- Email: [email protected]
Image quality characteristics for virtual monoenergetic images using dual-layer spectral detector CT: Comparison with conventional tube-voltage images
Physica Medica 49 (2018) 5–10
Contents lists available at ScienceDirect
Physica Medica journal homepage: www.elsevier.com/locate/ejmp
Original paper
Image quality characteristics for virtual monoenergetic images using duallayer spectral detector CT: Comparison with conventional tube-voltage images
T
⁎
Daisuke Sakabea,b, Yoshinori Funamac, , Katsuyuki Taguchid, Takeshi Nakaurae, Daisuke Utsunomiyae, Seitaro Odae, Masafumi Kidohe, Yasunori Nagayamae, Yasuyuki Yamashitae a
Graduate School of Health Sciences, Kumamoto University, Kumamoto, Japan Department of Radiology, Kumamoto University Hospital, Kumamoto, Japan Department of Medical Physics, Faculty of Life Sciences, Kumamoto University, Kumamoto, Japan d The Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, USA e Department of Diagnostic Radiology, Faculty of Life Sciences, Kumamoto University, Kumamoto, Japan b c
A R T I C LE I N FO
A B S T R A C T
Keywords: Dual-layer spectral detector CT Virtual monoenergetic image Image quality Iodine Different object sizes
Purpose: To investigate the image quality characteristics for virtual monoenergetic images compared with conventional tube-voltage image with dual-layer spectral CT (DLCT). Methods: Helical scans were performed using a first-generation DLCT scanner, two different sizes of acrylic cylindrical phantoms, and a Catphan phantom. Three different iodine concentrations were inserted into the phantom center. The single-tube voltage for obtaining virtual monoenergetic images was set to 120 or 140 kVp. Conventional 120- and 140-kVp images and virtual monoenergetic images (40–200-keV images) were reconstructed from slice thicknesses of 1.0 mm. The CT number and image noise were measured for each iodine concentration and water on the 120-kVp images and virtual monoenergetic images. The noise power spectrum (NPS) was also calculated. Results: The iodine CT numbers for the iodinated enhancing materials were similar regardless of phantom size and acquisition method. Compared with the iodine CT numbers of the conventional 120-kVp images, those for the monoenergetic 40-, 50-, and 60-keV images increased by approximately 3.0-, 1.9-, and 1.3-fold, respectively. The image noise values for each virtual monoenergetic image were similar (for example, 24.6 HU at 40 keV and 23.3 HU at 200 keV obtained at 120 kVp and 30-cm phantom size). The NPS curves of the 70-keV and 120-kVp images for a 1.0-mm slice thickness over the entire frequency range were similar. Conclusion: Virtual monoenergetic images represent stable image noise over the entire energy spectrum and improved the contrast-to-noise ratio than conventional tube voltage using the dual-layer spectral detector CT.
1. Introduction Dual-energy computed tomography (CT) is performed using different scan techniques, including dual-spin, tube potential switching, and dual-source beam techniques, and two different X-ray spectrums are acquired [1–3]. These techniques have been used in valuable clinical applications, such as virtual monoenergetic imaging, iodine mapping, determination of effective atomic number, and measurement of electron density [4–9]. Recently, a raw data-based dual-energy CT called dual-layer spectral detector CT has become available [10–14] as a new device. Unlike
⁎
tube voltage beam method, dual-layer spectral detector CT is the first commercially available detector-based CT that uses a single-tube voltage beam. With dual-layer spectral detector CT, low-energy photons are absorbed by the first detector layer, and high-energy photons are absorbed by the second. Therefore, dual-layer spectral detector CT not only provides valuable dual-energy information but is also routinely used in conventional tube-voltage images (e.g., 120 or 140 kVp) simultaneously. So far, substantially increased image noise in lower monoenergetic images (e.g., 40–50 keV) has been a critical issue in clinical situations because increasing image noise degrades image quality, and
Corresponding author at: Department of Medical Physics, Faculty of Life Sciences, Kumamoto University, 4-24-1 Kuhonji, Kumamoto 862-0976, Japan. E-mail address: [email protected] (Y. Funama).
https://doi.org/10.1016/j.ejmp.2018.04.388 Received 31 October 2017; Received in revised form 9 April 2018; Accepted 10 April 2018 1120-1797/ © 2018 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
0.75 s; beam pitch, 0.8; and display field-of-view (FOV), 22.0 cm for 20cm acrylic cylindrical and Catphan phantoms and 32 cm for a 30-cm acrylic cylindrical phantom. The tube current-time product was set to 110 mAs at 120 kVp, 80 mAs at 140 kVp for a 20-cm acrylic cylindrical phantoms and 800 mAs at 120 kVp, 550 mAs at 140 kVp for a 30-cm acrylic cylindrical phantom. Conventional 120- and 140-kVp images and virtual monoenergetic images (40–200-keV images) were reconstructed at slice thicknesses of 1.0 and at a slice interval of 1.0 mm with an abdomen standard kernel (C). In addition, the virtual monoenergetic images were also reconstructed using corresponding denoising levels 2 and 4(iDose4 level 2 and 4), which decreased the image noise with increasing level.
deteriorated virtual monoenergetic images cannot provide a sustainable clinical image [15,16]. To decrease image noise of virtual monoenergetic images, dual-layer spectral detector CT introduces an “anticorrelated statistical reconstruction algorithm” in spectral reconstruction [17,18] which operates in the image and projection domains (postdecomposition), the latter being possible only in a projection-based decomposition technology (with a perfect alignment raw data sets) [19]. Then the method starts with converting a pair of high and low energy datasets obtained from dual-layer detector to Compton scatter and photoelectric data. It then performs spectral reconstruction, combining the following three points: the first point is maximizing a regularized log-likelihood. The second point is based on the fact that the noise of Compton and photoelectric data were negatively correlated. The noise can be effectively decreased by canceling out the correlated noise. The third point is using a synthesized data via a weighted sum of Compton and photoelectric data. This algorithm is effective in suppressing increased image noise by performing basis material decomposition on the spectral data. The spectral reconstruction for dual-layer CT made the amount of noise rather consistent regardless of the effective energies of monoenergetic keV images. Therefore, decreased image noise is expected in virtual monoenergetic images relative to the noise in other dual-energy mode images. In addition, a stable iodine CT number is also expected for virtual monoenergetic images regardless of the object size and acquisition method (120 or 140 kVp), and even conventional polychromatic image noise is decreased with increasing object size [4,20]. The improvement in the image quality with duallayer spectral detector CT is expected to be routinely utilized for poor vessel enhancement, low iodine-contrast detectability, and so on. However, the image quality of dual-layer spectral detector CT virtual monoenergetic images is uncertain with respect to specific characteristics, such as image noise, iodine enhancement at different object sizes, and improvement of the contrast-to-noise ratio. The purpose of this study was to investigate the image quality characteristics for comparison of virtual monoenergetic images with conventional tube-voltage images in a dual-layer spectral detector CT scanner.
2.3. Measurement of the CT number, image noise, and CNR Using a slice thickness of 1.0 mm, the CT number was measured for each iodinated enhancing material in the acrylic cylindrical phantom on conventional 120- and 140-kVp, and virtual monoenergetic images (40–200 keV), including those images with denoising levels 2 and 4 (iDose4 level 2 and 4). One of the authors (D.S) measured 20 consecutive images of the center region along the z-axis using a 4.0 mmdiameter region of interest (ROI) and calculated the mean value. Image noise was also measured at the center position using the inserted water object at a slice thickness of 1.0 mm, and the mean value was calculated from 20 consecutive images of the center region. Image noise was calculated as the root mean square value of the standard deviation of the CT numbers in the ROI of a 4.0-mm diameter. CNR values were calculated as follows:
CNR = (ROIm−ROIb)/SDb where ROIm was the mean CT number of the iodinated enhancing material, and ROIb and SDb are, respectively, the mean CT number of the water object and mean standard deviation of the CT numbers of the background [21,23]. In addition, the relative CNR was calculated by the CNR for each keV image divided by the CNR for 120-kVp images without denoising.
2. Materials and methods
2.4. Noise power spectrum (NPS)
2.1. Phantom For identifying the effect of iodine in different phantom sizes, two acrylic cylindrical phantoms with different size: small, 20 cm and 15 cm (diameter and height) and large, 30 cm and 15 cm were completely scanned. For our measurements, we made a concentration of iodinated enhancing material at 3.8 mg I/ml, 7.5 mg I/ml, and 15 mg I/ml (Iopamiron 300; Bayer Healthcare, Osaka). Each iodinated enhancing material was separately added along with distilled water to cylindrical silicon tubes of 1.0 cm (diameter) × 10 cm (length). The cylindrical silicon tube including the iodinated enhancing material or water was separately inserted into the small and large acrylic cylindrical phantoms in the same central position. A Catphan phantom (Phantom Laboratory, Cambridge, NY) with a CTP712 uniformity module was acquired to calculate the noise power spectrum (NPS) [21,22]. The diameter and height (length along the z-axis) of the phantom were 20.0 cm.
For the conventional 120-kVp images and each virtual monoenergetic image from 40, 70, and 100 keV, including denoising levels of 2 and 4 at slice thicknesses of 1.0 mm, 30 noise images along the z-axis were produced by subtracting the middle image. The 15 noise images from top slice levels and 15 from bottom slice levels (total 30 noise images) were chosen to avoid the noise correlation introduced by subtracting the middle image. There was 5-mm gap between the middle image and the first used image of top or bottom side. A central portion of the noise images with 256 × 256 pixels was chosen as the ROI, and the power spectrum of the ROI was obtained by the squared magnitude of the two-dimensional Fourier transform of the noisy ROI image divided by the physical area of the ROI [24,25]. The process was repeated for 30 noise images, and the mean of the spectrum over 30 noise realizations was calculated and defined as the NPS. In addition, the expected image noise for 120 kVp, 40 keV, and 70 keV was calculated by a square-root of the area-under-the-NPS-curve, which was then normalized against that of 70 keV.
2.2. Helical scanning and image reconstruction
3. Results
Helical scans were performed using a first-generation dual-layer spectral detector CT scanner (IQon Spectral CT; Philips Healthcare, Cleveland, OH). The single-tube voltage for obtaining virtual monoenergetic images was set to 120 or 140 kVp. Two acrylic cylindrical phantoms with different size and a Catphan phantom were completely scanned. The following scan parameters were used: detector configuration, 64 × 0.625 mm (detector collimation); gantry rotation time,
3.1. Iodine CT number Fig. 1 shows the iodine CT numbers from 40- to 200-keV virtual monoenergetic images at different iodinated enhancing and phantom sizes acquired at 120 kVp (Fig. 1a) and 140 kVp (Fig. 1b). The iodine CT number at different iodinated enhancing was similar with respect to phantom size and acquisition method. At 15.0 mg I/ml and 120-kVp 6
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
Fig. 1. Iodine CT numbers from 40 to 200 keV monoenergetic images for different iodinated enhancing materials and phantom sizes obtained at (a) 120 kVp and (b) 140 kVp.
acquisition, the iodine CT numbers for the 20- and 30-cm phantom sizes were 1210 and 1174 HU, respectively, for the 40-keV image and 381 and 372 HU, respectively, for the 70-keV image (Table 1). The iodine CT number at 120 kVp was 444 HU at 20-cm phantom size and 404 HU at 30 cm. Compared with the iodine CT number of the conventional 120-kVp image, those for the monoenergetic 40-, 50-, and 60-keV images increased by approximately 3.0-, 1.9-, and 1.3-fold (Fig. 2), respectively. The iodine CT numbers in the conventional 120- and 140kVp were close to that for the monoenergetic 66-keV image at 20-cm and the 68-keV image at 30-cm phantom size and to that for the monoenergetic 71-keV image at 20-cm and the 74-keV image at 30-cm phantom size, respectively. At 120-kVp acquisition the difference in CT numbers between the 20- and 30-cm phantom sizes was 3.0% for the 40-keV images and 2.4% for the 70-keV images (Table 1). In contrast,
Table 1 Difference of iodine CT number between phantom sizes at virtual monoenergetic and conventional tube-voltage images. Dual energy scan method (kVp)
Virtual monoenergetic and conventional tube voltage image (keV)
120
40 70
140
40 70
Conventional tube-voltage image (kVp)
Iodine CT number (HU)
Difference in CT number between phantom sizes (%)
20 cm
30 cm
120
1210 381 444
1174 372 404
3.0 2.4 9.0
140
1235 388 382
1195 371 338
3.2 4.4 11.5
Fig. 2. Images obtained at virtual monoenergetic energies of 40, 50, 60, and 70 keV and conventional tube-voltage energy of 120 kVp on 20-cm phantom size. The iodine concentration at 15 mg I/ml was represented in each image. The window width and level were 1200 HU and 200 HU. 7
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
Table 2 Image noise at different phantom sizes obtained at virtual monoenergetic and conventional tube-voltage images. Phantom size
20 cm
Scan technique
120 kVp
140 kVp
30 cm
120 kVp
140 kVp
Denoising level
w/o Level Level w/o Level Level w/o Level Level w/o Level Level
2 4 2 4 2 4 2 4
kVp image (HU)
Virtual monoenergetic image (HU) 40
50
60
70
80
90
100
120
140
160
180
200
25.5 21.5 17.7 26.3 22.0 18.6
23.5 20.2 17.5 23.1 19.2 16.5
21.8 18.1 15.2 21.9 18.1 15.2
21.2 17.4 14.4 21.4 17.6 14.7
21.1 17.2 14.2 21.2 17.4 14.5
20.9 17.2 14.2 21.1 17.3 14.4
20.9 17.1 14.1 21.0 17.3 14.5
20.9 17.2 14.2 21.0 17.3 14.5
21.0 17.3 14.3 21.0 17.4 14.5
21.0 17.3 14.3 20.9 17.3 14.5
21.0 17.3 14.3 20.9 17.3 14.4
21.0 17.3 14.4 20.9 17.3 14.4
21.1 17.4 14.5 20.8 17.3 14.5
26.5 22.1 18.7 27.2 22.2 18.6
24.6 20.5 17.2 25.4 21.0 17.7
23.9 19.8 16.6 24.6 20.1 16.8
23.6 19.6 16.4 24.1 19.7 16.4
23.5 19.4 16.2 23.9 19.4 16.1
23.4 19.4 16.2 23.7 19.3 16.0
23.4 19.3 16.1 23.6 19.2 16.0
23.3 19.2 16.1 23.6 19.2 15.9
23.3 19.2 16.1 23.5 19.1 15.8
23.3 19.3 16.1 23.5 19.1 15.8
23.4 19.3 16.1 23.4 19.0 15.8
23.4 19.3 16.0 23.4 19.0 15.7
23.3 19.2 16.0 23.4 19.0 15.7
frequency range, although the image noise magnitude differs substantially. In contrast, the NPS of the 40-keV image had more low-frequency components and was different from that of the 120-kVp image. The expected image noise was calculated by a square-root of the areaunder-the-NPS-curve, which was then normalized against that of 70 keV. The expected normalized noise was in a good agreement with the measured normalized noise (Table 2, image noise at 120 kVp without denoising using 20 cm phantom): 1.10 for the expected noise versus 1.11 for the measured noise, both for 40 keV, and 1.25 versus 1.21 for 120 kVp. When denoising was performed on 70-keV images at different strength levels, the NPS of the denoised images showed decreased magnitudes over the entire frequency range depending on the strength of the process (denoising level 0 had a weaker effect, and thus, had larger noise). Although the slight shift of the maximum NPS to lower frequencies was observed, the entire shape of the NPS curves remained unchanged (Fig. 5).
the difference in iodine CT numbers between the 20- and 30-cm phantom sizes was 9.0% for the conventional 120-kVp images and 11.5% for the 140-kVp images. 3.2. Image noise The image noise values for each virtual monoenergetic image without denoising were similar from 24.6 HU at 40 keV to 23.3 HU at 200 keV acquired at 120 kVp (Table 2) at the 30-cm phantom. At 140 kVp, the image noise also remained constant from 25.4 HU at 40 keV to 23.4 HU at 200 keV. With increasing denoising levels 2 and 4, the image noise entirely shifted to lower values. The image noise values for the conventional 120- and 140-kVp images were 26.5 and 27.2 HU, respectively. Compared with the image noise of the conventional 120kVp, the image noise of monochromatic 70-keV images was decreased by 11.3% without denoising and by 26.8% and 38.9% at denoising levels 2 and 4, respectively. In the case of the 20-cm phantom, the image noise acquired at 120kVp was 23.5 HU at 40 keV, 21.1 HU at 200 keV, 25.5 HU at 120 kVp (Table 2) and exhibited the same tendencies as those of images obtained at 140 kVp (23.1 HU at 40 keV, 20.8 HU at 200 keV, 26.3 HU at 140 kVp). Compared with the image noise of the conventional 120-kVp, image noise of monochromatic 70-keV images was decreased by 17.7% without denoising and by 32.6% and 44.3% at denoising levels 2 and 4, respectively.
4. Discussion In this study, we investigated the performance of virtual monoenergetic images using dual-layer spectral detector CT. Dual-layer spectral detector CT is the first commercially available detector-based CT that uses a single-tube voltage beam (120 or 140 kVp acquisitions). The dual-energy information and routinely used conventional tubevoltage images are retrospectively available for every acquisition at the same time and hence, no decision to perform dual-energy acquisition prior to the examination is needed. Furthermore, additional radiation dose and adjustments of scan parameters, such as FOV, scan range, and automatic exposure control settings, are not needed to obtain dualenergy information because the same acquisition method as that for conventional CT scans is used. The stable iodine CT numbers are obtained across the virtual monoenergetic images regardless of the difference in phantom sizes compared with conventional tube-voltage image. In our study, we compared the CT numbers between a smaller 20-cm phantom and a larger 30-cm phantom. At 15.0 mg I/ml and 120-kVp acquisition, the difference in the CT numbers between the small and large phantoms was 3.0% at monochromatic 40 keV and 2.4% at 70 keV in contrast to 9.0% at the conventional 120 kVp. In addition, the theoretical CT number was calculated as 1233 HU at 40 keV and 387 HU at 70 keV [26] and the difference in the CT number for 40 keV was 1.9% at 20-cm and 4.8% at 30-cm phantoms. The fact that the pixel values are less dependent on the object sizes indicates that the beam-hardening effects are minimized on virtual monoenergetic images. In addition, we observe no difference in CT numbers of iodine inserts with the same phantom size between the two acquisition methods (120 and 140 kVp). In contrast, van Hamersvelt et al. [10] reported a different result. By
3.3. Relative CNR Fig. 3 shows relative CNR for 15 mg I/ml at different phantom sizes. The highest relative CNR was obtained for the monochromatic 40-keV image and gradually decreased with increasing virtual monoenergetic images. The CNR at denoising level 4 was higher than that at denoising level 2 and without denoising because of the greater decrease in image noise. Compared with the CNR at 120 kVp without denoising, CNR values at level 4 were 4.7-, 3.1-, 2.1-, and 1.5-fold for the 30-cm phantom size (Fig. 3a) and 4.1-, 3.1-, 2.2-, and 1.5-fold for the 20-cm phantom size (Fig. 3b) on the 40-, 50-, 60-, and 70-keV images. In addition, relative CNR was almost 1.0 at 70 keV, regardless of phantom size. 3.4. Noise power spectrum (NPS) The NPS presented in Fig. 4 reflects the difference in the texture of image noise between the conventional 120-kVp and virtual monoenergetic images. The entire shape distribution (e.g., substantial increase in low-frequency magnitude and spatial frequency point at the maximum magnitude) for NPS of the 70-keV images with a 1.0-mm slice thickness was similar to that of the 120-kVp images over the entire 8
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
Fig. 3. Relative contrast-to-noise ratio (CNR) for virtual monoenergetic images at different phantom sizes of (a) 30 cm and (b) 20 cm obtained at 120 kVp.
Fig. 4. The noise power spectrum (NPS) from virtual monoenergetic images acquired at 40, 70, and 100 keV and conventional tube-voltage images at 120 kVp. The NPS of 40 keV was shifted to the low-frequency range for 1-mm slice thicknesses. The Catphan phantom with the uniformity module was used to calculate NPS.
Fig. 5. The NPS of conventional tube-voltage images for 120 kVp and virtual monoenergetic images for 70 keV at different denoising levels. The NPS levels for a slice thickness of 1 mm entirely decreased with increasing denoising levels of 2 and 4. The Catphan phantom with the uniformity module was used to calculate NPS.
scanning gadolinium solutions using 140 and 120 kVp, 140 kVp acquisitions resulted in higher CT numbers at monochromatic 40-keV images than for 120-kVp acquisitions: the difference in CT numbers between 120- and 140-kVp acquisitions was as large as 1250 HU (15.7–26.3 mg/ml) at 40 keV. We have confirmed van Hamersvelt’s finding in a separate study [10]. The mechanism for the discrepancy is not clear at this moment, although it seems logical to think that the major reason may be the difference between iodine and gadolinium. Fig. 3 of Ref. [8] showed that the difference between 120 kVp and 140 kVp significantly increased below 60 keV, which is near and below the K-edge of gadolinium at 50.2 keV. It is also possible that the X-ray intensities detected at the two layers become unbalanced when the attenuation is high—higher energy windows (mostly > 50 keV) may suffer from severer photon starvations with gadolinium than with
iodine—and that resulted in the kVp-dependent enhancement. The image noise of virtual monoenergetic images was nearly constant throughout the image and can be reduced by using image smoothing. The NPS’s of 40-keV images had larger low-frequency components and were significantly different from the NPS of 120 kVp images acquired at the same condition. We believe that it is primarily because of the anti-correlation noise reduction algorithm, which is a noise reduction method that uses the negative correlation between the photoelectric effect and the Compton scattering effect. There is a potential risk of altering the granularity and texture of images, which may affect the clinical value of the images. Therefore, careful assessment of the clinical task-specific image quality is required. In general, an intrinsic tradeoff exists between iodine contrast and image noise. Although iodine contrast is maximized at lower energies, image noise is 9
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
also maximized at lower energies. However, image noise in monoenergetic 40-keV images has been found to be 5-fold higher than that in 70-keV monoenergetic images using different acquisition techniques [15,16]. To decrease the image noise in lower monoenergetic images, energy domain noise reduction is applied using another dual-energy technique [16]. Dual-layer spectral detector CT utilizes a unique statistical property of noise, i.e. anti-correlation of noise between two spectral base images [17,18] for properly identifying and reducing the overall noise level in available spectral results by applying a dedicated spectral reconstruction algorithm which operates in the image and projection domains (post-decomposition), the latter being possible only in a projection-based decomposition technology (with a perfect alignment raw data sets) [19]. In addition, the obtained virtual monoenergetic images are also available to reduce the remaining quantum noise using a denoising tool (see Table 2, levels 2 and 4). Dual-layer spectral detector CT achieves CNR improvements because the level of image noise does not change much. Clinically, when poor iodine enhancements are obtained on conventional 120-kVp images, such as those in CT angiographic examinations, dynamic contrast-enhanced CT examinations, and unexpected patient conditions, lower monoenergetic images are helpful because of the increase in iodine enhancements in various situations, which can help avoid misdiagnosis. In addition, dual-layer spectral detector CT allows decrease in the iodine load using lower virtual monoenergetic images and our institution achieves a 50% reduction in the iodine load during multiphasic hepatic CT without degradation of the image quality because of a stable image noise [27]. Our study had some limitations. First, we focused on measuring physical performance indices and did not discuss the effects on clinical tasks. As stated earlier, we plan to perform observer performance tests, including low-contrast detectability, image appearance, and accuracy of diagnosis, using lower monoenergetic images and compare them with conventional images. Next, because we focused on iodine enhancements useful in clinical situations, we did not study iodine concentrations higher than 1500 HU for monoenergetic 40-keV images, which corresponds to > 500 HU at 120 kVp. In conclusion, using the dual-layer spectral detector CT, virtual monoenergetic images provide stable image noise over the entire energy spectrum and improve the contrast-to-noise ratio than conventional tube voltage.
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17] [18]
[19]
[20]
[21]
[22]
[23]
References [1] McCollough CH, Leng S, Yu L, Fletcher JG. Dual- and multi-energy CT: principles, technical approaches, and clinical applications. Radiology 2015;276:637–53. [2] Johnson TR. Dual-energy CT: general principles. Am J Roentgenol 2012;199:S3–8. [3] Kaza RK, Platt JF, Cohan RH, Caoili EM, Al-Hawary MM, Wasnik A. Dual-energy CT with single- and dual-source scanners: current applications in evaluating the genitourinary tract. Radiographics 2012;32:353–69. [4] Almeida IP, Schyns LE, Ollers MC, van Elmpt W, Parodi K, Landry G, et al. Dualenergy CT quantitative imaging: a comparison study between twin-beam and dualsource CT scanners. Med Phys 2017;44:171–9. [5] Faby S, Kuchenbecker S, Sawall S, Simons D, Schlemmer HP, Lell M, et al. Performance of today's dual energy CT and future multi energy CT in virtual noncontrast imaging and in iodine quantification: a simulation study. Med Phys 2015;42:4349–66. [6] Ju Y, Liu A, Dong Y, Liu Y, Wang H, Sun M, et al. The value of nonenhanced single-
[24]
[25]
[26]
[27]
10
source dual-energy CT for differentiating metastases from adenoma in adrenal glands. Acad Radiol 2015;22:834–9. Gonzalez-Perez V, Arana E, Barrios M, Bartres A, Cruz J, Montero R, et al. Differentiation of benign and malignant lung lesions: dual-energy computed tomography findings. Eur J Radiol 2016;85:1765–72. Mullner M, Schlattl H, Hoeschen C, Dietrich O. Feasibility of spectral CT imaging for the detection of liver lesions with gold-based contrast agents – a simulation study. Phys Med 2015;31:875–81. Nasirudin RA, Penchev P, Mei K, Rummeny EJ, Fiebich M, Noel PB. A Monte Carlo software bench for simulation of spectral k-edge CT imaging: initial results. Phys Med 2015;31:398–405. van Hamersvelt RW, Willemink MJ, de Jong PA, Milles J, Vlassenbroek A, Schilham AM, et al. Feasibility and accuracy of dual-layer spectral detector computed tomography for quantification of gadolinium: a phantom study. Eur Radiol 2017. Hickethier T, Baessler B, Kroeger JR, Doerner J, Pahn G, Maintz D, et al. Monoenergetic reconstructions for imaging of coronary artery stents using spectral detector CT: in-vitro experience and comparison to conventional images. J Cardiovasc Comput Tomogr 2017;11:33–9. Pelgrim GJ, van Hamersvelt RW, Willemink MJ, Schmidt BT, Flohr T, Schilham A, et al. Accuracy of iodine quantification using dual energy CT in latest generation dual source and dual layer CT. Eur Radiol 2017. Wellenberg RH, Boomsma MF, van Osch JA, Vlassenbroek A, Milles J, Edens MA, et al. Quantifying metal artefact reduction using virtual monochromatic dual-layer detector spectral CT imaging in unilateral and bilateral total hip prostheses. Eur J Radiol 2017;88:61–70. Ananthakrishnan L, Rajiah P, Ahn R, Rassouli N, Xi Y, Soesbe TC, et al. Spectral detector CT-derived virtual non-contrast images: comparison of attenuation values with unenhanced CT. Abdom Radiol (NY) 2017;42:702–9. Matsumoto K, Jinzaki M, Tanami Y, Ueno A, Yamada M, Kuribayashi S. Virtual monochromatic spectral imaging with fast kilovoltage switching: improved image quality as compared with that obtained with conventional 120-kVp CT. Radiology 2011;259:257–62. Leng S, Yu L, Fletcher JG, McCollough CH. Maximizing iodine contrast-to-noise ratios in abdominal CT imaging through use of energy domain noise reduction and virtual monoenergetic dual-energy CT. Radiology 2015;276:562–70. Kalender WA, Klotz E, Kostaridou L. An algorithm for noise suppression in dual energy CT material density images. IEEE Trans Med Imaging 1988;7:218–24. Brown KM GS, Zabic S. Impact of spectral separation in dual-energy CT with anticorrelated statistical reconstruction. In The 13th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine. 2015:491–4. Maass C, Baer M, Kachelriess M. Image-based dual energy CT using optimized precorrection functions: a practical new approach of material decomposition in image domain. Med Phys 2009;36:3818–29. Yu L, Christner JA, Leng S, Wang J, Fletcher JG, McCollough CH. Virtual monochromatic imaging in dual-source dual-energy CT: radiation dose and image quality. Med Phys 2011;38:6371–9. Gupta AK, Nelson RC, Johnson GA, Paulson EK, Delong DM, Yoshizumi TT. Optimization of eight-element multi-detector row helical CT technology for evaluation of the abdomen. Radiology 2003;227:739–45. Verdun FR, Denys A, Valley JF, Schnyder P, Meuli RA. Detection of low-contrast objects: experimental comparison of single- and multi-detector row CT with a phantom. Radiology 2002;223:426–31. Brix G, Lechel U, Sudarski S, Trumm C, Henzler T. Spectral optimization of iodineenhanced CT: quantifying the effect of tube voltage on image quality and radiation exposure determined at an anthropomorphic phantom. Phys Med 2016;32:999–1006. Boedeker KL, Cooper VN, McNitt-Gray MF. Application of the noise power spectrum in modern diagnostic MDCT: part I. Measurement of noise power spectra and noise equivalent quanta. Phys Med Biol 2007;52:4027–46. Boedeker KL, McNitt-Gray MF. Application of the noise power spectrum in modern diagnostic MDCT: part II. Noise power spectra and signal to noise. Phys Med Biol 2007;52:4047–61. Goodsitt MM, Christodoulou EG, Larson SC. Accuracies of the synthesized monochromatic CT numbers and effective atomic numbers obtained with a rapid kVp switching dual energy CT scanner. Med Phys 2011;38:2222–32. Nagayama Y, Nakaura T, Oda S, Utsunomiya D, Funama Y, Iyama Y, et al. Duallayer DECT for multiphasic hepatic CT with 50 percent iodine load: a matched-pair comparison with a 120 kVp protocol. Eur Radiol 2017.
Contents lists available at ScienceDirect
Physica Medica journal homepage: www.elsevier.com/locate/ejmp
Original paper
Image quality characteristics for virtual monoenergetic images using duallayer spectral detector CT: Comparison with conventional tube-voltage images
T
⁎
Daisuke Sakabea,b, Yoshinori Funamac, , Katsuyuki Taguchid, Takeshi Nakaurae, Daisuke Utsunomiyae, Seitaro Odae, Masafumi Kidohe, Yasunori Nagayamae, Yasuyuki Yamashitae a
Graduate School of Health Sciences, Kumamoto University, Kumamoto, Japan Department of Radiology, Kumamoto University Hospital, Kumamoto, Japan Department of Medical Physics, Faculty of Life Sciences, Kumamoto University, Kumamoto, Japan d The Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, USA e Department of Diagnostic Radiology, Faculty of Life Sciences, Kumamoto University, Kumamoto, Japan b c
A R T I C LE I N FO
A B S T R A C T
Keywords: Dual-layer spectral detector CT Virtual monoenergetic image Image quality Iodine Different object sizes
Purpose: To investigate the image quality characteristics for virtual monoenergetic images compared with conventional tube-voltage image with dual-layer spectral CT (DLCT). Methods: Helical scans were performed using a first-generation DLCT scanner, two different sizes of acrylic cylindrical phantoms, and a Catphan phantom. Three different iodine concentrations were inserted into the phantom center. The single-tube voltage for obtaining virtual monoenergetic images was set to 120 or 140 kVp. Conventional 120- and 140-kVp images and virtual monoenergetic images (40–200-keV images) were reconstructed from slice thicknesses of 1.0 mm. The CT number and image noise were measured for each iodine concentration and water on the 120-kVp images and virtual monoenergetic images. The noise power spectrum (NPS) was also calculated. Results: The iodine CT numbers for the iodinated enhancing materials were similar regardless of phantom size and acquisition method. Compared with the iodine CT numbers of the conventional 120-kVp images, those for the monoenergetic 40-, 50-, and 60-keV images increased by approximately 3.0-, 1.9-, and 1.3-fold, respectively. The image noise values for each virtual monoenergetic image were similar (for example, 24.6 HU at 40 keV and 23.3 HU at 200 keV obtained at 120 kVp and 30-cm phantom size). The NPS curves of the 70-keV and 120-kVp images for a 1.0-mm slice thickness over the entire frequency range were similar. Conclusion: Virtual monoenergetic images represent stable image noise over the entire energy spectrum and improved the contrast-to-noise ratio than conventional tube voltage using the dual-layer spectral detector CT.
1. Introduction Dual-energy computed tomography (CT) is performed using different scan techniques, including dual-spin, tube potential switching, and dual-source beam techniques, and two different X-ray spectrums are acquired [1–3]. These techniques have been used in valuable clinical applications, such as virtual monoenergetic imaging, iodine mapping, determination of effective atomic number, and measurement of electron density [4–9]. Recently, a raw data-based dual-energy CT called dual-layer spectral detector CT has become available [10–14] as a new device. Unlike
⁎
tube voltage beam method, dual-layer spectral detector CT is the first commercially available detector-based CT that uses a single-tube voltage beam. With dual-layer spectral detector CT, low-energy photons are absorbed by the first detector layer, and high-energy photons are absorbed by the second. Therefore, dual-layer spectral detector CT not only provides valuable dual-energy information but is also routinely used in conventional tube-voltage images (e.g., 120 or 140 kVp) simultaneously. So far, substantially increased image noise in lower monoenergetic images (e.g., 40–50 keV) has been a critical issue in clinical situations because increasing image noise degrades image quality, and
Corresponding author at: Department of Medical Physics, Faculty of Life Sciences, Kumamoto University, 4-24-1 Kuhonji, Kumamoto 862-0976, Japan. E-mail address: [email protected] (Y. Funama).
https://doi.org/10.1016/j.ejmp.2018.04.388 Received 31 October 2017; Received in revised form 9 April 2018; Accepted 10 April 2018 1120-1797/ © 2018 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
0.75 s; beam pitch, 0.8; and display field-of-view (FOV), 22.0 cm for 20cm acrylic cylindrical and Catphan phantoms and 32 cm for a 30-cm acrylic cylindrical phantom. The tube current-time product was set to 110 mAs at 120 kVp, 80 mAs at 140 kVp for a 20-cm acrylic cylindrical phantoms and 800 mAs at 120 kVp, 550 mAs at 140 kVp for a 30-cm acrylic cylindrical phantom. Conventional 120- and 140-kVp images and virtual monoenergetic images (40–200-keV images) were reconstructed at slice thicknesses of 1.0 and at a slice interval of 1.0 mm with an abdomen standard kernel (C). In addition, the virtual monoenergetic images were also reconstructed using corresponding denoising levels 2 and 4(iDose4 level 2 and 4), which decreased the image noise with increasing level.
deteriorated virtual monoenergetic images cannot provide a sustainable clinical image [15,16]. To decrease image noise of virtual monoenergetic images, dual-layer spectral detector CT introduces an “anticorrelated statistical reconstruction algorithm” in spectral reconstruction [17,18] which operates in the image and projection domains (postdecomposition), the latter being possible only in a projection-based decomposition technology (with a perfect alignment raw data sets) [19]. Then the method starts with converting a pair of high and low energy datasets obtained from dual-layer detector to Compton scatter and photoelectric data. It then performs spectral reconstruction, combining the following three points: the first point is maximizing a regularized log-likelihood. The second point is based on the fact that the noise of Compton and photoelectric data were negatively correlated. The noise can be effectively decreased by canceling out the correlated noise. The third point is using a synthesized data via a weighted sum of Compton and photoelectric data. This algorithm is effective in suppressing increased image noise by performing basis material decomposition on the spectral data. The spectral reconstruction for dual-layer CT made the amount of noise rather consistent regardless of the effective energies of monoenergetic keV images. Therefore, decreased image noise is expected in virtual monoenergetic images relative to the noise in other dual-energy mode images. In addition, a stable iodine CT number is also expected for virtual monoenergetic images regardless of the object size and acquisition method (120 or 140 kVp), and even conventional polychromatic image noise is decreased with increasing object size [4,20]. The improvement in the image quality with duallayer spectral detector CT is expected to be routinely utilized for poor vessel enhancement, low iodine-contrast detectability, and so on. However, the image quality of dual-layer spectral detector CT virtual monoenergetic images is uncertain with respect to specific characteristics, such as image noise, iodine enhancement at different object sizes, and improvement of the contrast-to-noise ratio. The purpose of this study was to investigate the image quality characteristics for comparison of virtual monoenergetic images with conventional tube-voltage images in a dual-layer spectral detector CT scanner.
2.3. Measurement of the CT number, image noise, and CNR Using a slice thickness of 1.0 mm, the CT number was measured for each iodinated enhancing material in the acrylic cylindrical phantom on conventional 120- and 140-kVp, and virtual monoenergetic images (40–200 keV), including those images with denoising levels 2 and 4 (iDose4 level 2 and 4). One of the authors (D.S) measured 20 consecutive images of the center region along the z-axis using a 4.0 mmdiameter region of interest (ROI) and calculated the mean value. Image noise was also measured at the center position using the inserted water object at a slice thickness of 1.0 mm, and the mean value was calculated from 20 consecutive images of the center region. Image noise was calculated as the root mean square value of the standard deviation of the CT numbers in the ROI of a 4.0-mm diameter. CNR values were calculated as follows:
CNR = (ROIm−ROIb)/SDb where ROIm was the mean CT number of the iodinated enhancing material, and ROIb and SDb are, respectively, the mean CT number of the water object and mean standard deviation of the CT numbers of the background [21,23]. In addition, the relative CNR was calculated by the CNR for each keV image divided by the CNR for 120-kVp images without denoising.
2. Materials and methods
2.4. Noise power spectrum (NPS)
2.1. Phantom For identifying the effect of iodine in different phantom sizes, two acrylic cylindrical phantoms with different size: small, 20 cm and 15 cm (diameter and height) and large, 30 cm and 15 cm were completely scanned. For our measurements, we made a concentration of iodinated enhancing material at 3.8 mg I/ml, 7.5 mg I/ml, and 15 mg I/ml (Iopamiron 300; Bayer Healthcare, Osaka). Each iodinated enhancing material was separately added along with distilled water to cylindrical silicon tubes of 1.0 cm (diameter) × 10 cm (length). The cylindrical silicon tube including the iodinated enhancing material or water was separately inserted into the small and large acrylic cylindrical phantoms in the same central position. A Catphan phantom (Phantom Laboratory, Cambridge, NY) with a CTP712 uniformity module was acquired to calculate the noise power spectrum (NPS) [21,22]. The diameter and height (length along the z-axis) of the phantom were 20.0 cm.
For the conventional 120-kVp images and each virtual monoenergetic image from 40, 70, and 100 keV, including denoising levels of 2 and 4 at slice thicknesses of 1.0 mm, 30 noise images along the z-axis were produced by subtracting the middle image. The 15 noise images from top slice levels and 15 from bottom slice levels (total 30 noise images) were chosen to avoid the noise correlation introduced by subtracting the middle image. There was 5-mm gap between the middle image and the first used image of top or bottom side. A central portion of the noise images with 256 × 256 pixels was chosen as the ROI, and the power spectrum of the ROI was obtained by the squared magnitude of the two-dimensional Fourier transform of the noisy ROI image divided by the physical area of the ROI [24,25]. The process was repeated for 30 noise images, and the mean of the spectrum over 30 noise realizations was calculated and defined as the NPS. In addition, the expected image noise for 120 kVp, 40 keV, and 70 keV was calculated by a square-root of the area-under-the-NPS-curve, which was then normalized against that of 70 keV.
2.2. Helical scanning and image reconstruction
3. Results
Helical scans were performed using a first-generation dual-layer spectral detector CT scanner (IQon Spectral CT; Philips Healthcare, Cleveland, OH). The single-tube voltage for obtaining virtual monoenergetic images was set to 120 or 140 kVp. Two acrylic cylindrical phantoms with different size and a Catphan phantom were completely scanned. The following scan parameters were used: detector configuration, 64 × 0.625 mm (detector collimation); gantry rotation time,
3.1. Iodine CT number Fig. 1 shows the iodine CT numbers from 40- to 200-keV virtual monoenergetic images at different iodinated enhancing and phantom sizes acquired at 120 kVp (Fig. 1a) and 140 kVp (Fig. 1b). The iodine CT number at different iodinated enhancing was similar with respect to phantom size and acquisition method. At 15.0 mg I/ml and 120-kVp 6
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
Fig. 1. Iodine CT numbers from 40 to 200 keV monoenergetic images for different iodinated enhancing materials and phantom sizes obtained at (a) 120 kVp and (b) 140 kVp.
acquisition, the iodine CT numbers for the 20- and 30-cm phantom sizes were 1210 and 1174 HU, respectively, for the 40-keV image and 381 and 372 HU, respectively, for the 70-keV image (Table 1). The iodine CT number at 120 kVp was 444 HU at 20-cm phantom size and 404 HU at 30 cm. Compared with the iodine CT number of the conventional 120-kVp image, those for the monoenergetic 40-, 50-, and 60-keV images increased by approximately 3.0-, 1.9-, and 1.3-fold (Fig. 2), respectively. The iodine CT numbers in the conventional 120- and 140kVp were close to that for the monoenergetic 66-keV image at 20-cm and the 68-keV image at 30-cm phantom size and to that for the monoenergetic 71-keV image at 20-cm and the 74-keV image at 30-cm phantom size, respectively. At 120-kVp acquisition the difference in CT numbers between the 20- and 30-cm phantom sizes was 3.0% for the 40-keV images and 2.4% for the 70-keV images (Table 1). In contrast,
Table 1 Difference of iodine CT number between phantom sizes at virtual monoenergetic and conventional tube-voltage images. Dual energy scan method (kVp)
Virtual monoenergetic and conventional tube voltage image (keV)
120
40 70
140
40 70
Conventional tube-voltage image (kVp)
Iodine CT number (HU)
Difference in CT number between phantom sizes (%)
20 cm
30 cm
120
1210 381 444
1174 372 404
3.0 2.4 9.0
140
1235 388 382
1195 371 338
3.2 4.4 11.5
Fig. 2. Images obtained at virtual monoenergetic energies of 40, 50, 60, and 70 keV and conventional tube-voltage energy of 120 kVp on 20-cm phantom size. The iodine concentration at 15 mg I/ml was represented in each image. The window width and level were 1200 HU and 200 HU. 7
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
Table 2 Image noise at different phantom sizes obtained at virtual monoenergetic and conventional tube-voltage images. Phantom size
20 cm
Scan technique
120 kVp
140 kVp
30 cm
120 kVp
140 kVp
Denoising level
w/o Level Level w/o Level Level w/o Level Level w/o Level Level
2 4 2 4 2 4 2 4
kVp image (HU)
Virtual monoenergetic image (HU) 40
50
60
70
80
90
100
120
140
160
180
200
25.5 21.5 17.7 26.3 22.0 18.6
23.5 20.2 17.5 23.1 19.2 16.5
21.8 18.1 15.2 21.9 18.1 15.2
21.2 17.4 14.4 21.4 17.6 14.7
21.1 17.2 14.2 21.2 17.4 14.5
20.9 17.2 14.2 21.1 17.3 14.4
20.9 17.1 14.1 21.0 17.3 14.5
20.9 17.2 14.2 21.0 17.3 14.5
21.0 17.3 14.3 21.0 17.4 14.5
21.0 17.3 14.3 20.9 17.3 14.5
21.0 17.3 14.3 20.9 17.3 14.4
21.0 17.3 14.4 20.9 17.3 14.4
21.1 17.4 14.5 20.8 17.3 14.5
26.5 22.1 18.7 27.2 22.2 18.6
24.6 20.5 17.2 25.4 21.0 17.7
23.9 19.8 16.6 24.6 20.1 16.8
23.6 19.6 16.4 24.1 19.7 16.4
23.5 19.4 16.2 23.9 19.4 16.1
23.4 19.4 16.2 23.7 19.3 16.0
23.4 19.3 16.1 23.6 19.2 16.0
23.3 19.2 16.1 23.6 19.2 15.9
23.3 19.2 16.1 23.5 19.1 15.8
23.3 19.3 16.1 23.5 19.1 15.8
23.4 19.3 16.1 23.4 19.0 15.8
23.4 19.3 16.0 23.4 19.0 15.7
23.3 19.2 16.0 23.4 19.0 15.7
frequency range, although the image noise magnitude differs substantially. In contrast, the NPS of the 40-keV image had more low-frequency components and was different from that of the 120-kVp image. The expected image noise was calculated by a square-root of the areaunder-the-NPS-curve, which was then normalized against that of 70 keV. The expected normalized noise was in a good agreement with the measured normalized noise (Table 2, image noise at 120 kVp without denoising using 20 cm phantom): 1.10 for the expected noise versus 1.11 for the measured noise, both for 40 keV, and 1.25 versus 1.21 for 120 kVp. When denoising was performed on 70-keV images at different strength levels, the NPS of the denoised images showed decreased magnitudes over the entire frequency range depending on the strength of the process (denoising level 0 had a weaker effect, and thus, had larger noise). Although the slight shift of the maximum NPS to lower frequencies was observed, the entire shape of the NPS curves remained unchanged (Fig. 5).
the difference in iodine CT numbers between the 20- and 30-cm phantom sizes was 9.0% for the conventional 120-kVp images and 11.5% for the 140-kVp images. 3.2. Image noise The image noise values for each virtual monoenergetic image without denoising were similar from 24.6 HU at 40 keV to 23.3 HU at 200 keV acquired at 120 kVp (Table 2) at the 30-cm phantom. At 140 kVp, the image noise also remained constant from 25.4 HU at 40 keV to 23.4 HU at 200 keV. With increasing denoising levels 2 and 4, the image noise entirely shifted to lower values. The image noise values for the conventional 120- and 140-kVp images were 26.5 and 27.2 HU, respectively. Compared with the image noise of the conventional 120kVp, the image noise of monochromatic 70-keV images was decreased by 11.3% without denoising and by 26.8% and 38.9% at denoising levels 2 and 4, respectively. In the case of the 20-cm phantom, the image noise acquired at 120kVp was 23.5 HU at 40 keV, 21.1 HU at 200 keV, 25.5 HU at 120 kVp (Table 2) and exhibited the same tendencies as those of images obtained at 140 kVp (23.1 HU at 40 keV, 20.8 HU at 200 keV, 26.3 HU at 140 kVp). Compared with the image noise of the conventional 120-kVp, image noise of monochromatic 70-keV images was decreased by 17.7% without denoising and by 32.6% and 44.3% at denoising levels 2 and 4, respectively.
4. Discussion In this study, we investigated the performance of virtual monoenergetic images using dual-layer spectral detector CT. Dual-layer spectral detector CT is the first commercially available detector-based CT that uses a single-tube voltage beam (120 or 140 kVp acquisitions). The dual-energy information and routinely used conventional tubevoltage images are retrospectively available for every acquisition at the same time and hence, no decision to perform dual-energy acquisition prior to the examination is needed. Furthermore, additional radiation dose and adjustments of scan parameters, such as FOV, scan range, and automatic exposure control settings, are not needed to obtain dualenergy information because the same acquisition method as that for conventional CT scans is used. The stable iodine CT numbers are obtained across the virtual monoenergetic images regardless of the difference in phantom sizes compared with conventional tube-voltage image. In our study, we compared the CT numbers between a smaller 20-cm phantom and a larger 30-cm phantom. At 15.0 mg I/ml and 120-kVp acquisition, the difference in the CT numbers between the small and large phantoms was 3.0% at monochromatic 40 keV and 2.4% at 70 keV in contrast to 9.0% at the conventional 120 kVp. In addition, the theoretical CT number was calculated as 1233 HU at 40 keV and 387 HU at 70 keV [26] and the difference in the CT number for 40 keV was 1.9% at 20-cm and 4.8% at 30-cm phantoms. The fact that the pixel values are less dependent on the object sizes indicates that the beam-hardening effects are minimized on virtual monoenergetic images. In addition, we observe no difference in CT numbers of iodine inserts with the same phantom size between the two acquisition methods (120 and 140 kVp). In contrast, van Hamersvelt et al. [10] reported a different result. By
3.3. Relative CNR Fig. 3 shows relative CNR for 15 mg I/ml at different phantom sizes. The highest relative CNR was obtained for the monochromatic 40-keV image and gradually decreased with increasing virtual monoenergetic images. The CNR at denoising level 4 was higher than that at denoising level 2 and without denoising because of the greater decrease in image noise. Compared with the CNR at 120 kVp without denoising, CNR values at level 4 were 4.7-, 3.1-, 2.1-, and 1.5-fold for the 30-cm phantom size (Fig. 3a) and 4.1-, 3.1-, 2.2-, and 1.5-fold for the 20-cm phantom size (Fig. 3b) on the 40-, 50-, 60-, and 70-keV images. In addition, relative CNR was almost 1.0 at 70 keV, regardless of phantom size. 3.4. Noise power spectrum (NPS) The NPS presented in Fig. 4 reflects the difference in the texture of image noise between the conventional 120-kVp and virtual monoenergetic images. The entire shape distribution (e.g., substantial increase in low-frequency magnitude and spatial frequency point at the maximum magnitude) for NPS of the 70-keV images with a 1.0-mm slice thickness was similar to that of the 120-kVp images over the entire 8
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
Fig. 3. Relative contrast-to-noise ratio (CNR) for virtual monoenergetic images at different phantom sizes of (a) 30 cm and (b) 20 cm obtained at 120 kVp.
Fig. 4. The noise power spectrum (NPS) from virtual monoenergetic images acquired at 40, 70, and 100 keV and conventional tube-voltage images at 120 kVp. The NPS of 40 keV was shifted to the low-frequency range for 1-mm slice thicknesses. The Catphan phantom with the uniformity module was used to calculate NPS.
Fig. 5. The NPS of conventional tube-voltage images for 120 kVp and virtual monoenergetic images for 70 keV at different denoising levels. The NPS levels for a slice thickness of 1 mm entirely decreased with increasing denoising levels of 2 and 4. The Catphan phantom with the uniformity module was used to calculate NPS.
scanning gadolinium solutions using 140 and 120 kVp, 140 kVp acquisitions resulted in higher CT numbers at monochromatic 40-keV images than for 120-kVp acquisitions: the difference in CT numbers between 120- and 140-kVp acquisitions was as large as 1250 HU (15.7–26.3 mg/ml) at 40 keV. We have confirmed van Hamersvelt’s finding in a separate study [10]. The mechanism for the discrepancy is not clear at this moment, although it seems logical to think that the major reason may be the difference between iodine and gadolinium. Fig. 3 of Ref. [8] showed that the difference between 120 kVp and 140 kVp significantly increased below 60 keV, which is near and below the K-edge of gadolinium at 50.2 keV. It is also possible that the X-ray intensities detected at the two layers become unbalanced when the attenuation is high—higher energy windows (mostly > 50 keV) may suffer from severer photon starvations with gadolinium than with
iodine—and that resulted in the kVp-dependent enhancement. The image noise of virtual monoenergetic images was nearly constant throughout the image and can be reduced by using image smoothing. The NPS’s of 40-keV images had larger low-frequency components and were significantly different from the NPS of 120 kVp images acquired at the same condition. We believe that it is primarily because of the anti-correlation noise reduction algorithm, which is a noise reduction method that uses the negative correlation between the photoelectric effect and the Compton scattering effect. There is a potential risk of altering the granularity and texture of images, which may affect the clinical value of the images. Therefore, careful assessment of the clinical task-specific image quality is required. In general, an intrinsic tradeoff exists between iodine contrast and image noise. Although iodine contrast is maximized at lower energies, image noise is 9
Physica Medica 49 (2018) 5–10
D. Sakabe et al.
also maximized at lower energies. However, image noise in monoenergetic 40-keV images has been found to be 5-fold higher than that in 70-keV monoenergetic images using different acquisition techniques [15,16]. To decrease the image noise in lower monoenergetic images, energy domain noise reduction is applied using another dual-energy technique [16]. Dual-layer spectral detector CT utilizes a unique statistical property of noise, i.e. anti-correlation of noise between two spectral base images [17,18] for properly identifying and reducing the overall noise level in available spectral results by applying a dedicated spectral reconstruction algorithm which operates in the image and projection domains (post-decomposition), the latter being possible only in a projection-based decomposition technology (with a perfect alignment raw data sets) [19]. In addition, the obtained virtual monoenergetic images are also available to reduce the remaining quantum noise using a denoising tool (see Table 2, levels 2 and 4). Dual-layer spectral detector CT achieves CNR improvements because the level of image noise does not change much. Clinically, when poor iodine enhancements are obtained on conventional 120-kVp images, such as those in CT angiographic examinations, dynamic contrast-enhanced CT examinations, and unexpected patient conditions, lower monoenergetic images are helpful because of the increase in iodine enhancements in various situations, which can help avoid misdiagnosis. In addition, dual-layer spectral detector CT allows decrease in the iodine load using lower virtual monoenergetic images and our institution achieves a 50% reduction in the iodine load during multiphasic hepatic CT without degradation of the image quality because of a stable image noise [27]. Our study had some limitations. First, we focused on measuring physical performance indices and did not discuss the effects on clinical tasks. As stated earlier, we plan to perform observer performance tests, including low-contrast detectability, image appearance, and accuracy of diagnosis, using lower monoenergetic images and compare them with conventional images. Next, because we focused on iodine enhancements useful in clinical situations, we did not study iodine concentrations higher than 1500 HU for monoenergetic 40-keV images, which corresponds to > 500 HU at 120 kVp. In conclusion, using the dual-layer spectral detector CT, virtual monoenergetic images provide stable image noise over the entire energy spectrum and improve the contrast-to-noise ratio than conventional tube voltage.
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17] [18]
[19]
[20]
[21]
[22]
[23]
References [1] McCollough CH, Leng S, Yu L, Fletcher JG. Dual- and multi-energy CT: principles, technical approaches, and clinical applications. Radiology 2015;276:637–53. [2] Johnson TR. Dual-energy CT: general principles. Am J Roentgenol 2012;199:S3–8. [3] Kaza RK, Platt JF, Cohan RH, Caoili EM, Al-Hawary MM, Wasnik A. Dual-energy CT with single- and dual-source scanners: current applications in evaluating the genitourinary tract. Radiographics 2012;32:353–69. [4] Almeida IP, Schyns LE, Ollers MC, van Elmpt W, Parodi K, Landry G, et al. Dualenergy CT quantitative imaging: a comparison study between twin-beam and dualsource CT scanners. Med Phys 2017;44:171–9. [5] Faby S, Kuchenbecker S, Sawall S, Simons D, Schlemmer HP, Lell M, et al. Performance of today's dual energy CT and future multi energy CT in virtual noncontrast imaging and in iodine quantification: a simulation study. Med Phys 2015;42:4349–66. [6] Ju Y, Liu A, Dong Y, Liu Y, Wang H, Sun M, et al. The value of nonenhanced single-
[24]
[25]
[26]
[27]
10
source dual-energy CT for differentiating metastases from adenoma in adrenal glands. Acad Radiol 2015;22:834–9. Gonzalez-Perez V, Arana E, Barrios M, Bartres A, Cruz J, Montero R, et al. Differentiation of benign and malignant lung lesions: dual-energy computed tomography findings. Eur J Radiol 2016;85:1765–72. Mullner M, Schlattl H, Hoeschen C, Dietrich O. Feasibility of spectral CT imaging for the detection of liver lesions with gold-based contrast agents – a simulation study. Phys Med 2015;31:875–81. Nasirudin RA, Penchev P, Mei K, Rummeny EJ, Fiebich M, Noel PB. A Monte Carlo software bench for simulation of spectral k-edge CT imaging: initial results. Phys Med 2015;31:398–405. van Hamersvelt RW, Willemink MJ, de Jong PA, Milles J, Vlassenbroek A, Schilham AM, et al. Feasibility and accuracy of dual-layer spectral detector computed tomography for quantification of gadolinium: a phantom study. Eur Radiol 2017. Hickethier T, Baessler B, Kroeger JR, Doerner J, Pahn G, Maintz D, et al. Monoenergetic reconstructions for imaging of coronary artery stents using spectral detector CT: in-vitro experience and comparison to conventional images. J Cardiovasc Comput Tomogr 2017;11:33–9. Pelgrim GJ, van Hamersvelt RW, Willemink MJ, Schmidt BT, Flohr T, Schilham A, et al. Accuracy of iodine quantification using dual energy CT in latest generation dual source and dual layer CT. Eur Radiol 2017. Wellenberg RH, Boomsma MF, van Osch JA, Vlassenbroek A, Milles J, Edens MA, et al. Quantifying metal artefact reduction using virtual monochromatic dual-layer detector spectral CT imaging in unilateral and bilateral total hip prostheses. Eur J Radiol 2017;88:61–70. Ananthakrishnan L, Rajiah P, Ahn R, Rassouli N, Xi Y, Soesbe TC, et al. Spectral detector CT-derived virtual non-contrast images: comparison of attenuation values with unenhanced CT. Abdom Radiol (NY) 2017;42:702–9. Matsumoto K, Jinzaki M, Tanami Y, Ueno A, Yamada M, Kuribayashi S. Virtual monochromatic spectral imaging with fast kilovoltage switching: improved image quality as compared with that obtained with conventional 120-kVp CT. Radiology 2011;259:257–62. Leng S, Yu L, Fletcher JG, McCollough CH. Maximizing iodine contrast-to-noise ratios in abdominal CT imaging through use of energy domain noise reduction and virtual monoenergetic dual-energy CT. Radiology 2015;276:562–70. Kalender WA, Klotz E, Kostaridou L. An algorithm for noise suppression in dual energy CT material density images. IEEE Trans Med Imaging 1988;7:218–24. Brown KM GS, Zabic S. Impact of spectral separation in dual-energy CT with anticorrelated statistical reconstruction. In The 13th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine. 2015:491–4. Maass C, Baer M, Kachelriess M. Image-based dual energy CT using optimized precorrection functions: a practical new approach of material decomposition in image domain. Med Phys 2009;36:3818–29. Yu L, Christner JA, Leng S, Wang J, Fletcher JG, McCollough CH. Virtual monochromatic imaging in dual-source dual-energy CT: radiation dose and image quality. Med Phys 2011;38:6371–9. Gupta AK, Nelson RC, Johnson GA, Paulson EK, Delong DM, Yoshizumi TT. Optimization of eight-element multi-detector row helical CT technology for evaluation of the abdomen. Radiology 2003;227:739–45. Verdun FR, Denys A, Valley JF, Schnyder P, Meuli RA. Detection of low-contrast objects: experimental comparison of single- and multi-detector row CT with a phantom. Radiology 2002;223:426–31. Brix G, Lechel U, Sudarski S, Trumm C, Henzler T. Spectral optimization of iodineenhanced CT: quantifying the effect of tube voltage on image quality and radiation exposure determined at an anthropomorphic phantom. Phys Med 2016;32:999–1006. Boedeker KL, Cooper VN, McNitt-Gray MF. Application of the noise power spectrum in modern diagnostic MDCT: part I. Measurement of noise power spectra and noise equivalent quanta. Phys Med Biol 2007;52:4027–46. Boedeker KL, McNitt-Gray MF. Application of the noise power spectrum in modern diagnostic MDCT: part II. Noise power spectra and signal to noise. Phys Med Biol 2007;52:4047–61. Goodsitt MM, Christodoulou EG, Larson SC. Accuracies of the synthesized monochromatic CT numbers and effective atomic numbers obtained with a rapid kVp switching dual energy CT scanner. Med Phys 2011;38:2222–32. Nagayama Y, Nakaura T, Oda S, Utsunomiya D, Funama Y, Iyama Y, et al. Duallayer DECT for multiphasic hepatic CT with 50 percent iodine load: a matched-pair comparison with a 120 kVp protocol. Eur Radiol 2017.