & Nigel Anderson & Gregor Pache & Hans Scheffel & Simon Wildermuth & Hatem Alkadhi & Paul Stolzmann (2011) Dual-and multi-energy CT: approach to functional imaging. Insight into Imaging 2(2):149-159.
  9. Kevin P. Murphy, Kathryn E. Darras, Savvas Nicolaou, Patrick D. McLaughlin (2016) Emergency Abdominal Applications of DECT. Current Radiology Reports 4(10).
  10. Joseph R. Grajo; Manuel Patino; Andrea Prochowski; Dushyant V. Sahani (2016) Dual energy CT in practice: Basic principles and applications. Applied Radiology 45(7):6-12.
  11. Alvin C. Silva, Brian G. Morse, Amy K. Hara, Robert G. Paden, Norio Hongo, William Pavlicek (2011) Dual-Energy (Spectral) CT: Applications in Abdominal Imaging. Radiographics 31(4):1031-46; discussion 1047-50.
  12. Ravi K. Kaza, Joel F. Platt, Richard H. Cohan, Elaine M. Caoili, Mahmoud M. Al-Hawary, Ashish Wasnik (2012) Dual-Energy CT with Single- and Dual-Source Scanners:  Current Applications in Evaluating the Genitourinary Tract. Radiographics 32(2):353-69.
  13. Daniele Marin, Daniel T Boll, Achille Mileto, Rendon C. Nelson (2014) State of the Art: Dual-Energy CT of the Abdomen. Radiology 271(2):327-42
  14. Joseph R. Grajo, Dushyant V. Sahani, (2017) Dual-Energy CT of the Abdomen and Pelvis: Radiation Dose Considerations. Journal of the American College of Radiology 15(8).
  15. Anushri Parakh, MDa, Francesco Macri, MD, PhDa,b, Dushyant Sahani (2018) Dual-Energy Computed Tomography Dose Reduction, Series Reduction, and Contrast Load Reduction in Dual-Energy Computed Tomography. Radiologic Clinics of North America 56(4):601-624.
  16. S. Jamali, N. Michoux, E. Coche, C.A. Dragean (2019) Virtual unenhanced phase with spectral dual-energy CT: Is it an alternative to conventional true unenhanced phase for abdominal tissues? Diagnostic and Interventional Imaging 100(9).
  17. Han A Lee, Young Hwan Lee, Kwon-Ha Yoon, Dong-Ho Bang, Dong Eun Park (2016) Comparison of Virtual Unenhanced Images Derived from Dual-Energy CT with True Unenhanced Images in Evaluation of Gallstone Disease. American Journal of Roentgenology 206(1):74-80.