- Email: [email protected]
Differential diagnosis of osteoblastic metastases from bone islands in patients with lung cancer by single-source dual-energy CT: Advantages of spectral CT imaging
European Journal of Radiology 84 (2015) 901–907
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Differential diagnosis of osteoblastic metastases from bone islands in patients with lung cancer by single-source dual-energy CT: Advantages of spectral CT imaging Dong Yue a , Zheng Shaowei a , Haruhiko Machida b , Wang Bing a , Liu Ailian a,∗ , Liu Yijun a , Zhang Xin c a
Department of Radiology, The First Affiliated Hospital of Dalian Medical University, LiaoNing 116011, China Department of Radiology, Tokyo Women’s Medical University Medical Center East, Tokyo 116-8567, Japan c Department of Nuclear Medicine, The First Affiliated Hospital of Dalian Medical University, Liaoning 116011, China b
a r t i c l e
i n f o
Article history: Received 23 September 2014 Received in revised form 23 December 2014 Accepted 5 January 2015 Keywords: Dual-energy CT Monochromatic image Osteoblastic metastasis Bone island Lung cancer
a b s t r a c t Objectives: To evaluate the diagnostic efficacy of spectral CT for the differentiation of osteoblastic metastases (OBMs) from bone islands (BIs) in patients with lung cancer. Methods: In 94 patients with lung cancer who underwent spectral CT, focal hyperdense lesions in vertebral bodies were diagnosed as OBMs or BIs. Regions of interest were placed within each lesion to measure the mean CT value and its standard deviation (SD) on polychromatic single-energy CT (SECT) at 140 kVp and dual-energy virtual monochromatic spectral (VMS) images. The mean bone (Dbone(wa) ) and water densities (Dwa(bone) ) of each lesion were also measured. The slope (k) of the spectral curve was calculated. Independent-sample t-test was used to compare those values between OBMs and BIs. Receiver operator characteristic analysis was performed to compare the area under curve (AUC) for the differentiation of OBMs from BIs. Results: A total of 79 OBMs and 43 BIs were confirmed. The CT and SD values on SECT at 140 kVp and VMS images at 50–130 keV, k value, and Dbone(wa) for OBMs were significantly lower than for BIs; Dwa(bone) was significantly higher for OBMs than for BIs (p < 0.05 for all). The AUC for the SD value at 110 keV was the highest among those parameters. The optimal cut-off value for this differentiation was 68.6 HU for the SD value on VMS images at 110 keV with sensitivity of 93.0% and specificity of 93.3%. Conclusion: Spectral CT is helpful for the differentiation of OBMs from BIs in patients with lung cancer, particularly using SD of the CT value on high-energy VMS images. © 2015 Elsevier Ireland Ltd. All rights reserved.
1. Introduction According to the International Agency for Research on Cancer (IARC) GLOBOCAN World Cancer Report, lung cancer affects more than 1 million people a year worldwide [1]. Lung cancer frequently spreads to the bones. Specifically, bone metastases were reported
Abbreviations: AUC, area under the curve; BI, bone island; OBM, osteoblastic metastasis; Dbone(wa) , bone (water) density; Dwa(bone) , water (bone) density; ROI, region of interest; SECT, single-energy CT; ssDECT, single-source dual-energy CT; VMS, virtual monochromatic spectral. ∗ Corresponding author. Tel.: +86 0411 83635963; fax: +86 0411 83622844. E-mail addresses: [email protected] (Y. Dong), [email protected] (S. Zheng), [email protected] (H. Machida), [email protected] (B. Wang), dmu [email protected] (A. Liu), [email protected] (Y. Liu), [email protected] (X. Zhang). http://dx.doi.org/10.1016/j.ejrad.2015.01.007 0720-048X/© 2015 Elsevier Ireland Ltd. All rights reserved.
to be evident at post-mortem in up to 36% of patients [2]; and bone marrow micrometastases were found in 22–60% [3]. As the life expectancy of patients with lung cancer increases, symptomcontrol therapies are growing in importance. Physicians are needed to be increasingly aware of and early manage bone metastases and thus to prevent potentially debilitating and costly skeletal complications [4–6]. In clinical practice, thoraco-abdominal CT is an important follow-up study in patients with lung cancer. The CT images often disclose small, focal hyperdense vertebral lesions in these patients. Whether these lesions are osteoblastic metastases (OBMs) or not influences the therapeutic approach and patient prognosis [6]. Both OBMs and bone islands (BIs) can appear as focal hyperdense vertebral lesions with clear margins on conventional single-energy CT (SECT) images. The accurate final diagnosis of the focal hyperdense vertebral lesions can often pose a dilemma if the criteria are
902
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
based only on their size, margin, location, and CT value [7]. Presumably, atypical BIs show misleading imaging features, such as a huge expansive mass with the presence of symptoms or “hot” radionuclide uptake on bone scintigraphs [7,8]. Further examinations using magnetic resonance imaging (MRI), bone scintigraphy, or positron emission tomography (PET)/PET-CT are required for a comprehensive assessment [9–11]. However, both bone scintigraphy and FDG-PET can cause false-positive or false-negative results in the diagnosis of OBMs [11–13]. Insufficient specificity and high cost of MRI may also lead to diagnostic problems [14,15]. Recently, single-source dual-energy CT (ssDECT) system with spectral CT technique allows the acquisition of CT data using two different photon spectra at 80 and 140 kVp simultaneously [16]. In contrast to SECT, spectral CT by ssDECT provides not only SECT images at 140 kVp but also virtual monochromatic spectral (VMS) images at 40–140 keV, material decomposition images, and effective atomic number images. Thus, ssDECT transforms CT scan from single-parameter imaging into multi-parameter imaging, and offers novel strategies for the assessment of various diseases [17]. VMS images are less susceptible to beam-hardening and metal streak artifacts while also providing more accurate and reproducible attenuation measurement [18]. Iodine metric analysis by spectral CT has been demonstrated to be superior to SECT in the diagnosis and differentiation of various pathologies, including hepatic, lung, and thyroid tumors [19–22]. To our knowledge, however, little investigation has been performed for the differentiation of OBMs from BIs. The aim of the present study was to evaluate the clinical feasibility of ssDECT for the differential diagnosis of focal hyperdense vertebral lesions in patients with lung cancer. 2. Materials and methods 2.1. Research subjects The protocol was approved by the institutional review board of our hospital and all the patients gave informed consent. A retrospective review was performed for patients attending this hospital from May 2011 to May 2013, according to the following patient selection criteria: (1) lung cancer had been confirmed by histopathological examination after biopsy or surgery; (2) non-contrast chest or abdominal CT scan was performed by ssDECT at this hospital; (3) previous imaging examinations (CT and bone scintigraphs or PET-CT) had not shown the features of bone metastases at this hospital; (4) all the patients had never received chemotherapy, radiotherapy, or the treatment of bisphosphonate-type medications before the DECT examination; (5) focal hyperdense lesions with the maximal diameter of <2 cm were identified within the vertebral body in this CT scan by a radiologist (Z.SW.) with 5-year experience in diagnostic orthopedic imaging; and (6) bone scintigraphs (or PET-CT) and follow-up CT images with duration of 6 months or longer performed at this hospital were available for the diagnosis of vertebral lesions without pathological results. 2.2. Reference standards Two radiologists (D.Y. and L.AL.) with 10-year experience in diagnostic orthopedic imaging and one nuclear medicine physician (Z.X.) with 10-year clinical experience reviewed bone scintigraphs (or PET-CT) and follow-up CT findings of all the lesions identified in the ssDECT examinations. The 3 readers established the diagnostic criteria of OBMs and BIs based on their typical imaging findings as follows: OBMs were defined as osteoblastic lesions that had indeterminate morphologic features with ill-defined margins, increased radionuclide uptake on bone scintigraphs (or PET-CT),
and an increase in the size and/or density on follow-up CT images. BIs were diagnosed in the presence of their characteristic features including a focus of cortical bone attenuation in the medullary cavity with interdigitations or “thorny radiations” along the trabeculae, no radionuclide uptake on bone scintigraphs (or PET-CT), and no change in the size and density on follow-up CT images [7]. The 3 readers, in consensus, characterized the imaging findings of each lesion and diagnosed it as OBM or BI based on the aforementioned criteria. Any equivocal lesion with (1) negative on bone scintigraphs (or PET-CT) but an increase in the size and/or density on follow-up CT images or (2) positive on bone scintigraphs (or PET-CT) but no change in the size and density on follow-up CT images was excluded from the present study. 2.3. ssDECT imaging protocol All the patients enrolled in this study underwent a non-contrast chest or abdominal CT scan with an ssDECT scanner (Discovery CT750 HD; GE Healthcare, Milwaukee, WI) using spectral CT imaging scan mode with rapidly switching tube voltage between 80 and 140 kVp. The other spectral imaging parameters were as follows: tube current of 550 mA; pitch of 1.375:1; rotation time of 0.8 s. The corresponding volumetric CT dose index (CTDIvol) was 18.28 mGy. Using raw data from the spectral CT imaging, 3 types of image sets were generated as follows: (1) a set of polychromatic 140 kVp SECT images, (2) 101 sets of VMS images, and (3) material decomposition image sets using bone and water as the base material pair. Both image slice thickness and interval were 1.25 mm. 2.4. Quantitative analysis One clinical fellow (W.B.) with 2-year experience in diagnostic orthopedic imaging at our radiology department, blinded to the final diagnosis, performed quantitative analysis using a dedicated software package for spectral CT analysis (Gemstone Spectral Imaging [GSI] Viewer; GE Healthcare) in our image processing workstation (AW4.5; GE Healthcare). A circular or elliptical region of interest (ROI) was placed within each lesion on VMS images. A copy-and-paste function was applied to obtain ROIs of the same size and location on the corresponding SECT images at 140 kVp. The ROIs encompassed the central two-thirds of the lesion’s area avoiding the edges. The GSI Viewer automatically generated a spectral attenuation curve for each lesion with the x-axis representing the energy level from 40 to 140 keV (one-keV interval) and the y-axis representing the attenuation value (CT value) in Hounsfield unit (HU). The spectral curve slope (k) of each lesion was calculated as follows: k = (HU40 keV – HU100 keV )/60 HU, where CT40 keV and CT100 keV represent the CT value at 40 and 100 keV, respectively. The GSI Viewer also automatically calculated the mean CT value and its standard deviation (SD) in HU for each lesion on SECT at 140 kVp and VMS images at 50–130 keV (20-keV interval). Material density was measured on material decomposition images with bone and water as the base material pair to measure the mean bone density (Dbone(wa) ) and water density (Dwa(bone) ). The SD value represents the uniformity of the CT value of each pixel within the ROI of each lesion. All the measurements were performed twice on 2 consecutive slices; the averaged value was calculated to reduce a variation of the measurements. 2.5. Statistical analysis All continuous variables were expressed as mean ± SD. Statistical analysis was performed using SPSS 19.0 (SPSS Inc., Chicago, IL). Independent sample t-test was used to compare patients’ age, height, body weight, and body mass index (BMI) and the area of ROI and the calculated measurements consisting of the k value,
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
903
Table 1 Patient characteristics.
OBMs (n = 55) BIs (n = 39)
Male/female
Age (years)
Height (cm)
Body weight (kg)
BMI (kg/m2 )
Lesion number
ROI area (cm2 )
29/26
36–80 (59.8 ± 19.2) 35–76 (61.1 ± 20.8)
156–183 (174 ± 7) 155–183 (173 ± 7)
52–85 (68 ± 12) 53–89 (67 ± 10)
20.8–28.5 (24.6 ± 4.2) 21.6–29.1 (24.9 ± 3.7)
79 (T: 38, L: 41)
0.18–0.41 (0.31 ± 0.09) 0.16–0.42 (0.29 ± 0.11)
24/15
43 (T: 17, L: 26)
BIs, bone islands; BMI, body mass index; L, lumber spine; OBMs, osteoblastic metastases; T, thoracic spine.
Table 2 CT values on 140-kVp and virtual monochromatic spectral images for OBMs and BIs. BIs (HU) 140 kVp 50 keV 70 keV 90 keV 110 keV 130 keV
840.6 1380.2 856.9 646.5 545.7 493.9
± ± ± ± ± ±
OBMs (HU) 211.2 400.6 247.4 186.2 157.5 143.0
664.6 1078.5 678.8 517.5 440.6 401.0
± ± ± ± ± ±
170.7 321.4 197.6 148.2 125.3 113.8
t value
p value
4.238 3.886 3.723 3.584 3.453 3.361
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Table 3 SD values on 140-kVp and virtual monochromatic spectral images for OBMs and BIs. BIs (HU) 140 kVp 50 keV 70 keV 90 keV 110 keV 130 keV
120.08 202.16 129.13 100.51 87.38 80.83
± ± ± ± ± ±
OBMs (HU) 47.32 132.75 83.16 67.35 58.81 54.61
75.99 110.30 70.68 55.39 48.41 44.98
± ± ± ± ± ±
30.03 46.11 28.24 21.48 18.39 16.88
t value
p value
4.927 6.433 6.527 6.534 6.562 3.576
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001
BIs, bone islands; OBMs, osteoblastic metastases. Data are mean ± standard deviation (shown in HU).
BIs, bone islands; OBMs, osteoblastic metastases; SD, standard deviation. Data are mean ± SD (shown in HU).
CT value, SD value, Dbone(wa) , and Dwa(bone) between OBM- and BIgroups. Receiver operator characteristic (ROC) curve analysis was performed to compare area under the curve (AUC) for the differential diagnostic efficacy of OBMs from BIs using the k value, CT and SD values on SECT at 140 kVp and VMS imaging at 50–130 keV, and Dbone(wa) , and Dwa(bone) values. The optimal cut-off value of each parameter was determined for this differential diagnosis; the sensitivity and specificity were calculated when using this cut-off value. A p value of less than 0.05 was regarded as statistically significant.
3.2. Spectral curves
3. Results 3.1. Patient characteristics This study included 94 patients with lung cancer (53 patients, adenocarcinoma; 15, squamous cell carcinoma; 11, small cell cancers; 15, undetermined pathologic type) with a total of 122 focal hyperdense vertebral lesions and the mean age of 60.3 ± 20.2 years (range, 35–80 years). In the 122 lesions, 79 OBMs and 43 BIs were determined. Five patients had both OBMs and BIs. The lesion diameter ranged from 0.8 to 1.6 cm (mean, 1.1 ± 0.3 cm). The area of ROIs ranged from 0.16 to 0.42 cm2 (mean, 0.30 ± 0.10 cm2 ). There was no significant difference in patients’ age (p = 0.72), height (p = 0.66), body weight (p = 0.77), BMI (p = 0.72), and ROI area (p = 0.79) between OBM- and BI-groups. Detailed information about patient characteristics is shown in Table 1.
According to the spectral curves, CT value steadily decreased for both OBMs and BIs, as the energy level was increased, but the decreasing degree differed at different energy levels (Figs. 1, 4 and 5). The k value for OBMs (16.8 ± 5.3) was significantly lower than that for BIs (22.0 ± 6.5) (p < 0.001). 3.3. CT values, SD values, and material densities Both CT and SD values for BIs on SECT at 140 kVp and VMS images at 50–130 keV were significantly higher than all the corresponding values for OBMs (p < 0.05) (Tables 2 and 3). Dwa(bone) of OBMs (223.3 ± 270.3 mg/cm3 ) was significantly higher than that of BIs (−46.6 ± 327.5 mg/cm3 ) (p < 0.05). Dbone(wa) of OBMs (1208.1 ± 380.8 mg/cm3 ) was significantly lower than that of BIs (1579.7 ± 469.8 mg/cm3 ) (p < 0.05) (Figs. 2, 4 and 5). 3.4. Diagnostic efficacy of each parameter in differentiating OBMs from BIs The AUC for SD values on VMS images at 110 keV was 0.972. This AUC was significantly higher than that on VMS images at any other energy levels (0.861–0.891 at 50, 70, 90, and 130 keV) and on SECT at 140 kVp (0.791); and also higher than the AUC for CT values at 50–130 keV and 140 kVp (0.696–0.741) (p < 0.05). The AUC
Fig. 1. Spectral curves for bone islands and osteoblastic metastases
904
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
Fig. 2. Box-whisker plots of Dbone (wa) (a) and Dwa (bone) (b) for osteoblastic metastases (OBMs) and bone islands (BIs).
Table 4 AUC value, the optimal cut-off value, sensitivity and specificity in each parameter for differentiating OBMs from BIs.
AUC value Optimal cut-off value Sensitivity Specificity
CT value at 140 kVp
SD value at 110 keV
k value
Dbone(wa)
Dwa(bone)
0.741 734.2 HU 67.4% 62.2%
0.972 68.6 HU 93.0% 93.3%
0.733 18.7 72.1% 64.4%
0.734 1433.3 mg/cm3 69.8% 75.6%
0.739 94.0 mg/cm3 71.1% 67.4%
AUC, area under the curve; BIs, bone islands; OBMs, osteoblastic metastases; SD, standard deviation.
for the k value, Dbone(wa) and Dwa(bone) was 0.733, 0.734, and 0.739, respectively (Table 4 and Fig. 3). Thus, SD value obtained on VMS images at 110 keV showed the highest AUC value and a maximal diagnostic efficacy in differentiating OBMs from BIs. The optimal cut-off value for this differentiation was 68.6 HU for SD values on VMS images at 110 keV with sensitivity of 93.0% and specificity of 93.3% (Table 4 and Fig. 3). 4. Discussion Each pure substance has its specific spectral attenuation curve, since mass attenuation is shown as a function of photon energy. The X-ray attenuation of various tissues can be expressed by a pair of known pure substances. Based on these principles, spectral CT by ssDECT allows accurate quantitative analysis and material decomposition for regions of interest [16,25,26]. Previous studies
Fig. 3. ROC curves for SD-110 keV, 140 kVp, bone (water) density and k value by ssDECT in differentiating osteoblastic metastases (OBMs) and bone islands (BIs).
have shown the usefulness of the slope of the spectral curve for the differential diagnosis of benign or malignant thyroid nodules [22,23]. Although OBMs and BIs can show similar imaging features and CT values on SECT images at 140 kVp, they showed different spectral curves in the present study due to the difference in material composition. The spectral curves of both OBMs and BIs showed a steady decrease with higher energy levels but the slope between 40 and 100 keV was significantly different between OBMs and BIs, reflecting the following facts: even though both OBMs and BIs contain osteoid tissue, BIs consist of normal bone tissue structure; in contrast, OBMs contain abnormal osteoid tissue. In this study, both the CT and SD values of BIs were significantly higher than those of OBMs on SECT at 140 kVp and VMS images. BI is an asymptomatic hamartomatous malformation and histologically defined as a discrete focus of compact bone within the spongiosa [7]. The vast majority of BIs range from 1 mm to 2 cm in the maximal diameter, and show hyperdense similar to the normal cortical bone on CT [7]. Thus, this study only included focal hyperdense vertebral lesions with the maximal diameter of less than 2 cm. Some BIs were excluded from this study, because they showed atypical imaging features and thus did not reach an agreement for the diagnosis by the 3 readers. On the other hand, OBM is histologically defined as abnormal proliferation of osteoid tissue with abundant tumor cells and incomplete deposition of calcium salts. OBMs do not contain normal cortical bone structure, and are relatively less dense than BIs [24]. A CT image is an array of pixels, and each pixel is assigned a CT value. The CT value within an ROI is the mean of CT values of all the individual pixels included in the ROI. Its SD value indicates the degree of divergence of the individual pixel values from the mean value, and reflects the homogeneity within the ROI. Lower SD value means more homogenous nature within the ROI. The size of ROIs ranged from 0.15 to 0.42 cm2 for both OBM- and BI-groups and was comparable between the 2 groups. Nevertheless, the SD values for BIs on SECT at 140 kVp and VMS images were significantly higher than those of OBMs. With histologically showing a discrete focus of compact bone within the spongiosa, BIs are composed of
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
905
Fig. 4. (a–e) A bone island in the 12th thoracic vertebral body in a 58-year-old man with lung adenocarcinoma for 2 years. (a and b) Virtual monochromatic spectral images at 70 keV in December 2011 (a) and March 2012 (b) similarly shows a focal hyperdense lesion with clear margin and radiating bony streaks. CT value of this lesion was 677.6 HU in (a) and 629.6 HU in (b). (c) Spectral curve for this lesion shows a steadily decreased CT value with increasing energy level (shown in keV). (d) Bone (water) density image reveals the density of this lesion as 1095.8 mg/cm3 . (e) Water (bone) density image, as 25.6 mg/cm3 .
mature cortical bone growing along the cancellous bone trabeculae, resulting in a mixture of cancellous bone trabeculae, normal bone marrow, and cortical bone [7]. Characteristic radiating bony streaks are aligned with the axis of the host bone’s trabeculae, blending with the surrounding trabeculae in a feathered or brush like fashion [7]. Higher SD values of BIs reflect their histological inhomogeneity. In contrast, OBMs contain abnormal osteoid tissue with a relatively regular composition, thus resulting in a more uniform density and lower SD values compared to BIs. For ssDECT with fast switching of tube voltage, low- and highkVp projections are interleaved during the acquisition. As a result,
they incur a small angular offset relative to each other, and are interpolated to the same angular positions as a pre-processing step. These paired projections are then decomposed into density integrals of base material pair. During the system calibration, material decomposition by the pair can be reconstructed, and material density images by the pair can be obtained using the standard tomographic reconstruction technique. The two major components of the vertebral body were chosen as the pair, bone and water, for the quantitative analysis using material density images in this study. OBMs completely destroy original bone structures and significantly reduce their bone content and increase their water
906
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
Fig. 5. (a–d) An osteoblastic metastasis in the 1st lumbar vertebral body in a 68-year-old woman with lung adenocarcinoma for 3 years. (a) Virtual monochromatic spectral image at 70 keV shows a focal homogeneous hyperdense lesion with clear margin. CT value of this lesion was 614.7 HU. (b) Spectral curve for this lesion shows a steadily decreased CT value with increasing energy level (shown in keV). (c) Bone (water) density image reveals the density of this lesion as 1055.8 mg/cm3 . (d) Water (bone) density image, as 338.3 mg/cm3 .
content. In contrast, BIs consist of normal cortical bone within the cancellous trabecula and hence more bone content and less water content. Thus, BIs showed significantly higher Dbone(wa) and lower Dwa(bone) compared to OBMs. Bone- and water-based material decomposition images can provide an accurate and objective assessment for the differential diagnosis of these focal hyperdense lesions. According to the ROC curve analysis, CT values on VMS images at 50–130 keV were not better than those on SECT at 140 kVp for differentiating OBMs from BIs. However, SD values on VMS images at 50–130 keV demonstrated better diagnostic accuracy than those on SECT at 140 kVp. Particularly, the diagnostic accuracy by SD values at 110 keV was higher than any other parameters. The optimal cut-off value of SD values at 110 keV was 68.6 HU, providing the sensitivity and specificity of 93.0% and 93.3%, respectively. The use of high-energy VMS images can effectively reduce beam-hardening effect and image noise, thus providing more accurate measurement of CT and SD values and aiding the differentiation between OBMs and BIs. Although SD values at 110 keV were more useful for distinguishing OBM from BI than any other parameters, the final diagnosis could not just rely on the SD value. Of course, lesion morphology and CT value also should be taken into account. Other than ssDECT, bone scintigraphy is still a modality of first choice for the diagnosis of vertebral metastases because
of its high sensitivity and easy accessibility. Although diffusionweighted imaging and dynamic-enhanced T1-weighted imaging were reported to be useful for the differential diagnosis of spinal lesions [27–29], some contraindications and a high expense may limit a clinical use of MRI. At present, ssDECT is not suitable as a modality of first choice for the diagnosis of vertebral metastases due to the need of radiation exposure and special equipment. However, when compared to conventional CT, ssDECT is useful for its higher diagnostic accuracy. The various parameters evaluated in spectral CT by ssDECT can play an important role for follow-up imaging studies for patients with malignant tumors. There are several limiting aspects of this study that need to be considered. First of all, the imaging features of smaller lesions of both OBMs and BIs tended to be atypical, and the definitive diagnosis was more difficult. Thus, this study included only 122 hyperdense lesions with the maximal diameter of less than 2 cm. This research reflected our preliminary experience using a small number of patients. Our results are needed to be verified through further studies using a larger sample size in the future. Second, the diagnosis of all the lesions was performed based on imaging findings of bone scintigraphy (or PET-CT) and follow-up CT without pathological “gold standard” for the diagnosis of both OBMs and BIs in this study. Further clinical trials need to be performed to validate our quantitative data based on pathological diagnosis. Third, in this
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
study all CT examinations were done without IV contrast because of the limitation of clinicians demand, whereas in most clinical practices, initial and followup thoracoabdominal CT for patients with lung cancer is done with IV contrast, and the further research would focus on this problem. Forth, the fact that the mean radiation dose of ssDECT is higher than that of SECT limits its wider applications. With the second-generation ssDECT, the use of adaptive statistical iterative reconstruction is expected to reduce radiation dose. Finally, although ssDECT can display a variety of parameters with a better diagnostic efficacy compared to SECT, further studies may be needed to correlate findings of ssDECT imaging with those of other modalities, including MRI, bone scintigraphy, or PET-CT for this differential diagnosis of hyperdense vertebral lesions in patients with various malignant tumors. This will be useful to reduce or replace MRI, bone scintigraphy or PET-CT examinations for follow-up studies in these patients. In conclusion, ssDECT can generate VMS images, spectral curves and perform quantitative material composition analysis. It is useful for the accurate differentiation of OBMs from BIs. Particularly, VMS image at 110 keV is optimal for providing this accurate diagnosis and has a sufficient potential for differentiating benign and malignant hyperdense vertebral lesions. Authors’ contribution Dong Yue and Liu Ailian were involved in study conception. Dong Yue and Zheng Shaowei were participated in study design. Wang Bing, Zheng Shaowei and Liu Yijun were exclusively made data acquisition. Dong Yue, Liu Ailian and Zhang Xin had quality control of data and algorithms. Dong Yue, Liu Ailian and Zhang xin made data analysis and interpretation. Zheng Shaowei and Wang Bing were involved statistical analysis. Zheng Shaowei and Wang Bing prepared the manuscript. Manuscript was edited by Dong Yue, Zheng Shaowei, Haruhiko Machida. Dong Yue, Haruhiko Machida and Liu Ailian reviewed the manuscript. Conflicts of interest We declared that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled. References [1] Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer 2010;127:2893–917. [2] Coleman RE. Clinical features of metastatic bone disease and risk of skeletal morbidity. Clin Cancer Res 2006;12:6243–9. [3] Coello MC, Luketich JD, Litle VR, Godfrey TE. Prognostic significance of micrometastases in non-small-cell lung cancer. Clin Lung Cancer 2004;5:214–25. [4] Scagliotti GV, Hirsh V, Siena S, Henry DH, Woll PJ, Manegold C, et al. Overall survival improvement in patients with lung cancer and bone metastases treated with denosumab versus zoledronic acid: subgroup analysis from a randomized phase 3 study. J Thorac Oncol 2012;7:1823–9.
907
[5] Rossi A, Gridelli C, Ricciardi S, de Marinis F. Bone metastases and non-small cell lung cancer: from bisphosphonates to targeted therapy. Curr Med Chem 2012;19:5524–35. [6] Brodowicz T, O’Byrne K, Manegold C. Bone matters in lung cancer. Ann Oncol 2012;23:2215–22. [7] Greenspan A. Bone island(enostosis):current concept-a review. Skeletal Radiol 1995;2:111–5. [8] Yanagawa T, Watanabe H, Shinozaki T, Ahmed AR, Shirakura K, Takagishi K. The natural history of disappearing bone tumours and tumour-like conditions. Clin Radiol 2001;56:877–86. [9] Howe BM, Johnson GB, Wenger DE. Current concepts in MRI of focal and diffuse malignancy of bone marrow. Semin Musculoskelet Radiol 2013;17: 137–44. [10] Savelli G, Maffioli L, Maccauro M, De Deckere E, Bombardieri E. Bone scintigraphy and the added value of SPECT (single photon emission tomography) in detecting skeletal lesions. Q J Nucl Med 2001;45:27–37. [11] Evangelista L, Panunzio A, Polverosi R, Ferretti A, Chondrogiannis S, Pomerri F, et al. Early bone marrow metastasis detection: the additional value of FDGPET/CT vs. CT imaging. Biomed Pharmacother 2012;66:448–53. [12] Huyge V, Garcia C, Vanderstappen A, Alexiou J, Gil T, Flamen P. Progressive osteoblastic bone metastases in breast cancer negative on FDG-PET. Clin Nucl Med 2009;34:417–20. [13] Taoka T, Mayr NA, Lee HJ, Yuh WT, Simonson TM, Rezai K, et al. Factors influencing visualization of vertebral metastases on MR imaging versus bone scintigraphy. AJR Am J Roentgenol 2001;176:1523–30. [14] Shih WJ. Vertebral SPECT bone scintigraphy should be compared with MR imaging for vertebral metastases. AJR Am J Roentgenol 2001;177: 1482. [15] Donald Z, William M, Mark S, Parellada JA, Carrino JA. Benign and malignant processes: normal values and differentiation with chemical shift MR imaging in vertebral marrow. Radiology 2005;237:590–6. [16] YehB M, Shepherd JA, Wang ZJ, Teh HS, Hartman RP, Prevrhal S. Dualenergy and low kVp CT in the abdomen. AJR Am J Roentgenol 2009;193: 47–54. [17] Hurrell MA, Butler AP, Cook NJ, Butler PH, Ronaldson JP, Zainon R. Spectral Hounsfield units: a new radiological concept. Eur Radiol 2012;22:1008–13. [18] Matsuda I, Akahane M, Sato J, Katsura M, Kiryu S, Yoshioka N, et al. Precision of measurement of CT numbers: comparison of dual-energy CT spectral imaging with fast kVP switching and conventional CT with phantoms. Jpn J Radiol 2012;30:34–9. [19] 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. [20] Yoshitake Y, Masahiro J, Yutaka T, Abe T, Kuribayashi S. Virtual monochromatic spectral imaging for the evaluation of hypovascular hepatic metastases. Invest Radiol 2012;47:292–8. [21] Remy-Jardin M, Faivre JB, Pontana F, Hachulla AL, Tacelli N, Santangelo T, et al. Thoracic applications of dual energy. Radiol Clin North Am 2010;48:193–205. [22] Li M, Zheng X, Li J, Yang Y, Lu C, Xu H, et al. Dual-energy computed tomography imaging of thyroid nodule specimens. Invest Radiol 2011;45:780–1. [23] Ni MF, Wang LJ, Dong Y, Miao YW, Zhang JW. Spectral CT imaging in differential diagnosis of benign and malignant thyroid nodules. Chin J Med Imaging Technol 2012;28:1642–5. [24] Clezardin P, Teti A. Bone metastasis: pathogenesis and therapeutic implications. Clin Exp Metastasis 2007;24:599–608. [25] Ascenti G, Siragusa C, Racchiusa S, Ielo I, Privitera G, Midili F, et al. Stonetargeted dual energy CT: a new diagnostic approach to urinary calculosis. AJR Am J Roentgenol 2010;195:953–8. [26] Chandarana H, Megibow AJ, Cohen BA, Srinivasan R, Kim D, Leidecker C, et al. Iodine quantification with dual-energy CT: phantom study and preliminary experience with renal masses. AJR Am J Roentgenol 2011;196:693–700. [27] Khadem NR, Karimi S, Peck KK, Yamada Y, Lis E, Lyo J, et al. Characterizing hypervascular and hypovascular metastases and normal bone marrow of the spine using dynamic contrast-enhanced MR imaging. AJNR Am J Neuroradiol 2012;33:2178–85. [28] Eiber M, Holzapfel K, Ganter C, Epple K, Metz S, Geinitz H, et al. Whole-body MRI including diffusion-weighted imaging (DWI) for patients with recurring prostate cancer: technical feasibility and assessment of lesion conspicuity in DWI. J Magn Reson Imaging 2011;33:1160–70. [29] Messiou C, Collins DJ, Morgan VA, Robson MD, deBono JS, Bydder GM, et al. Quantifying sclerotic bone metastases with 2D ultra short TE MRI: a feasibility study. Cancer Biomark 2010;7:211–8.
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Differential diagnosis of osteoblastic metastases from bone islands in patients with lung cancer by single-source dual-energy CT: Advantages of spectral CT imaging Dong Yue a , Zheng Shaowei a , Haruhiko Machida b , Wang Bing a , Liu Ailian a,∗ , Liu Yijun a , Zhang Xin c a
Department of Radiology, The First Affiliated Hospital of Dalian Medical University, LiaoNing 116011, China Department of Radiology, Tokyo Women’s Medical University Medical Center East, Tokyo 116-8567, Japan c Department of Nuclear Medicine, The First Affiliated Hospital of Dalian Medical University, Liaoning 116011, China b
a r t i c l e
i n f o
Article history: Received 23 September 2014 Received in revised form 23 December 2014 Accepted 5 January 2015 Keywords: Dual-energy CT Monochromatic image Osteoblastic metastasis Bone island Lung cancer
a b s t r a c t Objectives: To evaluate the diagnostic efficacy of spectral CT for the differentiation of osteoblastic metastases (OBMs) from bone islands (BIs) in patients with lung cancer. Methods: In 94 patients with lung cancer who underwent spectral CT, focal hyperdense lesions in vertebral bodies were diagnosed as OBMs or BIs. Regions of interest were placed within each lesion to measure the mean CT value and its standard deviation (SD) on polychromatic single-energy CT (SECT) at 140 kVp and dual-energy virtual monochromatic spectral (VMS) images. The mean bone (Dbone(wa) ) and water densities (Dwa(bone) ) of each lesion were also measured. The slope (k) of the spectral curve was calculated. Independent-sample t-test was used to compare those values between OBMs and BIs. Receiver operator characteristic analysis was performed to compare the area under curve (AUC) for the differentiation of OBMs from BIs. Results: A total of 79 OBMs and 43 BIs were confirmed. The CT and SD values on SECT at 140 kVp and VMS images at 50–130 keV, k value, and Dbone(wa) for OBMs were significantly lower than for BIs; Dwa(bone) was significantly higher for OBMs than for BIs (p < 0.05 for all). The AUC for the SD value at 110 keV was the highest among those parameters. The optimal cut-off value for this differentiation was 68.6 HU for the SD value on VMS images at 110 keV with sensitivity of 93.0% and specificity of 93.3%. Conclusion: Spectral CT is helpful for the differentiation of OBMs from BIs in patients with lung cancer, particularly using SD of the CT value on high-energy VMS images. © 2015 Elsevier Ireland Ltd. All rights reserved.
1. Introduction According to the International Agency for Research on Cancer (IARC) GLOBOCAN World Cancer Report, lung cancer affects more than 1 million people a year worldwide [1]. Lung cancer frequently spreads to the bones. Specifically, bone metastases were reported
Abbreviations: AUC, area under the curve; BI, bone island; OBM, osteoblastic metastasis; Dbone(wa) , bone (water) density; Dwa(bone) , water (bone) density; ROI, region of interest; SECT, single-energy CT; ssDECT, single-source dual-energy CT; VMS, virtual monochromatic spectral. ∗ Corresponding author. Tel.: +86 0411 83635963; fax: +86 0411 83622844. E-mail addresses: [email protected] (Y. Dong), [email protected] (S. Zheng), [email protected] (H. Machida), [email protected] (B. Wang), dmu [email protected] (A. Liu), [email protected] (Y. Liu), [email protected] (X. Zhang). http://dx.doi.org/10.1016/j.ejrad.2015.01.007 0720-048X/© 2015 Elsevier Ireland Ltd. All rights reserved.
to be evident at post-mortem in up to 36% of patients [2]; and bone marrow micrometastases were found in 22–60% [3]. As the life expectancy of patients with lung cancer increases, symptomcontrol therapies are growing in importance. Physicians are needed to be increasingly aware of and early manage bone metastases and thus to prevent potentially debilitating and costly skeletal complications [4–6]. In clinical practice, thoraco-abdominal CT is an important follow-up study in patients with lung cancer. The CT images often disclose small, focal hyperdense vertebral lesions in these patients. Whether these lesions are osteoblastic metastases (OBMs) or not influences the therapeutic approach and patient prognosis [6]. Both OBMs and bone islands (BIs) can appear as focal hyperdense vertebral lesions with clear margins on conventional single-energy CT (SECT) images. The accurate final diagnosis of the focal hyperdense vertebral lesions can often pose a dilemma if the criteria are
902
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
based only on their size, margin, location, and CT value [7]. Presumably, atypical BIs show misleading imaging features, such as a huge expansive mass with the presence of symptoms or “hot” radionuclide uptake on bone scintigraphs [7,8]. Further examinations using magnetic resonance imaging (MRI), bone scintigraphy, or positron emission tomography (PET)/PET-CT are required for a comprehensive assessment [9–11]. However, both bone scintigraphy and FDG-PET can cause false-positive or false-negative results in the diagnosis of OBMs [11–13]. Insufficient specificity and high cost of MRI may also lead to diagnostic problems [14,15]. Recently, single-source dual-energy CT (ssDECT) system with spectral CT technique allows the acquisition of CT data using two different photon spectra at 80 and 140 kVp simultaneously [16]. In contrast to SECT, spectral CT by ssDECT provides not only SECT images at 140 kVp but also virtual monochromatic spectral (VMS) images at 40–140 keV, material decomposition images, and effective atomic number images. Thus, ssDECT transforms CT scan from single-parameter imaging into multi-parameter imaging, and offers novel strategies for the assessment of various diseases [17]. VMS images are less susceptible to beam-hardening and metal streak artifacts while also providing more accurate and reproducible attenuation measurement [18]. Iodine metric analysis by spectral CT has been demonstrated to be superior to SECT in the diagnosis and differentiation of various pathologies, including hepatic, lung, and thyroid tumors [19–22]. To our knowledge, however, little investigation has been performed for the differentiation of OBMs from BIs. The aim of the present study was to evaluate the clinical feasibility of ssDECT for the differential diagnosis of focal hyperdense vertebral lesions in patients with lung cancer. 2. Materials and methods 2.1. Research subjects The protocol was approved by the institutional review board of our hospital and all the patients gave informed consent. A retrospective review was performed for patients attending this hospital from May 2011 to May 2013, according to the following patient selection criteria: (1) lung cancer had been confirmed by histopathological examination after biopsy or surgery; (2) non-contrast chest or abdominal CT scan was performed by ssDECT at this hospital; (3) previous imaging examinations (CT and bone scintigraphs or PET-CT) had not shown the features of bone metastases at this hospital; (4) all the patients had never received chemotherapy, radiotherapy, or the treatment of bisphosphonate-type medications before the DECT examination; (5) focal hyperdense lesions with the maximal diameter of <2 cm were identified within the vertebral body in this CT scan by a radiologist (Z.SW.) with 5-year experience in diagnostic orthopedic imaging; and (6) bone scintigraphs (or PET-CT) and follow-up CT images with duration of 6 months or longer performed at this hospital were available for the diagnosis of vertebral lesions without pathological results. 2.2. Reference standards Two radiologists (D.Y. and L.AL.) with 10-year experience in diagnostic orthopedic imaging and one nuclear medicine physician (Z.X.) with 10-year clinical experience reviewed bone scintigraphs (or PET-CT) and follow-up CT findings of all the lesions identified in the ssDECT examinations. The 3 readers established the diagnostic criteria of OBMs and BIs based on their typical imaging findings as follows: OBMs were defined as osteoblastic lesions that had indeterminate morphologic features with ill-defined margins, increased radionuclide uptake on bone scintigraphs (or PET-CT),
and an increase in the size and/or density on follow-up CT images. BIs were diagnosed in the presence of their characteristic features including a focus of cortical bone attenuation in the medullary cavity with interdigitations or “thorny radiations” along the trabeculae, no radionuclide uptake on bone scintigraphs (or PET-CT), and no change in the size and density on follow-up CT images [7]. The 3 readers, in consensus, characterized the imaging findings of each lesion and diagnosed it as OBM or BI based on the aforementioned criteria. Any equivocal lesion with (1) negative on bone scintigraphs (or PET-CT) but an increase in the size and/or density on follow-up CT images or (2) positive on bone scintigraphs (or PET-CT) but no change in the size and density on follow-up CT images was excluded from the present study. 2.3. ssDECT imaging protocol All the patients enrolled in this study underwent a non-contrast chest or abdominal CT scan with an ssDECT scanner (Discovery CT750 HD; GE Healthcare, Milwaukee, WI) using spectral CT imaging scan mode with rapidly switching tube voltage between 80 and 140 kVp. The other spectral imaging parameters were as follows: tube current of 550 mA; pitch of 1.375:1; rotation time of 0.8 s. The corresponding volumetric CT dose index (CTDIvol) was 18.28 mGy. Using raw data from the spectral CT imaging, 3 types of image sets were generated as follows: (1) a set of polychromatic 140 kVp SECT images, (2) 101 sets of VMS images, and (3) material decomposition image sets using bone and water as the base material pair. Both image slice thickness and interval were 1.25 mm. 2.4. Quantitative analysis One clinical fellow (W.B.) with 2-year experience in diagnostic orthopedic imaging at our radiology department, blinded to the final diagnosis, performed quantitative analysis using a dedicated software package for spectral CT analysis (Gemstone Spectral Imaging [GSI] Viewer; GE Healthcare) in our image processing workstation (AW4.5; GE Healthcare). A circular or elliptical region of interest (ROI) was placed within each lesion on VMS images. A copy-and-paste function was applied to obtain ROIs of the same size and location on the corresponding SECT images at 140 kVp. The ROIs encompassed the central two-thirds of the lesion’s area avoiding the edges. The GSI Viewer automatically generated a spectral attenuation curve for each lesion with the x-axis representing the energy level from 40 to 140 keV (one-keV interval) and the y-axis representing the attenuation value (CT value) in Hounsfield unit (HU). The spectral curve slope (k) of each lesion was calculated as follows: k = (HU40 keV – HU100 keV )/60 HU, where CT40 keV and CT100 keV represent the CT value at 40 and 100 keV, respectively. The GSI Viewer also automatically calculated the mean CT value and its standard deviation (SD) in HU for each lesion on SECT at 140 kVp and VMS images at 50–130 keV (20-keV interval). Material density was measured on material decomposition images with bone and water as the base material pair to measure the mean bone density (Dbone(wa) ) and water density (Dwa(bone) ). The SD value represents the uniformity of the CT value of each pixel within the ROI of each lesion. All the measurements were performed twice on 2 consecutive slices; the averaged value was calculated to reduce a variation of the measurements. 2.5. Statistical analysis All continuous variables were expressed as mean ± SD. Statistical analysis was performed using SPSS 19.0 (SPSS Inc., Chicago, IL). Independent sample t-test was used to compare patients’ age, height, body weight, and body mass index (BMI) and the area of ROI and the calculated measurements consisting of the k value,
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
903
Table 1 Patient characteristics.
OBMs (n = 55) BIs (n = 39)
Male/female
Age (years)
Height (cm)
Body weight (kg)
BMI (kg/m2 )
Lesion number
ROI area (cm2 )
29/26
36–80 (59.8 ± 19.2) 35–76 (61.1 ± 20.8)
156–183 (174 ± 7) 155–183 (173 ± 7)
52–85 (68 ± 12) 53–89 (67 ± 10)
20.8–28.5 (24.6 ± 4.2) 21.6–29.1 (24.9 ± 3.7)
79 (T: 38, L: 41)
0.18–0.41 (0.31 ± 0.09) 0.16–0.42 (0.29 ± 0.11)
24/15
43 (T: 17, L: 26)
BIs, bone islands; BMI, body mass index; L, lumber spine; OBMs, osteoblastic metastases; T, thoracic spine.
Table 2 CT values on 140-kVp and virtual monochromatic spectral images for OBMs and BIs. BIs (HU) 140 kVp 50 keV 70 keV 90 keV 110 keV 130 keV
840.6 1380.2 856.9 646.5 545.7 493.9
± ± ± ± ± ±
OBMs (HU) 211.2 400.6 247.4 186.2 157.5 143.0
664.6 1078.5 678.8 517.5 440.6 401.0
± ± ± ± ± ±
170.7 321.4 197.6 148.2 125.3 113.8
t value
p value
4.238 3.886 3.723 3.584 3.453 3.361
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Table 3 SD values on 140-kVp and virtual monochromatic spectral images for OBMs and BIs. BIs (HU) 140 kVp 50 keV 70 keV 90 keV 110 keV 130 keV
120.08 202.16 129.13 100.51 87.38 80.83
± ± ± ± ± ±
OBMs (HU) 47.32 132.75 83.16 67.35 58.81 54.61
75.99 110.30 70.68 55.39 48.41 44.98
± ± ± ± ± ±
30.03 46.11 28.24 21.48 18.39 16.88
t value
p value
4.927 6.433 6.527 6.534 6.562 3.576
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001
BIs, bone islands; OBMs, osteoblastic metastases. Data are mean ± standard deviation (shown in HU).
BIs, bone islands; OBMs, osteoblastic metastases; SD, standard deviation. Data are mean ± SD (shown in HU).
CT value, SD value, Dbone(wa) , and Dwa(bone) between OBM- and BIgroups. Receiver operator characteristic (ROC) curve analysis was performed to compare area under the curve (AUC) for the differential diagnostic efficacy of OBMs from BIs using the k value, CT and SD values on SECT at 140 kVp and VMS imaging at 50–130 keV, and Dbone(wa) , and Dwa(bone) values. The optimal cut-off value of each parameter was determined for this differential diagnosis; the sensitivity and specificity were calculated when using this cut-off value. A p value of less than 0.05 was regarded as statistically significant.
3.2. Spectral curves
3. Results 3.1. Patient characteristics This study included 94 patients with lung cancer (53 patients, adenocarcinoma; 15, squamous cell carcinoma; 11, small cell cancers; 15, undetermined pathologic type) with a total of 122 focal hyperdense vertebral lesions and the mean age of 60.3 ± 20.2 years (range, 35–80 years). In the 122 lesions, 79 OBMs and 43 BIs were determined. Five patients had both OBMs and BIs. The lesion diameter ranged from 0.8 to 1.6 cm (mean, 1.1 ± 0.3 cm). The area of ROIs ranged from 0.16 to 0.42 cm2 (mean, 0.30 ± 0.10 cm2 ). There was no significant difference in patients’ age (p = 0.72), height (p = 0.66), body weight (p = 0.77), BMI (p = 0.72), and ROI area (p = 0.79) between OBM- and BI-groups. Detailed information about patient characteristics is shown in Table 1.
According to the spectral curves, CT value steadily decreased for both OBMs and BIs, as the energy level was increased, but the decreasing degree differed at different energy levels (Figs. 1, 4 and 5). The k value for OBMs (16.8 ± 5.3) was significantly lower than that for BIs (22.0 ± 6.5) (p < 0.001). 3.3. CT values, SD values, and material densities Both CT and SD values for BIs on SECT at 140 kVp and VMS images at 50–130 keV were significantly higher than all the corresponding values for OBMs (p < 0.05) (Tables 2 and 3). Dwa(bone) of OBMs (223.3 ± 270.3 mg/cm3 ) was significantly higher than that of BIs (−46.6 ± 327.5 mg/cm3 ) (p < 0.05). Dbone(wa) of OBMs (1208.1 ± 380.8 mg/cm3 ) was significantly lower than that of BIs (1579.7 ± 469.8 mg/cm3 ) (p < 0.05) (Figs. 2, 4 and 5). 3.4. Diagnostic efficacy of each parameter in differentiating OBMs from BIs The AUC for SD values on VMS images at 110 keV was 0.972. This AUC was significantly higher than that on VMS images at any other energy levels (0.861–0.891 at 50, 70, 90, and 130 keV) and on SECT at 140 kVp (0.791); and also higher than the AUC for CT values at 50–130 keV and 140 kVp (0.696–0.741) (p < 0.05). The AUC
Fig. 1. Spectral curves for bone islands and osteoblastic metastases
904
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
Fig. 2. Box-whisker plots of Dbone (wa) (a) and Dwa (bone) (b) for osteoblastic metastases (OBMs) and bone islands (BIs).
Table 4 AUC value, the optimal cut-off value, sensitivity and specificity in each parameter for differentiating OBMs from BIs.
AUC value Optimal cut-off value Sensitivity Specificity
CT value at 140 kVp
SD value at 110 keV
k value
Dbone(wa)
Dwa(bone)
0.741 734.2 HU 67.4% 62.2%
0.972 68.6 HU 93.0% 93.3%
0.733 18.7 72.1% 64.4%
0.734 1433.3 mg/cm3 69.8% 75.6%
0.739 94.0 mg/cm3 71.1% 67.4%
AUC, area under the curve; BIs, bone islands; OBMs, osteoblastic metastases; SD, standard deviation.
for the k value, Dbone(wa) and Dwa(bone) was 0.733, 0.734, and 0.739, respectively (Table 4 and Fig. 3). Thus, SD value obtained on VMS images at 110 keV showed the highest AUC value and a maximal diagnostic efficacy in differentiating OBMs from BIs. The optimal cut-off value for this differentiation was 68.6 HU for SD values on VMS images at 110 keV with sensitivity of 93.0% and specificity of 93.3% (Table 4 and Fig. 3). 4. Discussion Each pure substance has its specific spectral attenuation curve, since mass attenuation is shown as a function of photon energy. The X-ray attenuation of various tissues can be expressed by a pair of known pure substances. Based on these principles, spectral CT by ssDECT allows accurate quantitative analysis and material decomposition for regions of interest [16,25,26]. Previous studies
Fig. 3. ROC curves for SD-110 keV, 140 kVp, bone (water) density and k value by ssDECT in differentiating osteoblastic metastases (OBMs) and bone islands (BIs).
have shown the usefulness of the slope of the spectral curve for the differential diagnosis of benign or malignant thyroid nodules [22,23]. Although OBMs and BIs can show similar imaging features and CT values on SECT images at 140 kVp, they showed different spectral curves in the present study due to the difference in material composition. The spectral curves of both OBMs and BIs showed a steady decrease with higher energy levels but the slope between 40 and 100 keV was significantly different between OBMs and BIs, reflecting the following facts: even though both OBMs and BIs contain osteoid tissue, BIs consist of normal bone tissue structure; in contrast, OBMs contain abnormal osteoid tissue. In this study, both the CT and SD values of BIs were significantly higher than those of OBMs on SECT at 140 kVp and VMS images. BI is an asymptomatic hamartomatous malformation and histologically defined as a discrete focus of compact bone within the spongiosa [7]. The vast majority of BIs range from 1 mm to 2 cm in the maximal diameter, and show hyperdense similar to the normal cortical bone on CT [7]. Thus, this study only included focal hyperdense vertebral lesions with the maximal diameter of less than 2 cm. Some BIs were excluded from this study, because they showed atypical imaging features and thus did not reach an agreement for the diagnosis by the 3 readers. On the other hand, OBM is histologically defined as abnormal proliferation of osteoid tissue with abundant tumor cells and incomplete deposition of calcium salts. OBMs do not contain normal cortical bone structure, and are relatively less dense than BIs [24]. A CT image is an array of pixels, and each pixel is assigned a CT value. The CT value within an ROI is the mean of CT values of all the individual pixels included in the ROI. Its SD value indicates the degree of divergence of the individual pixel values from the mean value, and reflects the homogeneity within the ROI. Lower SD value means more homogenous nature within the ROI. The size of ROIs ranged from 0.15 to 0.42 cm2 for both OBM- and BI-groups and was comparable between the 2 groups. Nevertheless, the SD values for BIs on SECT at 140 kVp and VMS images were significantly higher than those of OBMs. With histologically showing a discrete focus of compact bone within the spongiosa, BIs are composed of
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
905
Fig. 4. (a–e) A bone island in the 12th thoracic vertebral body in a 58-year-old man with lung adenocarcinoma for 2 years. (a and b) Virtual monochromatic spectral images at 70 keV in December 2011 (a) and March 2012 (b) similarly shows a focal hyperdense lesion with clear margin and radiating bony streaks. CT value of this lesion was 677.6 HU in (a) and 629.6 HU in (b). (c) Spectral curve for this lesion shows a steadily decreased CT value with increasing energy level (shown in keV). (d) Bone (water) density image reveals the density of this lesion as 1095.8 mg/cm3 . (e) Water (bone) density image, as 25.6 mg/cm3 .
mature cortical bone growing along the cancellous bone trabeculae, resulting in a mixture of cancellous bone trabeculae, normal bone marrow, and cortical bone [7]. Characteristic radiating bony streaks are aligned with the axis of the host bone’s trabeculae, blending with the surrounding trabeculae in a feathered or brush like fashion [7]. Higher SD values of BIs reflect their histological inhomogeneity. In contrast, OBMs contain abnormal osteoid tissue with a relatively regular composition, thus resulting in a more uniform density and lower SD values compared to BIs. For ssDECT with fast switching of tube voltage, low- and highkVp projections are interleaved during the acquisition. As a result,
they incur a small angular offset relative to each other, and are interpolated to the same angular positions as a pre-processing step. These paired projections are then decomposed into density integrals of base material pair. During the system calibration, material decomposition by the pair can be reconstructed, and material density images by the pair can be obtained using the standard tomographic reconstruction technique. The two major components of the vertebral body were chosen as the pair, bone and water, for the quantitative analysis using material density images in this study. OBMs completely destroy original bone structures and significantly reduce their bone content and increase their water
906
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
Fig. 5. (a–d) An osteoblastic metastasis in the 1st lumbar vertebral body in a 68-year-old woman with lung adenocarcinoma for 3 years. (a) Virtual monochromatic spectral image at 70 keV shows a focal homogeneous hyperdense lesion with clear margin. CT value of this lesion was 614.7 HU. (b) Spectral curve for this lesion shows a steadily decreased CT value with increasing energy level (shown in keV). (c) Bone (water) density image reveals the density of this lesion as 1055.8 mg/cm3 . (d) Water (bone) density image, as 338.3 mg/cm3 .
content. In contrast, BIs consist of normal cortical bone within the cancellous trabecula and hence more bone content and less water content. Thus, BIs showed significantly higher Dbone(wa) and lower Dwa(bone) compared to OBMs. Bone- and water-based material decomposition images can provide an accurate and objective assessment for the differential diagnosis of these focal hyperdense lesions. According to the ROC curve analysis, CT values on VMS images at 50–130 keV were not better than those on SECT at 140 kVp for differentiating OBMs from BIs. However, SD values on VMS images at 50–130 keV demonstrated better diagnostic accuracy than those on SECT at 140 kVp. Particularly, the diagnostic accuracy by SD values at 110 keV was higher than any other parameters. The optimal cut-off value of SD values at 110 keV was 68.6 HU, providing the sensitivity and specificity of 93.0% and 93.3%, respectively. The use of high-energy VMS images can effectively reduce beam-hardening effect and image noise, thus providing more accurate measurement of CT and SD values and aiding the differentiation between OBMs and BIs. Although SD values at 110 keV were more useful for distinguishing OBM from BI than any other parameters, the final diagnosis could not just rely on the SD value. Of course, lesion morphology and CT value also should be taken into account. Other than ssDECT, bone scintigraphy is still a modality of first choice for the diagnosis of vertebral metastases because
of its high sensitivity and easy accessibility. Although diffusionweighted imaging and dynamic-enhanced T1-weighted imaging were reported to be useful for the differential diagnosis of spinal lesions [27–29], some contraindications and a high expense may limit a clinical use of MRI. At present, ssDECT is not suitable as a modality of first choice for the diagnosis of vertebral metastases due to the need of radiation exposure and special equipment. However, when compared to conventional CT, ssDECT is useful for its higher diagnostic accuracy. The various parameters evaluated in spectral CT by ssDECT can play an important role for follow-up imaging studies for patients with malignant tumors. There are several limiting aspects of this study that need to be considered. First of all, the imaging features of smaller lesions of both OBMs and BIs tended to be atypical, and the definitive diagnosis was more difficult. Thus, this study included only 122 hyperdense lesions with the maximal diameter of less than 2 cm. This research reflected our preliminary experience using a small number of patients. Our results are needed to be verified through further studies using a larger sample size in the future. Second, the diagnosis of all the lesions was performed based on imaging findings of bone scintigraphy (or PET-CT) and follow-up CT without pathological “gold standard” for the diagnosis of both OBMs and BIs in this study. Further clinical trials need to be performed to validate our quantitative data based on pathological diagnosis. Third, in this
Y. Dong et al. / European Journal of Radiology 84 (2015) 901–907
study all CT examinations were done without IV contrast because of the limitation of clinicians demand, whereas in most clinical practices, initial and followup thoracoabdominal CT for patients with lung cancer is done with IV contrast, and the further research would focus on this problem. Forth, the fact that the mean radiation dose of ssDECT is higher than that of SECT limits its wider applications. With the second-generation ssDECT, the use of adaptive statistical iterative reconstruction is expected to reduce radiation dose. Finally, although ssDECT can display a variety of parameters with a better diagnostic efficacy compared to SECT, further studies may be needed to correlate findings of ssDECT imaging with those of other modalities, including MRI, bone scintigraphy, or PET-CT for this differential diagnosis of hyperdense vertebral lesions in patients with various malignant tumors. This will be useful to reduce or replace MRI, bone scintigraphy or PET-CT examinations for follow-up studies in these patients. In conclusion, ssDECT can generate VMS images, spectral curves and perform quantitative material composition analysis. It is useful for the accurate differentiation of OBMs from BIs. Particularly, VMS image at 110 keV is optimal for providing this accurate diagnosis and has a sufficient potential for differentiating benign and malignant hyperdense vertebral lesions. Authors’ contribution Dong Yue and Liu Ailian were involved in study conception. Dong Yue and Zheng Shaowei were participated in study design. Wang Bing, Zheng Shaowei and Liu Yijun were exclusively made data acquisition. Dong Yue, Liu Ailian and Zhang Xin had quality control of data and algorithms. Dong Yue, Liu Ailian and Zhang xin made data analysis and interpretation. Zheng Shaowei and Wang Bing were involved statistical analysis. Zheng Shaowei and Wang Bing prepared the manuscript. Manuscript was edited by Dong Yue, Zheng Shaowei, Haruhiko Machida. Dong Yue, Haruhiko Machida and Liu Ailian reviewed the manuscript. Conflicts of interest We declared that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled. References [1] Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer 2010;127:2893–917. [2] Coleman RE. Clinical features of metastatic bone disease and risk of skeletal morbidity. Clin Cancer Res 2006;12:6243–9. [3] Coello MC, Luketich JD, Litle VR, Godfrey TE. Prognostic significance of micrometastases in non-small-cell lung cancer. Clin Lung Cancer 2004;5:214–25. [4] Scagliotti GV, Hirsh V, Siena S, Henry DH, Woll PJ, Manegold C, et al. Overall survival improvement in patients with lung cancer and bone metastases treated with denosumab versus zoledronic acid: subgroup analysis from a randomized phase 3 study. J Thorac Oncol 2012;7:1823–9.
907
[5] Rossi A, Gridelli C, Ricciardi S, de Marinis F. Bone metastases and non-small cell lung cancer: from bisphosphonates to targeted therapy. Curr Med Chem 2012;19:5524–35. [6] Brodowicz T, O’Byrne K, Manegold C. Bone matters in lung cancer. Ann Oncol 2012;23:2215–22. [7] Greenspan A. Bone island(enostosis):current concept-a review. Skeletal Radiol 1995;2:111–5. [8] Yanagawa T, Watanabe H, Shinozaki T, Ahmed AR, Shirakura K, Takagishi K. The natural history of disappearing bone tumours and tumour-like conditions. Clin Radiol 2001;56:877–86. [9] Howe BM, Johnson GB, Wenger DE. Current concepts in MRI of focal and diffuse malignancy of bone marrow. Semin Musculoskelet Radiol 2013;17: 137–44. [10] Savelli G, Maffioli L, Maccauro M, De Deckere E, Bombardieri E. Bone scintigraphy and the added value of SPECT (single photon emission tomography) in detecting skeletal lesions. Q J Nucl Med 2001;45:27–37. [11] Evangelista L, Panunzio A, Polverosi R, Ferretti A, Chondrogiannis S, Pomerri F, et al. Early bone marrow metastasis detection: the additional value of FDGPET/CT vs. CT imaging. Biomed Pharmacother 2012;66:448–53. [12] Huyge V, Garcia C, Vanderstappen A, Alexiou J, Gil T, Flamen P. Progressive osteoblastic bone metastases in breast cancer negative on FDG-PET. Clin Nucl Med 2009;34:417–20. [13] Taoka T, Mayr NA, Lee HJ, Yuh WT, Simonson TM, Rezai K, et al. Factors influencing visualization of vertebral metastases on MR imaging versus bone scintigraphy. AJR Am J Roentgenol 2001;176:1523–30. [14] Shih WJ. Vertebral SPECT bone scintigraphy should be compared with MR imaging for vertebral metastases. AJR Am J Roentgenol 2001;177: 1482. [15] Donald Z, William M, Mark S, Parellada JA, Carrino JA. Benign and malignant processes: normal values and differentiation with chemical shift MR imaging in vertebral marrow. Radiology 2005;237:590–6. [16] YehB M, Shepherd JA, Wang ZJ, Teh HS, Hartman RP, Prevrhal S. Dualenergy and low kVp CT in the abdomen. AJR Am J Roentgenol 2009;193: 47–54. [17] Hurrell MA, Butler AP, Cook NJ, Butler PH, Ronaldson JP, Zainon R. Spectral Hounsfield units: a new radiological concept. Eur Radiol 2012;22:1008–13. [18] Matsuda I, Akahane M, Sato J, Katsura M, Kiryu S, Yoshioka N, et al. Precision of measurement of CT numbers: comparison of dual-energy CT spectral imaging with fast kVP switching and conventional CT with phantoms. Jpn J Radiol 2012;30:34–9. [19] 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. [20] Yoshitake Y, Masahiro J, Yutaka T, Abe T, Kuribayashi S. Virtual monochromatic spectral imaging for the evaluation of hypovascular hepatic metastases. Invest Radiol 2012;47:292–8. [21] Remy-Jardin M, Faivre JB, Pontana F, Hachulla AL, Tacelli N, Santangelo T, et al. Thoracic applications of dual energy. Radiol Clin North Am 2010;48:193–205. [22] Li M, Zheng X, Li J, Yang Y, Lu C, Xu H, et al. Dual-energy computed tomography imaging of thyroid nodule specimens. Invest Radiol 2011;45:780–1. [23] Ni MF, Wang LJ, Dong Y, Miao YW, Zhang JW. Spectral CT imaging in differential diagnosis of benign and malignant thyroid nodules. Chin J Med Imaging Technol 2012;28:1642–5. [24] Clezardin P, Teti A. Bone metastasis: pathogenesis and therapeutic implications. Clin Exp Metastasis 2007;24:599–608. [25] Ascenti G, Siragusa C, Racchiusa S, Ielo I, Privitera G, Midili F, et al. Stonetargeted dual energy CT: a new diagnostic approach to urinary calculosis. AJR Am J Roentgenol 2010;195:953–8. [26] Chandarana H, Megibow AJ, Cohen BA, Srinivasan R, Kim D, Leidecker C, et al. Iodine quantification with dual-energy CT: phantom study and preliminary experience with renal masses. AJR Am J Roentgenol 2011;196:693–700. [27] Khadem NR, Karimi S, Peck KK, Yamada Y, Lis E, Lyo J, et al. Characterizing hypervascular and hypovascular metastases and normal bone marrow of the spine using dynamic contrast-enhanced MR imaging. AJNR Am J Neuroradiol 2012;33:2178–85. [28] Eiber M, Holzapfel K, Ganter C, Epple K, Metz S, Geinitz H, et al. Whole-body MRI including diffusion-weighted imaging (DWI) for patients with recurring prostate cancer: technical feasibility and assessment of lesion conspicuity in DWI. J Magn Reson Imaging 2011;33:1160–70. [29] Messiou C, Collins DJ, Morgan VA, Robson MD, deBono JS, Bydder GM, et al. Quantifying sclerotic bone metastases with 2D ultra short TE MRI: a feasibility study. Cancer Biomark 2010;7:211–8.