ORIGINAL ARTICLE
Improving Image Quality for Lung Cancer Imaging With Optimal Monochromatic Energy Level in Dual Energy Spectral Computed Tomography Weishu Hou, MD,*† Xiangfang Sun, MD,‡ Yan Yin, MD,* Jiejun Cheng, MD,* Qing Zhang, MD,* Jianrong Xu, MD,* Yang Li, MD,* Wei Zhou, MD,* and Huawei Wu, MD* Objective: The aim of this study was to find optimal monochromatic spectral computed tomography (CT) level to improve image quality of lung cancer. Methods: Fifty patients with lung cancers were scanned by spectral CT; monochromatic images at 50, 60, 70 and 80 keV energy levels were generated; and objective analysis including image noise, lesion-to-lung contrastto-noise ratio, and CT number difference between central and peripheral regions of tumor (dCT value) were measured and compared. Subjective assessment about the overall image quality and inhomogeneity enhancement was compared. Results: The highest contrast-to-noise ratio value and subjective score of image quality were obtained at 70 keV, which were superior to those of 50- and 80-keV series (all P < 0.05). The subjective score of the inhomogeneity evaluation was peaked at 60-keV series and significantly higher than other energy levels (all P < 0.05). Conclusions: Both objective and subjective image analysis of lung cancers may be improved with the combined observation of 60 keV and 70 keV monochromatic images in spectral CT.
optimal energy level of monochromatic image in DESCT in patients with lung cancer.
MATERIALS AND METHODS Patients
Key Words: dual-energy CT, spectral imaging, lung cancer
Fifty-seven consecutive patients (34 men, 23 women; mean age, 62.1 years) with lung cancers underwent dual-phasic enhanced CT scans with DESCT mode from January 2011 to December 2013 were selected for the retrospective analysis. Seven patients were excluded because (a) the masses did not meet research requirement (round or oval shaped, solitary, no significant calcification or cavitation) (n = 5) and (b) unsatisfactory examination resulted from artifacts caused by contrast material (n = 2). Thus 50 patients were finally evaluable, including adenocarcinoma in 27 cases and squamous cell carcinoma in 23 cases (mean diameter, 3.3 cm [2.3–3.9 cm]). All patients were pathologically confirmed by either CT-guided percutaneous biopsy (n = 8), transbronchial lung biopsy (n = 15), or surgery (n = 27).
(J Comput Assist Tomogr 2016;40: 243–247)
CT Examinations
M
ultidetector row computed tomography (MDCT) has played an important role in the diagnosis of lung cancer noninvasively by demonstrating the morphology, interfaces, and enhancement type of masses.1 Because the inhomogeneous enhancement is a significant characteristic of lung cancer,2 a higher contrastto-noise ratio (CNR) for lung cancer, superior display for its enhancement details, and lower image noise are crucial in lung CT images. Recently, a new dual-energy spectral CT (DESCT) with fast tube voltage switching was introduced, which could reduce beam-hardening artifacts caused by polychromatic x-ray beam in the conventional MDCT, and may improve the image quality of lung masses. This technique has been used widely to improve image quality in imaging of pulmonary embolism, brain, and thoracic aneurysms by the selection of the optimal energy level.3–5 However, to our knowledge, the DESCT has not been specifically used in lung cancer to improve image quality. The purpose of this study was to evaluate the image quality improvement of using an
Computed tomography examinations (nonenhanced and contrast-enhanced scans) were performed to all patients on a Discovery CT750HD scanner (GE Healthcare, Wisconsin). The conventional nonenhanced helical CT scanning was performed first at a tube voltage of 120 kVp. Patients were then injected with 80 to 100 mL (1.35 mL/kg of body weight) nonionic iodinated contrast material (iopamidol 370 mg/mL; Shanghai Bracco Sine Phamaceutical Co, Ltd, China) via antecubital venous at a rate of 4.0 mL/s, and enhanced scans were obtained with dual-energy spectral mode began about 60 to 70 seconds after the injection of contrast medium. The other spectral CT parameters included the following: tube rotation time, 0.6 seconds; tube current, 600 mA; helical pitch, 1.375; field of view, 500 mm; collimation, 40 mm; and slice thickness and interval for axial images, 5 mm/5 mm. The volumetric CT dose index (CTDIvol) of the enhancement scan was 12.72 mGy (comparable to the 11.25 ± 3.28 mGy dose with conventional enhanced lung imaging for a normal-sized patient at our institution).
DESCT Postprocessing From the *Department of Radiology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai; †Department of Radiology, The First Affiliated Hospital of Anhui Medical University, Hefei; and ‡Department of CT and MRI, Zoucheng People's Hospital, Jining, Shandong Province, People's Republic of China. Received for publication August 10, 2015; accepted October 13, 2015. Correspondence to: Huawei Wu, MD, Department of Radiology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, No 160 Pujian Rd, Pudong, Shanghai 200127, People's Republic of China (e‐mail:
[email protected]). Supported by a research project of the Shanghai Municipal Health Bureau (20120176). The authors declare no conflict of interest. Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved. DOI: 10.1097/RCT.0000000000000357
The Gemstone Spectral Imaging Viewer software on an independent advanced workstation (Advantage Workstation 4.4, GE Healthcare) enables the automatic reconstruction of 101 sets of monochromatic images at photon energies ranging from 40 to 140 keV for every patient. In our research, a pre-evaluation was performed before the retrospective study, and 4 sets of monochromatic spectral images at the energy levels of 50, 60, 70, and 80 keV were chosen to evaluate image quality objectively and subjectively. We omitted lower and higher energy level because of the severe image noise in lower kiloelectron-volt and the markedly decreased iodine signal in tumors in higher kiloelectronvolt images.
J Comput Assist Tomogr • Volume 40, Number 2, March/April 2016
Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
243
J Comput Assist Tomogr • Volume 40, Number 2, March/April 2016
Hou et al
image slices, and the average values were calculated, with large vessels, obvious necrosis, and image artifacts avoided.
Subjective Image Analysis
FIGURE 1. Box-and-whisker plots of CNR in the monochromatic images at 50, 60, 70, and 80 keV. Boxes represent the middle 50% of the cases, horizontal lines within mark the median values, the whiskers represent minimal and maximal values. The CNR value peaked in the 70-keV series and was significantly higher for the 50 and 80 keV reconstructions series. There was no significant difference between 70- and 60-keV series.
Objective Image Analysis The CT attenuation values of monochromatic images from 40 to 140 keV can be calculated automatically by Gemstone Spectral Imaging Viewer software package once a region of interest (ROI) is selected. In our study, ROI was placed on the lesion and the adjacent nontumorous lung parenchyma to measure the mean CT number values (ROI ranged, 81–136 mm2; Fig. 1). The adjacent nontumorous lung parenchyma was used as the background, and a CNR value was defined using the following expression: CNR = (ROI lesion − ROI lung)/SD, where SD represented the standard deviation of CT numbers in the adjacent nontumorous lung parenchyma. Background noise was determined as the standard deviation of air measured presternally in front of the patient (ROI, 100 mm2). The half point from the center to the edge of each mass was used to divide the mass into the central and peripheral regions. Five ROIs were drawn, respectively, in 5 different locations both in the central and peripheral regions in each slice (ROI, 66.99 mm2), and all measurements were repeated at 3 consecutive image slices, then the average CT numbers were calculated. Computed tomography number difference (dCT) between the central and peripheral regions was calculated (│CTperipheral − CTcentral│) to reflect the inhomogeneous enhancement of masses. The 50 cases were divided into squamous cell carcinoma group and adenocarcinoma group according to different pathological types, and the dCT values of the 2 groups were calculated, respectively. All measurements were performed 3 times at consecutive
For subjective assessment, 2 radiologists (H.W. and Y.Y., with 14 and 6 years of experience in chest CT, respectively) interpreted the images, and their disagreements on the evaluation and score were resolved by consensus; the subjective score was decided by the 2 radiologists together. Observers were informed that all patients had a biopsy-proven lung cancer. Window settings were automatically set in soft tissue window (width, 400 Hounsfield unit [HU]; level, 100 HU) to show the enhancement details of lesions, but observers could adjust these values freely to improve tumor visualization. The 2 radiologists then evaluated and scored all monochromatic image series for the following attributes: overall image quality (1, nondiagnostic; 2, poor; 3, sufficient; 4, good; 5, excellent), inhomogeneity evaluation of the enhancement (ranging from 1, no visual internal contrast of enhancement, to 5, satisfactory visual internal contrast of enhancement).
Statistical Analysis All data were analyzed by using dedicated statistical software (SPSS for Windows, version 11.5). A value of P < 0.05 was considered statistically significant. Image noise values, CNRs, dCT values, and subjective image quality scores were presented as mean ± SD. A paired t test was performed on image noise values, CNRs, and dCT values from the different monochromatic image sets. Subjective score of the overall image quality and the inhomogeneity evaluation of the enhancement for lung cancers were tested using Wilcoxon rank sum test.
Ethics Statement The retrospective study was approved by the ethics commission of our institution (Shanghai Jiao Tong University, School of Medicine affiliated Renji Hospital) and conducted in accordance with the Declaration of Helsinki.
RESULTS The maximum CNR value (20.58 ± 7.20) occurred at 70 keV, followed by the value at 60 keV (18.83 ± 6.22), and there was no difference in the mean CNR value between the 60 and 70 keV images (P = 0.33) (Fig. 1, Table 1). Contrast-to-noise ratio values at the 50 keV (15.48 ± 6.90) and 80 keV (17.93 ± 6.40) energy levels both were significantly lower than that at 70 keV (both P < 0.05). The image noise decreased as the photon energy increases from 50 to 80 keV with the minimum value (8.28 ± 2.02 HU) for the monochromatic images observed at 80 keV (Fig. 2, Table 1). Compared with the 80 keV monochromatic images, the image noise increases were 95%, 43%, and 28% for the
TABLE 1. Quantitative Evaluation of CNR, Image Noise, and dCT Value in Different Energy Levels Energy Level CNR Image noise dCT value (SQ),* HU dCT value (AD),† HU
50 keV
60 keV
70 keV
80 keV
15.48 ± 6.90 16.34 ± 11.60 53.78 ± 25.18 41.36 ± 23.46
18.83 ± 6.22 11.85 ± 2.60 38.18 ± 19.18 27.75 ± 15.67
20.58 ± 7.20 10.67 ± 4.22 27.43 ± 15.20 20.12 ± 15.67
17.93 ± 6.40 8.28 ± 2.02 22.50 ± 13.00 18.07 ± 5.73
*dCT value of masses in squamous cell carcinoma group. †dCT value of masses in adenocarcinoma group. SQ indicates squamous cell carcinoma; AD, adenocarcinoma.
244
www.jcat.org
© 2016 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.
J Comput Assist Tomogr • Volume 40, Number 2, March/April 2016
Improving Image Quality in DESCT
which was significantly higher than for the other energy levels (both P < 0.05; Fig. 4, Table 2).
DISCUSSION
FIGURE 2. Box-and-whisker plots of image noise in the monochromatic images at 50, 60, 70, and 80 keV. Boxes represent the middle 50% of the cases, horizontal lines within mark the median values, the whiskers represent minimal and maximal values. The lowest image noise was in the 80-keV series and was significantly lower than other monochromatic energy levels.
monochromatic images at 50, 60, and 70 keV, respectively. The mean value of image noise in the 80 keV monochromatic images was statistically lower than all other energy levels (all P < 0.05). The dCT value decreased monotonically as the photon energy increases both in squamous cell carcinoma group and adenocarcinoma group with peak at 50 keV (53.78 ± 25.18 HU and 41.36 ± 23.46 HU, respectively, P > 0.05; Fig. 3, Table 1). In squamous cell carcinoma group, the second highest value of 38.18 ± 19.18 HU at 60 keV was significantly higher than that at 70 keV (27.43 ± 15.20 HU, P < 0.001) and 80 keV (22.50 ± 13.00, P < 0.001). In adenocarcinoma group, the second highest dCT value was also at 60 keV (27.75 ± 15.67 HU) and significantly higher than that at 70 keV (20.12 ± 15.67 HU, P = 0.002) and 80 keV (18.07 ± 5.73 HU, P = 0.003). The subjective score of overall image quality was the highest for the 70 keV monochromatic images (4.65 ± 0.53) and lowest for the 50 keV images (2.44 ± 0.71). The second highest score was at 60 keV (4.58 ± 0.48) and was significantly higher than that at the 50 and 80 keV image sets (both P < 0.05). The subjective evaluation of enhancement details peaked at 60 keV (4.79 ± 0.34),
In this study, we evaluated the monochromatic images of different energy levels of DESCT both objectively and subjectively to investigate the optimal energy levels for improving image quality for lung cancers. The DESCT imaging mode utilizes a single x-ray beam source that switches between 80 and 140 kVp data sets from view to view during a single rotation on the high-definition GE Discovery CT750 HD scanner, and allows the creation of the monochromatic spectral images at 101 energy levels ranging from 40 to 140 keV, which provides opportunity for selecting the optimal energy levels to improve image quality for lung cancer detection.6 The monochromatic spectral CT images can reduce beamhardening artifacts caused by polychromatic x-ray beam in conventional MDCT, thus improve CNR.7,8 Theoretically, the attenuation of iodine is greater at lower energies than at higher energies, due to the increasing photoelectric absorption that occurs with the decreasing photon energies. However, the highest CNR usually does not happen at the lowest energy level as image noise also increases as photon energy decreases.9,10 The higher CNR and lower image noise must be balanced on the monochromatic images for imaging lung cancers. According to our research results, the 70 keV monochromatic images yielded the highest CNR and second lowest image noise, which was significantly superior to the 50 keV energy level for the increased image noise and 80 keV for the much reduced attenuation. The subjective evaluation of overall image quality followed a similar pattern with the quantitative result, which indicated that 70 keV obtained the highest subjective score. In addition, our results were consistent with previous study of image quality with DESCT in liver by Lv et al11 and in pancreatic by Patel et al.6 Besides optimal CNR and image noise, the dCT value is also important for the evaluation of image quality objectively in charactering lung cancers. For some pathologic reasons such as fibrosis in adenocarcinoma, inappropriate angiogenesis and oxygen in the masses of squamous cell carcinoma, the enhancement in lung cancer is usually strong but inhomogeneous,12–15 which is different from the homogeneous enhancement in inflammatory masses and thin-rim enhancement in tuberculoma, thus the inhomogeneous enhancement would be a significant characteristic in
FIGURE 3. Box-and-whisker plots of dCT value in squamous cell carcinoma group (A) and adenocarcinoma group (B) in the monochromatic images at 50, 60, 70, and 80 keV. Boxes represent the interquartile range Q1 to Q3, horizontal lines within mark the median values, the whiskers represent minimal and maximal values. The greatest dCT value was in the 50-keV series and was significantly higher than other monochromatic energy levels. © 2016 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
245
J Comput Assist Tomogr • Volume 40, Number 2, March/April 2016
Hou et al
FIGURE 4. Axial monochromatic CT images of a 60-year-old male patient with pathologically proven pulmonary adenocarcinoma (A, 50 keV; B, 60 keV; C, 70 keV; D, 80 keV). Compared with the 60-keV monochromatic image, which showed superior CNR and inner contrast (B), reconstructions at 50 keV (A) showed significantly more obvious inner contrast of enhancement but also image noise. Monochromatic image at 70 keV (C) and 80 keV (D) demonstrate an inferior enhancement details about inner contrast, whereas image noise was also lower in these series.
the differentiation between lung cancer and benign masses with CT images. In this study, the dCT value was introduced to quantitatively characterize the inhomogeneous enhancement behaviors in lung cancers, and higher dCT value was more conducive to the objective evaluation of image quality. As previous studies had demonstrated, the greatest contrast value of lesions was usually found in the lowest viewing energy because the attenuation of iodine is greatest at lowest energies,11,16,17 our research also indicated that the greatest dCT values in squamous cell carcinoma group and adenocarcinoma group were both in 50 keV energy level and significantly higher than that in other energy levels. However, the evaluation of image quality was seriously influenced by the highest image noise at the 50 keVenergy level. The subjective evaluation of enhancement details showed that the highest
score for the inhomogeneity evaluation was at 60 keV instead of 50 keV; in other words, the enhancement differences (dCT value) between the central and peripheral regions of lung cancer at lower energy levels might not translate into conspicuity of inner contrast when readers evaluate the lesion because of the noise.18 In addition, the second highest CNR and subjective score of overall image quality were both obtained at 60 keV and were not significant lower than that at 70 keV. Based on the visibility of inhomogeneous enhancement in lung cancers, the 60 keV may also be an optimal energy level for the evaluation of lung cancer imaging objectively and subjectively. Our study had some limitations. First, the subjective assessment of our research was conducted by consensus between 2 readers, the analysis of interobserver variability was not included
TABLE 2. Subjective Scores for the Assessment of Allover Image Quality and Enhancement Details in Different Energy Level Energy Level Image quality P Enhancement details P
50 keV
60 keV
70 keV
80 keV
2.44 ± 0.71
