- Email: [email protected]
Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma
Clinical Radiology xxx (2016) e1ee9
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma G. Ascenti, C. Sofia*, S. Mazziotti, S. Silipigni, T. D’Angelo, S. Pergolizzi, E. Scribano Department of Biomedical Sciences and Morphologic and Functional Imaging, Policlinico “G. Martino”, University of Messina, Italy
art icl e i nformat ion Article history: Received 29 November 2015 Received in revised form 14 April 2016 Accepted 3 May 2016
AIM: To investigate the diagnostic accuracy of dual-energy multidetector computed tomography (MDCT) with iodine quantification compared to conventional enhancement measurements in distinguishing bland from neoplastic portal vein thrombosis in patients with hepatocellular carcinoma. MATERIAL AND METHODS: Thirty-four patients (26 men, eight women; mean age, 62 years) with hepatocellular carcinoma and portal vein thrombosis underwent contrast-enhanced dualenergy MDCT during the late hepatic arterial phase for the assessment of portal thrombosis (bland, n¼21; neoplastic, n¼13). Datasets were analysed separately by two different readers. Interobserver correlation and variability were calculated and compared with the BlandeAltman method. Diagnostic accuracy of conventional enhancement measurements and iodine quantification was calculated by setting either histopathology (n¼7) or a reference standard based on MDCT imaging criteria and thrombus evolutionary characteristics compared to a previous MDCT examination (n¼27). For iodine quantification threshold determination receiver operating characteristic (ROC) curves were drawn. p-Values <0.05 were considered significant. RESULTS: For conventional enhancement measurements and iodine quantification interobserver correlation was 98% and 96%. Enhancement measurement resulted in a sensitivity of 92.3%, specificity of 85.7%, positive predictive value (PPV) of 80%, and negative predictive value (NPV) of 94.7%. An iodine concentration of 0.9 mg/ml optimised discrimination between neoplastic and bland thrombi (area under the ROC [AUC] 0.993) resulting in a sensitivity of 100%, specificity of 95.2%, PPV of 92.9%, and NPV of 100%. The overall diagnostic accuracy of iodine quantification (97%) was significantly better than conventional enhancement measurements (88.2%; p<0.001). CONCLUSION: Compared to conventional enhancement measurements, iodine quantification improves the characterisation of portal vein thrombi during the late hepatic arterial phase in patients with hepatocellular carcinoma. Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
* Guarantor and correspondent: C. Sofia, Department of Biomedical Sciences and Morphologic and Functional Imaging, Policlinico “G. Martino”, University of Messina, 98100 Messina, Italy. Tel.: þ39 090 221 3092; fax: þ39 090 221 3720. E-mail address: carm.sofi[email protected] (C. Sofia). http://dx.doi.org/10.1016/j.crad.2016.05.002 0009-9260/Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e2
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
Introduction The distinction between bland and neoplastic portal vein thrombosis (PVT) represents a watershed for the clinical management of patients having hepatocellular carcinoma (HCC).1e3 Bland PVT, which is associated with a slowdown of the portal bloodstream, is reported in up to 40% of patients with HCC.4,5 Conversely, neoplastic thrombosis of the portal vein, which is determined by tumour invasion of the lumen of the portal vein or its tributaries, is found in 6.5e44% of patients with HCC.4,5 Beyond affecting tumour staging and patient prognosis, the occurrence of neoplastic thrombosis of the portal vein dramatically impacts the therapeutic planning by precluding a number of possible options, including surgical resection, orthotopic liver transplantation as well as chemoembolisation.1e5 Although the standard method for characterising a portal vein thrombus is represented by histopathology, imaging can play a decisive role and various imaging techniques, including Doppler ultrasonography, contrast-enhanced ultrasonography, multidetector-row computed tomography (MDCT) or magnetic resonance imaging, have long been utilised as yardsticks for identifying and characterising PVT.6e11 In MDCT, while the identification of the thrombus is largely reliant upon portal venous phase imaging, the late hepatic arterial phase is the ideal time point to determine the nature of the thrombus based on evidence of enhancement representing the most suggestive clue of thrombus malignancy.4,5,12e14 Nevertheless, inherent technical limitations of contrast-enhancement assessment, e.g., challenges in appropriate region-of-interest (ROI) placement on multiphase images, are such that radiodensity values can potentially fail to capture subtle changes in iodine uptake of the microvascular meshwork of the thrombus, thereby potentially undermining proper patient management.5,12e14 There is growing evidence that dualenergy MDCT with iodine quantification offers advantages over conventional CT for accurate quantification of iodine uptake in human tissues and provides a considerable potential for differentiating bland versus neoplastic PVT.15e18 The purpose of the present study was to investigate the diagnostic accuracy of dual-energy MDCT with iodine quantification compared to conventional enhancement measurements to distinguish between bland and neoplastic PVT.
Materials and methods This prospective, single-centre study was approved by the Institutional Review Board, and a waiver of informed consent was obtained.
Patients From April 2013 to April 2015, 45 consecutive patients with diagnosis of PVT at previous CT examination within the previous 6 months and biopsy diagnosis of HCC, who were referred for MDCT follow-up, were prospectively evaluated
using DECT. From the initial population, patients who underwent complete portal vein recanalisation and had no evidence of thrombosis were excluded (n¼11). The final study population consisted of 34 patients (26 men, eight women; mean age 62.18.9 yearsSD; age range, 48e79 years; mean body mass index, 244.7) with HCC (solitary, n¼17; multifocal, n¼11; diffusely infiltrative, n¼6; average primary lesion size: 21.417.4 mm, range, 8e80 mm), having a total of 34 thrombi (main portal vein, n¼14; right portal vein, n¼14; left portal vein, n¼6). Fig 1 shows the flowchart of study enrolment population. All these patients had undergone thrombolytic therapy after the first diagnosis of PVT and none had undergone chemotherapy (such as angiogenesis inhibitors) during the time interval between the first and the second CT examination.
Image technique Data acquisition All CT examinations of the liver were performed with a dual-source MDCT system (Somatom Definition Dual Source; Siemens Healthcare, Forchheim, Germany). The study protocol consisted of non-contrast, portal venous and equilibrium phases acquired in single-energy mode and the late hepatic arterial phase obtained in dual-energy mode (Table 1), using an automatic bolus-tracking technique (CARE Bolus, Siemens Healthcare; Forchheim; Germany) for the scanning delay. The late hepatic arterial phase was started automatically 12 seconds after the trigger threshold (100 HU) was reached at the level of the supracoeliac abdominal aorta. Portal venous and equilibrium phases were acquired at a fixed time delay of 80 and 180 seconds after the beginning of the contrast medium injection. All patients received 1.4 ml/kg body weight (560 mgI/kg of body weight) of a non-ionic contrast-medium (iomeprol, 400 mg iodine/ml; Iomeron 400; Bracco Diagnostics, Princeton, NJ, USA) that was administered with a dual-chamber mechanical power injector (Stellant D; Medrad, Indianola, PA, USA), at a rate of 5 ml/s through an intravenous catheter inserted into an antecubital vein. This was followed by a 30 ml saline chaser at the same injection rate. This protocol represents the standard CT approach for the study of patients with hepatic cirrhosis and HCC.
Data reconstruction From single-energy datasets, 3-mm-thick transverse images were reconstructed and only non-contrast images were utilised for the subsequent image analysis in this study. From dual-energy datasets, 3-mm-thick contiguous transverse 120-kVp simulated linear-blended images, which approximate the image quality of a standard 120-kVp scan of the abdomen, were obtained with a combination of image data averaging 30% from the low-kVp data and 70% from the high-kVp data, using a dual-energy dedicated convolution kernel (D30f). All datasets were networked to a secondary workstation where each patient’s late hepatic arterial phase study was
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
e3
Figure 1 Flowchart of the study enrolment population based on adopted inclusion and exclusion criteria and reference standard. y MDCT diagnosis was based on direct infiltration of primary tumour into the portal vein, expansion of the involved vessel, or unequivocal evidence of contrast enhancement [4, 5]. z Findings of the previous 6 months MDCT examinations relied upon depiction of a substantial growth in thrombus size, defined as an increase in the longest transverse diameter by greater than 30%. By comparison, stability in size or shrinkage was considered suggestive of bland nature of the thrombus.
post-processed with the commercially available dualenergy software (syngo Dual Energy version MMWP2010A, Siemens), using an iodine-subtraction algorithm (Liver VNC, Syngo Dual Energy version MMWP 2010A, Table 1 Acquisition parameters for single-energy and dual-energy multidetector computed tomography (MDCT). MDCT parameters
Single-energy MDCT
Dual-energy MDCT
Detector collimation (mm) Tube voltage (kV) Tube currentetime product (mAs) Gantry revolution time (s) Acquisition mode Pitch CTDIvol (mGy) Reconstructed section thickness (mm) Reconstruction algorithm Reconstruction kernel
640.6 120 200
141.2 140/80 95/510
0.5
0.5
Helical 0.75 11.1 3
Helical 0.75 13.9 3
Filtered back projection Soft tissue (B31f)
Filtered back projection Soft tissue (D30f)
CTDIvol, volume computed tomography dose index.
Siemens) as an image-based analysis of the low- and high-energy kVp images. Based on the three-material decomposition algorithm, this software enables spectral extraction of materials having a sufficiently high k-edge (e.g., iodine and calcium).17,19e20 The iodine-dependent slope for dual-energy three-material decomposition was also calibrated by placing a large ROI over the aortic lumen, at the level of the coeliac trunk origin (mean ROI size, 190 pixels). The manufacturer default value of 50% iodine overlay display, as well as the values for soft-tissue and fat attenuation at 80 and 140 kVp, were utilised for the analysis.
Image analysis All data were analysed independently and in a blinded fashion by two radiologists with 10 and 5 years of experience, respectively, in dual-energy abdominal imaging. The reviewers were blinded to the outcome, previously determined for each case. ROI measurements were obtained over each portal thrombus on non-contrast and contrast-enhanced late hepatic arterial phase images, both 120-kVp simulated
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e4
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
linear-blended and colour-coded iodine maps. The mean radiodensity values and iodine concentrations were recorded by drawing circular ROIs five times for each portal thrombus; average values from three ROI measurements were then used for statistical analyses, after the maximal and minimal values of five measurements were excluded (“winsorised mean” procedure).21 Each ROI included at least two-thirds of the area of the thrombus while encompassing as much of the areas showing either the highest attenuation (on 120-kVp simulated linearblended images) or highest coloured pixel content (on colour-coded iodine maps), as possible. During ROI measurements the wall of the vessel or prominent artefacts, if any, were carefully avoided. For the entire period of analysis, the precision of ROI measurements was further assured by a real-time guidance and thrombus tracking facilitated by the simultaneous display of portal venous images serving as a roadmap on an adjacent screen for each single case. Besides ensuring reproducibility of the data using repeated ROI samples, reliability of the results was also warranted in terms of consistency by keeping the size, shape, and position of the ROIs constant among the repeated measurements. While iodine quantification measurements were directly provided by the dual-energy software with the ROIs on colour-coded iodine maps, conventional contrastenhancement values of the thrombi were determined using the equation (HUpostcontrast HUnoncontrast)
where HUpostcontrast is the mean radiodensity value on the 120-kVp simulated linear-blended images and HUnoncontrast is the mean radiodensity value on non-contrast images.
Reference standard A histopathological proof of thrombus was available in seven of 34 patients (20,5%): autoptic specimens were available in four and surgical specimens were available in three patients who underwent en-bloc hepatic resection. In the remaining 27 portal vein thrombi, currently accepted MDCT criteria for characterisation of portal thrombus, together with comparative evaluation of thrombus features with a MDCT examination performed within a time-interval of 3e6 months, were used as a reference standard. CT criteria are represented by: (a) direct infiltration into the portal vein by the primary tumour4,5,11,18; (b) expansion of the involved vessel (vessel diameter, >2.6 cm for the main portal vein; >2.3 cm for the right portal vein; >2.2 cm for the left portal vein), along with a disproportionate enlargement of the vessel when compared with non-affected same-order branches in the same lobe4,5,11,18; (c) unequivocal evidence of enhancement on contrast-enhanced MDCT images.4,5,11,18 If, at least, two of the above-mentioned criteria were present, the thrombus was considered malignant; otherwise the thrombus was defined as bland.4,5,11,18
A substantial growth in thrombus size compared to prior MDCT examination d defined as an increase in the longest transverse diameter of thrombus by greater than 30% d was considered indicative of the neoplastic nature of thrombus. By contrast, stability in size or shrinkage was considered suggestive of bland nature.3
Statistical analysis Numerical values of continuous variables are expressed as means standard deviations, and categorical variables are expressed as frequencies or percentages. Interobserver variation in measurement of thrombus attenuation and iodine quantification was assessed in terms of Pearson’s correlation between the measurements provided by the two readers. Difference in measurements was determined by BlandeAltman analysis, by calculating the mean and 95% confidence interval (CI) of the arithmetic differences between measurements by two readers. If the differences are normally distributed, 95% of the differences will lie within a range of 1.96 SD of the mean difference. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) of conventional enhancement measurements and iodine quantification were calculated from c2 tests of contingency, setting as reference standard either histopathology or MDCT features together with comparative evaluation of thrombus characteristics with a prior CT examination (this latter not older than 6 months). With conventional enhancement measurements, thrombus enhancement was defined as a 20 HU increase in attenuation seen on 120-kVp simulated linear-blended late hepatic arterial phase when compared with noncontrast images.4,5 By comparison, in light of the discrepancy between the currently reported thresholds of iodine uptake using the iodine quantification for different abdominal applications, the optimal threshold for distinguishing bland from neoplastic thrombus was determined in a similar manner to Qian et al.15e18 In particular, receiver operating characteristic (ROC) curve analysis was utilised to establish the optimal threshold for iodine concentration; the area under the ROC curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and relative confidence interval (CI) were calculated. Differences in sensitivity and specificity between conventional enhancement measurements and iodine quantification were assessed using the McNemar test. Statistical significance was assumed to be p<0.05. All statistical analyses were performed using statistical software (MedCalc, version 12.7.1.0, MedCalc, Mariakerke, Belgium).
Results On histopathological examination, five out of seven thrombi were neoplastic and the remaining two were bland. According to conventional CT parameters, eight
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
thrombi out of the remaining 27 were neoplastic and 19 bland. In particular, all eight neoplastic thrombi showed substantial growth in size compared to the prior CT examination (average longest transverse diameter at baseline examination, 20.9 cm; average longest transverse diameter at re-examination, 3.70.5 cm; mean percentage of increase in size, 85%), while the remaining 19 bland portal vein thrombi displayed no substantial growth (average longest transverse diameter at baseline examination, 1.80.8 cm SD; average longest transverse diameter at reexamination, 1.90.2 cm; mean percentage of increase in size, 5.5%). Additionally, among neoplastic thrombi, direct neoplastic invasion of the portal vein was found in six out of eight thrombi, expansion of the involved vessel was reported in seven out of eight, and unequivocal evidence of enhancement was found in seven out of eight. In the remaining 19 thrombi, none of these conditions was met at conventional CT examination and were hence considered bland. In the presentseries bland thrombi (main portal vein, n¼11; right portal vein, n¼8; left portal vein, n¼2) showed a mean post-contrast attenuation increase of 8.68.5 HU for Reader I and of 8.87.1 HU for Reader II (p¼0.82). The average iodine concentration for this group was 0.50.4 mg/ml for Reader I and 0.60.4 mg/ml for Reader II (p 0.26). Conversely, neoplastic thrombi (main portal vein, n¼3; right portal vein, n¼6; left portal vein, n¼4) showed a mean post-contrast attenuation increase of 42.713.6 HU for Reader I and 43.311 HU for Reader II (p¼0.59). Average iodine concentration for this group was 2.30.6 mg/ml for Reader I and 2.40.4 mg/ml SD for Reader II (p¼0.18). Correlation among the two readers’ measurements was 98% for post-contrast attenuation increase (p<0.05) and 96% for iodine concentration (p<0.05). Interobserver mean difference was 0.3 for post-contrast attenuation increase (SD¼3.7; 95% CI: 0.9 to 1.6; p¼0.59) and 0.1 for iodine quantification (SD¼0.29; 95% CI: 0.01 to 0.19; p¼0.08). Using conventional enhancement measurements, there were 12 true-positive, three false-positive, one false-negative, and 18 true-negative results, yielding a sensitivity of 92.3% (95% CI: 63.9%, 98.7%), a specificity of 85.7% (95% CI: 63.6%, 96.8%), a PPV of 80% (95% CI: 51.9%, 95.4%), and a NPV of 94.7% (95% CI: 73.9%, 99.1%). The overall diagnostic accuracy was of 88.2%. The ROC analysis demonstrated that 0.9 mg/ml was the optimal threshold (AUC, 0.993; 95% CI: 0.88, 1.00) for discriminating between neoplastic and bland thrombi when using the iodine quantification (Fig 2), with 13 true-positive (Figs 3e4), one false-positive, no falsenegative, and 20 true-negative results. Overall, iodine quantification yielded a sensitivity of 100% (95% CI: 75.3%, 100%), a specificity of 95.2% (95% CI: 76.2%, 99.9%), a PPV of 92.9% (95% CI: 66.1%, 99.8%), and a NPV of 100% (95% CI: 83.2%, 100%), with an overall diagnostic accuracy of 97%. The McNemar test showed that, when compared with conventional enhancement measurements, iodine
e5
quantification allowed for significantly better diagnostic accuracy in distinguishing neoplastic from bland thrombi (p<0.001).
Discussion Imaging-based determination of the aetiology of PVT is of paramount value for clinical decision making in patients with HCC.1e11 The results of the present study indicate that, when compared with conventional enhancement measurements, dual-energy MDCT with iodine quantification yields significantly improved diagnostic performance for distinguishing bland from neoplastic PVT based on the iodine-uptake assessment. Of note, the present data indicates that an iodine concentration threshold of 0.9 mg/ ml optimises the discrimination between bland and neoplastic thrombi, in general accordance with preliminary data reported by Qian et al.,18 who showed the feasibility of dual-energy iodine quantification to differentiate between neoplastic and bland thrombi during the hepatic portal venous phase. It is important to note that the clinical perspective offered by Qian et al. was, somewhat limited by the lack of comparison to conventional enhancement measurements during the late hepatic arterial phase of enhancement.18 Despite the challenges encountered in portal thrombus detection, the latter remains the stronghold for characterising portal thrombus based on iodine uptake.22 As Qian et al.18 acknowledged in their study, poor thrombus enhancement in the hepatic portal venous phase can yield false-negative findings.12e14,18 It is well known that the venous phase is the most accurate for thrombus detection and that in several cases neoplastic thrombus vasculature may mimic an initial inflow of iodinate blood into the portal circulation during the late arterial phase. On the other hand, the arterial phase is fundamental to characterise the nature of the thrombus.22 In the present evaluation, ROI measurement was further assured by real-time guidance and thrombus tracking facilitated by simultaneous display of portal venous images serving as a roadmap on an adjacent screen for each single case. There are discrepancies in the optimal threshold for thrombus iodine-uptake identification proposed in the present study (0.9 mg/ml) and that recommended by Qian et al. (1.14 mg/ml).18 Beyond the variation in the selected imaging study phase (late hepatic arterial versus hepatic portal venous), it is conceivable that these differences could be related to the technical approaches adopted for iodine quantification, namely the image-space domain with the dual-source in the present study and projection-space domain with the single-source in the study of Qian et al.18 Although the causal factors determining those differences need further elucidation, emphasis should be placed on the highly heartening commonality in the achieved sensitivity and specificity (100% and 95% in the present study versus 100% and 91.7% in the study of Qian et al.18) for
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e6
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
Figure 2 (a) ROC curve representing the diagnostic performance of iodine quantification in discriminating between bland and neoplastic thrombi. Iodine quantification with a thrombus iodine concentration of 0.9 mg/ml yielded an area under the curve of 0.993 (95% CI: 0.88, 1.00). (b) Interactive dot diagram showing the individual data distribution using the optimised threshold of 0.9 mg/ml. Each dot represents the iodine concentration value for each of the portal thrombi. Bland thrombi (azure dots) are displayed in the left column. Neoplastic thrombi (green dots) are arrayed in the right column. Note that all the neoplastic thrombi (n¼13) showed a greater than 0.9 mg/ml iodine concentration. All (n¼20) but one bland thrombi had a less than 0.9 mg/ml iodine concentration.
portal thrombus characterisation.18 This may represent the starting point for a clinically substantiated utilisation of iodine quantification in the setting of PVT. In line with recent research efforts, the present results show that the iodine quantification technique can provide a more direct measure of tissue iodine uptake, which is less prone to technical factors (e.g., beam-hardening and related energy shift phenomena) traditionally posing a diagnostic
dilemma for the detection of subtle differences in iodine uptake when compared with conventional enhancement measurements15,17,23e25; however, the present data also show that iodine quantification is not immune to error. In particular, a false-positive result from the iodine quantification was observed in a case of coarsely calcified thrombus, which was correctly called as non-enhancing at conventional enhancement measurements.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
e7
Figure 3 CT transverse images of a 60-year-old man with “diffuse-type’’ HCC and neoplastic thrombosis of the portal vein. (a) Non-contrast, (b) contrast-enhanced late hepatic arterial phase 120-kVp simulated, (c) colour-coded iodine map, and (d) contrast-enhanced late hepatic arterial phase images carried out 5 months earlier. Both (a, b) conventional enhancement measurements (attenuation increase ¼ 41.5 HU) and (c) iodine uptake quantification (2.7 mg/ml) allow the correct characterisation of thrombosis as malignant. Thrombus is increased in extension compared to the prior CT examination.
This finding can be explained by a breakdown of the three-material decomposition algorithm. Calcium not being among the materials on which the algorithm is calibrated d such as fat, soft tissue, and iodine d it is held responsible for skewing the ROI data in favour of factitious iodine uptake.16,19e20 Nonetheless, if one had to contextualise such a finding in the setting of PVT, a calcified thrombus is far less likely to be malignant.26 The present data have important clinical implications. Given the dichotomous therapeutic outcome of HCC patients based on the determination of the nature of PVT, improved thrombus characterisation enabled by dualenergy iodine quantification can substantially affect patient management. The limitations of the present study have to be acknowledged. The foremost potential criticism of the present research is represented by the unavailability of histopathology results for many cases in the present series. The widespread utilisation of non-invasive imaging-based diagnosis d both in the literature and in the everyday clinical practice d has to be regarded as a necessary price to balance safety and efficacy in the characterisation of portal thrombosis in the overwhelming majority of HCC patients,
who often have underlying chronic liver diseases. In all patients lacking histopathological proof, portal thrombi were not characterised by thrombus contrast-enhancement criteria per se. Along with the adoption of stringent and substantiated imaging-based criteria, the availability of a previous multiphasic MDCT examination data, played a critical role in providing a reliable reference standard, as it is unlikely for a neoplastic thrombus without any anti-angiogenic therapy to be stable in size and characteristics when compared with a previous examination performed in a timeframe of 6 months. Finally, it could be advocated that the threshold for optimised diagnostic performance of dual-energy iodine quantification was not derived a priori but rather from the present analyses.27 To avoid over-interpretation of the present diagnostic performance data, future prospective investigation in a larger cohort is needed to validate the present findings. In conclusion, the results of the present study indicate that use of dual-energy MDCT with iodine quantification is a highly accurate non-invasive approach for distinguishing bland from neoplastic thrombi of the portal vein during the
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e8
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
Figure 4 CT transverse images of a 66-year-old man with “diffuse-type” HCC and histologically proven neoplastic thrombosis of the portal vein. (a) Non-contrast, (b) contrast-enhanced late hepatic arterial phase 120-kVp simulated, (c) colour-coded iodine map. Although (a, b) conventional enhancement measurements fail to capture the thrombus enhancement (attenuation increase ¼ 14.6 HU), the use of (c) iodine quantification allows for diagnosis of neoplastic thrombus, identifying an iodine uptake of 1.4 mg/ml.
late hepatic arterial phase in patients with HCC, yielding higher sensitivity and specificity compared to conventional enhancement measurements.
References 1. Ponziani FR, Zocco MA, Campanale C, et al. Portal vein thrombosis: insight into physiopathology, diagnosis, and treatment. World J Gastroenterol 2010;16(2):143e55. 2. Ogren M, Bergqvist D, Bjorck M, et al. Portal vein thrombosis: prevalence, patient characteristics and lifetime risk: a population study based on 23,796 consecutive autopsies. World J Gastroenterol 2006;12(13):2115e9. 3. Piscaglia F, Gianstefani A, Ravaioli M, et al. Criteria for diagnosing benign portal vein thrombosis in the assessment of patients with cirrhosis and hepatocellular carcinoma for liver transplantation. Liver Transpl 2010;16(5):658e67. 4. Shah ZK, McKernan MG, Hahn PF, et al. Enhancing and expansile portal vein thrombosis: value in the diagnosis of hepatocellular carcinoma in patients with multiple hepatic lesions. AJR Am J Roentgenol 2007;188(5):1320e3. 5. Catalano OA, Choy G, Zhu A, et al. Differentiation of malignant thrombus from bland thrombus of the portal vein in patients with hepatocellular carcinoma: application of diffusion-weighted MR imaging. Radiology 2010;254(1):154e62. 6. Tarantino L, Francica G, Sordelli I, et al. Diagnosis of benign and malignant portal vein thrombosis in cirrhotic patients with hepatocellular carcinoma: color Doppler US, contrast-enhanced US, and fine-needle biopsy. Abdom Imaging 2006;31(5):537e44.
7. Rossi S, Ghiottoni G, Ravetta V, et al. Contrast-enhanced ultrasonography and spiral computed tomography in the detection and characterization of portal vein thrombosis complicating hepatocellular carcinoma. Eur Radiol 2008;18(8):1749e56. 8. Sandrasegaran K, Tahir B, Nutakki K, et al. Usefulness of conventional MRI sequences and diffusion-weighted imaging in differentiating malignant from benign portal vein thrombus in cirrhotic patients. AJR Am J Roentgenol 2003;201(6):1211e9. 9. Parvey HR, Raval B, Sandler CM. Portal vein thrombosis: imaging findings. AJR Am J Roentgenol 1994;162(1):77e81. 10. Marn CS, Francis IR. CT of portal venous occlusion. AJR Am J Roentgenol 1992;159(4):717e26. 11. Tublin ME, Dodd III GD, Baron RL. Benign and malignant portal vein thrombosis: differentiation by CT characteristics. AJR Am J Roentgenol 1997;168(3):719e23. 12. Tajima T, Honda H, Taguchi K, et al. Sequential hemodynamic change in hepatocellular carcinoma and dysplastic nodules: CT angiography and pathologic correlation. AJR Am J Roentgenol 2002;178(4):885e97. 13. Akin O, Dixit D, Schwartz L. Bland and tumor thrombi in abdominal malignancies: magnetic resonance imaging assessment in a large oncologic patient population. Abdom Imaging 2011;36(1):62e8. 14. Raza SA, Jang HJ, Kim TK. Differentiating malignant from benign thrombosis in hepatocellular carcinoma: contrast-enhanced ultrasound. Abdom Imaging 2014;39(1):153e61. 15. Ascenti G, Mileto A, Krauss B, et al. Distinguishing enhancing from nonenhancing renal masses with dual-source dual-energy CT: iodine quantification versus standard enhancement measurements. Eur Radiol 2013;23(8):2288e95. 16. Mileto A, Marin D, Nelson RC, et al. Dual energy MDCT assessment of renal lesions: an overview. Eur Radiol 2014;24(2):353e62.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9 17. Chandarana H, Megibow AJ, Cohen BA, et al. Iodine quantification with dual-energy CT: phantom study and preliminary experience with renal masses. AJR Am J Roentgenol 2011;196(6):W693e700. 18. Qian LJ, Zhu J, Zhuang ZG, et al. Differentiation of neoplastic from bland macroscopic portal vein thrombi using dual-energy spectral CT imaging: a pilot study. Eur Radiol 2012;22(10):2178e85. 19. Graser A, Johnson TR, Hecht EM, et al. Dual-energy CT in patients suspected of having renal masses: can virtual nonenhanced images replace true nonenhanced images? Radiology 2009;252(2):433e40. 20. Ascenti G, Krauss B, Mazziotti S, et al. Dual-energy computed tomography (DECT) in renal masses: nonlinear versus linear blending. Acad Radiol 2012;19(10):1186e93. 21. Wilcox RR, Keselman HJ. Modern robust data analysis methods: measures of central tendency. Psychol Methods 2003;8(3):254e74. 22. Francis IR, Cohan RH, McNulty NJ, et al. Multidetector CT of the liver and hepatic neoplasms: effect of multiphasic imaging on tumor conspicuity and vascular enhancement. AJR Am J Roentgenol 2003;180(5):1217e24.
e9
23. Kaza RK, Caoili EM, Cohan RH, et al. Distinguishing enhancing from nonenhancing renal lesions with fast kilovoltage-switching dual-energy CT. AJR Am J Roentgenol 2011;197(6):1375e81. 24. Ascenti G, Mazziotti S, Mileto A, et al. Dual-source dual-energy CT evaluation of complex cystic renal masses. AJR Am J Roentgenol 2012;199(5):1026e34. 25. Mileto A, Marin D, Ramirez-Giraldo JC, et al. Accuracy of contrastenhanced dual-energy MDCT for the assessment of iodine uptake in renal lesions. AJR Am J Roentgenol 2014;202(5):W466e74. 26. Ayuso C, Luburich P, Vilana R, et al. Calcifications in the portal venous system: comparison of plain films, sonography and CT. AJR Am J Roentgenol 1992;159(2):321e3. 27. Ochodo EA, de Haan MC, Reitsma JB, et al. Overinterpretation and misreporting of diagnostic accuracy studies: evidence of “spin”. Radiology 2013;267(2):581e8.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma G. Ascenti, C. Sofia*, S. Mazziotti, S. Silipigni, T. D’Angelo, S. Pergolizzi, E. Scribano Department of Biomedical Sciences and Morphologic and Functional Imaging, Policlinico “G. Martino”, University of Messina, Italy
art icl e i nformat ion Article history: Received 29 November 2015 Received in revised form 14 April 2016 Accepted 3 May 2016
AIM: To investigate the diagnostic accuracy of dual-energy multidetector computed tomography (MDCT) with iodine quantification compared to conventional enhancement measurements in distinguishing bland from neoplastic portal vein thrombosis in patients with hepatocellular carcinoma. MATERIAL AND METHODS: Thirty-four patients (26 men, eight women; mean age, 62 years) with hepatocellular carcinoma and portal vein thrombosis underwent contrast-enhanced dualenergy MDCT during the late hepatic arterial phase for the assessment of portal thrombosis (bland, n¼21; neoplastic, n¼13). Datasets were analysed separately by two different readers. Interobserver correlation and variability were calculated and compared with the BlandeAltman method. Diagnostic accuracy of conventional enhancement measurements and iodine quantification was calculated by setting either histopathology (n¼7) or a reference standard based on MDCT imaging criteria and thrombus evolutionary characteristics compared to a previous MDCT examination (n¼27). For iodine quantification threshold determination receiver operating characteristic (ROC) curves were drawn. p-Values <0.05 were considered significant. RESULTS: For conventional enhancement measurements and iodine quantification interobserver correlation was 98% and 96%. Enhancement measurement resulted in a sensitivity of 92.3%, specificity of 85.7%, positive predictive value (PPV) of 80%, and negative predictive value (NPV) of 94.7%. An iodine concentration of 0.9 mg/ml optimised discrimination between neoplastic and bland thrombi (area under the ROC [AUC] 0.993) resulting in a sensitivity of 100%, specificity of 95.2%, PPV of 92.9%, and NPV of 100%. The overall diagnostic accuracy of iodine quantification (97%) was significantly better than conventional enhancement measurements (88.2%; p<0.001). CONCLUSION: Compared to conventional enhancement measurements, iodine quantification improves the characterisation of portal vein thrombi during the late hepatic arterial phase in patients with hepatocellular carcinoma. Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
* Guarantor and correspondent: C. Sofia, Department of Biomedical Sciences and Morphologic and Functional Imaging, Policlinico “G. Martino”, University of Messina, 98100 Messina, Italy. Tel.: þ39 090 221 3092; fax: þ39 090 221 3720. E-mail address: carm.sofi[email protected] (C. Sofia). http://dx.doi.org/10.1016/j.crad.2016.05.002 0009-9260/Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e2
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
Introduction The distinction between bland and neoplastic portal vein thrombosis (PVT) represents a watershed for the clinical management of patients having hepatocellular carcinoma (HCC).1e3 Bland PVT, which is associated with a slowdown of the portal bloodstream, is reported in up to 40% of patients with HCC.4,5 Conversely, neoplastic thrombosis of the portal vein, which is determined by tumour invasion of the lumen of the portal vein or its tributaries, is found in 6.5e44% of patients with HCC.4,5 Beyond affecting tumour staging and patient prognosis, the occurrence of neoplastic thrombosis of the portal vein dramatically impacts the therapeutic planning by precluding a number of possible options, including surgical resection, orthotopic liver transplantation as well as chemoembolisation.1e5 Although the standard method for characterising a portal vein thrombus is represented by histopathology, imaging can play a decisive role and various imaging techniques, including Doppler ultrasonography, contrast-enhanced ultrasonography, multidetector-row computed tomography (MDCT) or magnetic resonance imaging, have long been utilised as yardsticks for identifying and characterising PVT.6e11 In MDCT, while the identification of the thrombus is largely reliant upon portal venous phase imaging, the late hepatic arterial phase is the ideal time point to determine the nature of the thrombus based on evidence of enhancement representing the most suggestive clue of thrombus malignancy.4,5,12e14 Nevertheless, inherent technical limitations of contrast-enhancement assessment, e.g., challenges in appropriate region-of-interest (ROI) placement on multiphase images, are such that radiodensity values can potentially fail to capture subtle changes in iodine uptake of the microvascular meshwork of the thrombus, thereby potentially undermining proper patient management.5,12e14 There is growing evidence that dualenergy MDCT with iodine quantification offers advantages over conventional CT for accurate quantification of iodine uptake in human tissues and provides a considerable potential for differentiating bland versus neoplastic PVT.15e18 The purpose of the present study was to investigate the diagnostic accuracy of dual-energy MDCT with iodine quantification compared to conventional enhancement measurements to distinguish between bland and neoplastic PVT.
Materials and methods This prospective, single-centre study was approved by the Institutional Review Board, and a waiver of informed consent was obtained.
Patients From April 2013 to April 2015, 45 consecutive patients with diagnosis of PVT at previous CT examination within the previous 6 months and biopsy diagnosis of HCC, who were referred for MDCT follow-up, were prospectively evaluated
using DECT. From the initial population, patients who underwent complete portal vein recanalisation and had no evidence of thrombosis were excluded (n¼11). The final study population consisted of 34 patients (26 men, eight women; mean age 62.18.9 yearsSD; age range, 48e79 years; mean body mass index, 244.7) with HCC (solitary, n¼17; multifocal, n¼11; diffusely infiltrative, n¼6; average primary lesion size: 21.417.4 mm, range, 8e80 mm), having a total of 34 thrombi (main portal vein, n¼14; right portal vein, n¼14; left portal vein, n¼6). Fig 1 shows the flowchart of study enrolment population. All these patients had undergone thrombolytic therapy after the first diagnosis of PVT and none had undergone chemotherapy (such as angiogenesis inhibitors) during the time interval between the first and the second CT examination.
Image technique Data acquisition All CT examinations of the liver were performed with a dual-source MDCT system (Somatom Definition Dual Source; Siemens Healthcare, Forchheim, Germany). The study protocol consisted of non-contrast, portal venous and equilibrium phases acquired in single-energy mode and the late hepatic arterial phase obtained in dual-energy mode (Table 1), using an automatic bolus-tracking technique (CARE Bolus, Siemens Healthcare; Forchheim; Germany) for the scanning delay. The late hepatic arterial phase was started automatically 12 seconds after the trigger threshold (100 HU) was reached at the level of the supracoeliac abdominal aorta. Portal venous and equilibrium phases were acquired at a fixed time delay of 80 and 180 seconds after the beginning of the contrast medium injection. All patients received 1.4 ml/kg body weight (560 mgI/kg of body weight) of a non-ionic contrast-medium (iomeprol, 400 mg iodine/ml; Iomeron 400; Bracco Diagnostics, Princeton, NJ, USA) that was administered with a dual-chamber mechanical power injector (Stellant D; Medrad, Indianola, PA, USA), at a rate of 5 ml/s through an intravenous catheter inserted into an antecubital vein. This was followed by a 30 ml saline chaser at the same injection rate. This protocol represents the standard CT approach for the study of patients with hepatic cirrhosis and HCC.
Data reconstruction From single-energy datasets, 3-mm-thick transverse images were reconstructed and only non-contrast images were utilised for the subsequent image analysis in this study. From dual-energy datasets, 3-mm-thick contiguous transverse 120-kVp simulated linear-blended images, which approximate the image quality of a standard 120-kVp scan of the abdomen, were obtained with a combination of image data averaging 30% from the low-kVp data and 70% from the high-kVp data, using a dual-energy dedicated convolution kernel (D30f). All datasets were networked to a secondary workstation where each patient’s late hepatic arterial phase study was
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
e3
Figure 1 Flowchart of the study enrolment population based on adopted inclusion and exclusion criteria and reference standard. y MDCT diagnosis was based on direct infiltration of primary tumour into the portal vein, expansion of the involved vessel, or unequivocal evidence of contrast enhancement [4, 5]. z Findings of the previous 6 months MDCT examinations relied upon depiction of a substantial growth in thrombus size, defined as an increase in the longest transverse diameter by greater than 30%. By comparison, stability in size or shrinkage was considered suggestive of bland nature of the thrombus.
post-processed with the commercially available dualenergy software (syngo Dual Energy version MMWP2010A, Siemens), using an iodine-subtraction algorithm (Liver VNC, Syngo Dual Energy version MMWP 2010A, Table 1 Acquisition parameters for single-energy and dual-energy multidetector computed tomography (MDCT). MDCT parameters
Single-energy MDCT
Dual-energy MDCT
Detector collimation (mm) Tube voltage (kV) Tube currentetime product (mAs) Gantry revolution time (s) Acquisition mode Pitch CTDIvol (mGy) Reconstructed section thickness (mm) Reconstruction algorithm Reconstruction kernel
640.6 120 200
141.2 140/80 95/510
0.5
0.5
Helical 0.75 11.1 3
Helical 0.75 13.9 3
Filtered back projection Soft tissue (B31f)
Filtered back projection Soft tissue (D30f)
CTDIvol, volume computed tomography dose index.
Siemens) as an image-based analysis of the low- and high-energy kVp images. Based on the three-material decomposition algorithm, this software enables spectral extraction of materials having a sufficiently high k-edge (e.g., iodine and calcium).17,19e20 The iodine-dependent slope for dual-energy three-material decomposition was also calibrated by placing a large ROI over the aortic lumen, at the level of the coeliac trunk origin (mean ROI size, 190 pixels). The manufacturer default value of 50% iodine overlay display, as well as the values for soft-tissue and fat attenuation at 80 and 140 kVp, were utilised for the analysis.
Image analysis All data were analysed independently and in a blinded fashion by two radiologists with 10 and 5 years of experience, respectively, in dual-energy abdominal imaging. The reviewers were blinded to the outcome, previously determined for each case. ROI measurements were obtained over each portal thrombus on non-contrast and contrast-enhanced late hepatic arterial phase images, both 120-kVp simulated
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e4
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
linear-blended and colour-coded iodine maps. The mean radiodensity values and iodine concentrations were recorded by drawing circular ROIs five times for each portal thrombus; average values from three ROI measurements were then used for statistical analyses, after the maximal and minimal values of five measurements were excluded (“winsorised mean” procedure).21 Each ROI included at least two-thirds of the area of the thrombus while encompassing as much of the areas showing either the highest attenuation (on 120-kVp simulated linearblended images) or highest coloured pixel content (on colour-coded iodine maps), as possible. During ROI measurements the wall of the vessel or prominent artefacts, if any, were carefully avoided. For the entire period of analysis, the precision of ROI measurements was further assured by a real-time guidance and thrombus tracking facilitated by the simultaneous display of portal venous images serving as a roadmap on an adjacent screen for each single case. Besides ensuring reproducibility of the data using repeated ROI samples, reliability of the results was also warranted in terms of consistency by keeping the size, shape, and position of the ROIs constant among the repeated measurements. While iodine quantification measurements were directly provided by the dual-energy software with the ROIs on colour-coded iodine maps, conventional contrastenhancement values of the thrombi were determined using the equation (HUpostcontrast HUnoncontrast)
where HUpostcontrast is the mean radiodensity value on the 120-kVp simulated linear-blended images and HUnoncontrast is the mean radiodensity value on non-contrast images.
Reference standard A histopathological proof of thrombus was available in seven of 34 patients (20,5%): autoptic specimens were available in four and surgical specimens were available in three patients who underwent en-bloc hepatic resection. In the remaining 27 portal vein thrombi, currently accepted MDCT criteria for characterisation of portal thrombus, together with comparative evaluation of thrombus features with a MDCT examination performed within a time-interval of 3e6 months, were used as a reference standard. CT criteria are represented by: (a) direct infiltration into the portal vein by the primary tumour4,5,11,18; (b) expansion of the involved vessel (vessel diameter, >2.6 cm for the main portal vein; >2.3 cm for the right portal vein; >2.2 cm for the left portal vein), along with a disproportionate enlargement of the vessel when compared with non-affected same-order branches in the same lobe4,5,11,18; (c) unequivocal evidence of enhancement on contrast-enhanced MDCT images.4,5,11,18 If, at least, two of the above-mentioned criteria were present, the thrombus was considered malignant; otherwise the thrombus was defined as bland.4,5,11,18
A substantial growth in thrombus size compared to prior MDCT examination d defined as an increase in the longest transverse diameter of thrombus by greater than 30% d was considered indicative of the neoplastic nature of thrombus. By contrast, stability in size or shrinkage was considered suggestive of bland nature.3
Statistical analysis Numerical values of continuous variables are expressed as means standard deviations, and categorical variables are expressed as frequencies or percentages. Interobserver variation in measurement of thrombus attenuation and iodine quantification was assessed in terms of Pearson’s correlation between the measurements provided by the two readers. Difference in measurements was determined by BlandeAltman analysis, by calculating the mean and 95% confidence interval (CI) of the arithmetic differences between measurements by two readers. If the differences are normally distributed, 95% of the differences will lie within a range of 1.96 SD of the mean difference. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) of conventional enhancement measurements and iodine quantification were calculated from c2 tests of contingency, setting as reference standard either histopathology or MDCT features together with comparative evaluation of thrombus characteristics with a prior CT examination (this latter not older than 6 months). With conventional enhancement measurements, thrombus enhancement was defined as a 20 HU increase in attenuation seen on 120-kVp simulated linear-blended late hepatic arterial phase when compared with noncontrast images.4,5 By comparison, in light of the discrepancy between the currently reported thresholds of iodine uptake using the iodine quantification for different abdominal applications, the optimal threshold for distinguishing bland from neoplastic thrombus was determined in a similar manner to Qian et al.15e18 In particular, receiver operating characteristic (ROC) curve analysis was utilised to establish the optimal threshold for iodine concentration; the area under the ROC curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and relative confidence interval (CI) were calculated. Differences in sensitivity and specificity between conventional enhancement measurements and iodine quantification were assessed using the McNemar test. Statistical significance was assumed to be p<0.05. All statistical analyses were performed using statistical software (MedCalc, version 12.7.1.0, MedCalc, Mariakerke, Belgium).
Results On histopathological examination, five out of seven thrombi were neoplastic and the remaining two were bland. According to conventional CT parameters, eight
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
thrombi out of the remaining 27 were neoplastic and 19 bland. In particular, all eight neoplastic thrombi showed substantial growth in size compared to the prior CT examination (average longest transverse diameter at baseline examination, 20.9 cm; average longest transverse diameter at re-examination, 3.70.5 cm; mean percentage of increase in size, 85%), while the remaining 19 bland portal vein thrombi displayed no substantial growth (average longest transverse diameter at baseline examination, 1.80.8 cm SD; average longest transverse diameter at reexamination, 1.90.2 cm; mean percentage of increase in size, 5.5%). Additionally, among neoplastic thrombi, direct neoplastic invasion of the portal vein was found in six out of eight thrombi, expansion of the involved vessel was reported in seven out of eight, and unequivocal evidence of enhancement was found in seven out of eight. In the remaining 19 thrombi, none of these conditions was met at conventional CT examination and were hence considered bland. In the presentseries bland thrombi (main portal vein, n¼11; right portal vein, n¼8; left portal vein, n¼2) showed a mean post-contrast attenuation increase of 8.68.5 HU for Reader I and of 8.87.1 HU for Reader II (p¼0.82). The average iodine concentration for this group was 0.50.4 mg/ml for Reader I and 0.60.4 mg/ml for Reader II (p 0.26). Conversely, neoplastic thrombi (main portal vein, n¼3; right portal vein, n¼6; left portal vein, n¼4) showed a mean post-contrast attenuation increase of 42.713.6 HU for Reader I and 43.311 HU for Reader II (p¼0.59). Average iodine concentration for this group was 2.30.6 mg/ml for Reader I and 2.40.4 mg/ml SD for Reader II (p¼0.18). Correlation among the two readers’ measurements was 98% for post-contrast attenuation increase (p<0.05) and 96% for iodine concentration (p<0.05). Interobserver mean difference was 0.3 for post-contrast attenuation increase (SD¼3.7; 95% CI: 0.9 to 1.6; p¼0.59) and 0.1 for iodine quantification (SD¼0.29; 95% CI: 0.01 to 0.19; p¼0.08). Using conventional enhancement measurements, there were 12 true-positive, three false-positive, one false-negative, and 18 true-negative results, yielding a sensitivity of 92.3% (95% CI: 63.9%, 98.7%), a specificity of 85.7% (95% CI: 63.6%, 96.8%), a PPV of 80% (95% CI: 51.9%, 95.4%), and a NPV of 94.7% (95% CI: 73.9%, 99.1%). The overall diagnostic accuracy was of 88.2%. The ROC analysis demonstrated that 0.9 mg/ml was the optimal threshold (AUC, 0.993; 95% CI: 0.88, 1.00) for discriminating between neoplastic and bland thrombi when using the iodine quantification (Fig 2), with 13 true-positive (Figs 3e4), one false-positive, no falsenegative, and 20 true-negative results. Overall, iodine quantification yielded a sensitivity of 100% (95% CI: 75.3%, 100%), a specificity of 95.2% (95% CI: 76.2%, 99.9%), a PPV of 92.9% (95% CI: 66.1%, 99.8%), and a NPV of 100% (95% CI: 83.2%, 100%), with an overall diagnostic accuracy of 97%. The McNemar test showed that, when compared with conventional enhancement measurements, iodine
e5
quantification allowed for significantly better diagnostic accuracy in distinguishing neoplastic from bland thrombi (p<0.001).
Discussion Imaging-based determination of the aetiology of PVT is of paramount value for clinical decision making in patients with HCC.1e11 The results of the present study indicate that, when compared with conventional enhancement measurements, dual-energy MDCT with iodine quantification yields significantly improved diagnostic performance for distinguishing bland from neoplastic PVT based on the iodine-uptake assessment. Of note, the present data indicates that an iodine concentration threshold of 0.9 mg/ ml optimises the discrimination between bland and neoplastic thrombi, in general accordance with preliminary data reported by Qian et al.,18 who showed the feasibility of dual-energy iodine quantification to differentiate between neoplastic and bland thrombi during the hepatic portal venous phase. It is important to note that the clinical perspective offered by Qian et al. was, somewhat limited by the lack of comparison to conventional enhancement measurements during the late hepatic arterial phase of enhancement.18 Despite the challenges encountered in portal thrombus detection, the latter remains the stronghold for characterising portal thrombus based on iodine uptake.22 As Qian et al.18 acknowledged in their study, poor thrombus enhancement in the hepatic portal venous phase can yield false-negative findings.12e14,18 It is well known that the venous phase is the most accurate for thrombus detection and that in several cases neoplastic thrombus vasculature may mimic an initial inflow of iodinate blood into the portal circulation during the late arterial phase. On the other hand, the arterial phase is fundamental to characterise the nature of the thrombus.22 In the present evaluation, ROI measurement was further assured by real-time guidance and thrombus tracking facilitated by simultaneous display of portal venous images serving as a roadmap on an adjacent screen for each single case. There are discrepancies in the optimal threshold for thrombus iodine-uptake identification proposed in the present study (0.9 mg/ml) and that recommended by Qian et al. (1.14 mg/ml).18 Beyond the variation in the selected imaging study phase (late hepatic arterial versus hepatic portal venous), it is conceivable that these differences could be related to the technical approaches adopted for iodine quantification, namely the image-space domain with the dual-source in the present study and projection-space domain with the single-source in the study of Qian et al.18 Although the causal factors determining those differences need further elucidation, emphasis should be placed on the highly heartening commonality in the achieved sensitivity and specificity (100% and 95% in the present study versus 100% and 91.7% in the study of Qian et al.18) for
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e6
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
Figure 2 (a) ROC curve representing the diagnostic performance of iodine quantification in discriminating between bland and neoplastic thrombi. Iodine quantification with a thrombus iodine concentration of 0.9 mg/ml yielded an area under the curve of 0.993 (95% CI: 0.88, 1.00). (b) Interactive dot diagram showing the individual data distribution using the optimised threshold of 0.9 mg/ml. Each dot represents the iodine concentration value for each of the portal thrombi. Bland thrombi (azure dots) are displayed in the left column. Neoplastic thrombi (green dots) are arrayed in the right column. Note that all the neoplastic thrombi (n¼13) showed a greater than 0.9 mg/ml iodine concentration. All (n¼20) but one bland thrombi had a less than 0.9 mg/ml iodine concentration.
portal thrombus characterisation.18 This may represent the starting point for a clinically substantiated utilisation of iodine quantification in the setting of PVT. In line with recent research efforts, the present results show that the iodine quantification technique can provide a more direct measure of tissue iodine uptake, which is less prone to technical factors (e.g., beam-hardening and related energy shift phenomena) traditionally posing a diagnostic
dilemma for the detection of subtle differences in iodine uptake when compared with conventional enhancement measurements15,17,23e25; however, the present data also show that iodine quantification is not immune to error. In particular, a false-positive result from the iodine quantification was observed in a case of coarsely calcified thrombus, which was correctly called as non-enhancing at conventional enhancement measurements.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
e7
Figure 3 CT transverse images of a 60-year-old man with “diffuse-type’’ HCC and neoplastic thrombosis of the portal vein. (a) Non-contrast, (b) contrast-enhanced late hepatic arterial phase 120-kVp simulated, (c) colour-coded iodine map, and (d) contrast-enhanced late hepatic arterial phase images carried out 5 months earlier. Both (a, b) conventional enhancement measurements (attenuation increase ¼ 41.5 HU) and (c) iodine uptake quantification (2.7 mg/ml) allow the correct characterisation of thrombosis as malignant. Thrombus is increased in extension compared to the prior CT examination.
This finding can be explained by a breakdown of the three-material decomposition algorithm. Calcium not being among the materials on which the algorithm is calibrated d such as fat, soft tissue, and iodine d it is held responsible for skewing the ROI data in favour of factitious iodine uptake.16,19e20 Nonetheless, if one had to contextualise such a finding in the setting of PVT, a calcified thrombus is far less likely to be malignant.26 The present data have important clinical implications. Given the dichotomous therapeutic outcome of HCC patients based on the determination of the nature of PVT, improved thrombus characterisation enabled by dualenergy iodine quantification can substantially affect patient management. The limitations of the present study have to be acknowledged. The foremost potential criticism of the present research is represented by the unavailability of histopathology results for many cases in the present series. The widespread utilisation of non-invasive imaging-based diagnosis d both in the literature and in the everyday clinical practice d has to be regarded as a necessary price to balance safety and efficacy in the characterisation of portal thrombosis in the overwhelming majority of HCC patients,
who often have underlying chronic liver diseases. In all patients lacking histopathological proof, portal thrombi were not characterised by thrombus contrast-enhancement criteria per se. Along with the adoption of stringent and substantiated imaging-based criteria, the availability of a previous multiphasic MDCT examination data, played a critical role in providing a reliable reference standard, as it is unlikely for a neoplastic thrombus without any anti-angiogenic therapy to be stable in size and characteristics when compared with a previous examination performed in a timeframe of 6 months. Finally, it could be advocated that the threshold for optimised diagnostic performance of dual-energy iodine quantification was not derived a priori but rather from the present analyses.27 To avoid over-interpretation of the present diagnostic performance data, future prospective investigation in a larger cohort is needed to validate the present findings. In conclusion, the results of the present study indicate that use of dual-energy MDCT with iodine quantification is a highly accurate non-invasive approach for distinguishing bland from neoplastic thrombi of the portal vein during the
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
e8
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9
Figure 4 CT transverse images of a 66-year-old man with “diffuse-type” HCC and histologically proven neoplastic thrombosis of the portal vein. (a) Non-contrast, (b) contrast-enhanced late hepatic arterial phase 120-kVp simulated, (c) colour-coded iodine map. Although (a, b) conventional enhancement measurements fail to capture the thrombus enhancement (attenuation increase ¼ 14.6 HU), the use of (c) iodine quantification allows for diagnosis of neoplastic thrombus, identifying an iodine uptake of 1.4 mg/ml.
late hepatic arterial phase in patients with HCC, yielding higher sensitivity and specificity compared to conventional enhancement measurements.
References 1. Ponziani FR, Zocco MA, Campanale C, et al. Portal vein thrombosis: insight into physiopathology, diagnosis, and treatment. World J Gastroenterol 2010;16(2):143e55. 2. Ogren M, Bergqvist D, Bjorck M, et al. Portal vein thrombosis: prevalence, patient characteristics and lifetime risk: a population study based on 23,796 consecutive autopsies. World J Gastroenterol 2006;12(13):2115e9. 3. Piscaglia F, Gianstefani A, Ravaioli M, et al. Criteria for diagnosing benign portal vein thrombosis in the assessment of patients with cirrhosis and hepatocellular carcinoma for liver transplantation. Liver Transpl 2010;16(5):658e67. 4. Shah ZK, McKernan MG, Hahn PF, et al. Enhancing and expansile portal vein thrombosis: value in the diagnosis of hepatocellular carcinoma in patients with multiple hepatic lesions. AJR Am J Roentgenol 2007;188(5):1320e3. 5. Catalano OA, Choy G, Zhu A, et al. Differentiation of malignant thrombus from bland thrombus of the portal vein in patients with hepatocellular carcinoma: application of diffusion-weighted MR imaging. Radiology 2010;254(1):154e62. 6. Tarantino L, Francica G, Sordelli I, et al. Diagnosis of benign and malignant portal vein thrombosis in cirrhotic patients with hepatocellular carcinoma: color Doppler US, contrast-enhanced US, and fine-needle biopsy. Abdom Imaging 2006;31(5):537e44.
7. Rossi S, Ghiottoni G, Ravetta V, et al. Contrast-enhanced ultrasonography and spiral computed tomography in the detection and characterization of portal vein thrombosis complicating hepatocellular carcinoma. Eur Radiol 2008;18(8):1749e56. 8. Sandrasegaran K, Tahir B, Nutakki K, et al. Usefulness of conventional MRI sequences and diffusion-weighted imaging in differentiating malignant from benign portal vein thrombus in cirrhotic patients. AJR Am J Roentgenol 2003;201(6):1211e9. 9. Parvey HR, Raval B, Sandler CM. Portal vein thrombosis: imaging findings. AJR Am J Roentgenol 1994;162(1):77e81. 10. Marn CS, Francis IR. CT of portal venous occlusion. AJR Am J Roentgenol 1992;159(4):717e26. 11. Tublin ME, Dodd III GD, Baron RL. Benign and malignant portal vein thrombosis: differentiation by CT characteristics. AJR Am J Roentgenol 1997;168(3):719e23. 12. Tajima T, Honda H, Taguchi K, et al. Sequential hemodynamic change in hepatocellular carcinoma and dysplastic nodules: CT angiography and pathologic correlation. AJR Am J Roentgenol 2002;178(4):885e97. 13. Akin O, Dixit D, Schwartz L. Bland and tumor thrombi in abdominal malignancies: magnetic resonance imaging assessment in a large oncologic patient population. Abdom Imaging 2011;36(1):62e8. 14. Raza SA, Jang HJ, Kim TK. Differentiating malignant from benign thrombosis in hepatocellular carcinoma: contrast-enhanced ultrasound. Abdom Imaging 2014;39(1):153e61. 15. Ascenti G, Mileto A, Krauss B, et al. Distinguishing enhancing from nonenhancing renal masses with dual-source dual-energy CT: iodine quantification versus standard enhancement measurements. Eur Radiol 2013;23(8):2288e95. 16. Mileto A, Marin D, Nelson RC, et al. Dual energy MDCT assessment of renal lesions: an overview. Eur Radiol 2014;24(2):353e62.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002
G. Ascenti et al. / Clinical Radiology xxx (2016) e1ee9 17. Chandarana H, Megibow AJ, Cohen BA, et al. Iodine quantification with dual-energy CT: phantom study and preliminary experience with renal masses. AJR Am J Roentgenol 2011;196(6):W693e700. 18. Qian LJ, Zhu J, Zhuang ZG, et al. Differentiation of neoplastic from bland macroscopic portal vein thrombi using dual-energy spectral CT imaging: a pilot study. Eur Radiol 2012;22(10):2178e85. 19. Graser A, Johnson TR, Hecht EM, et al. Dual-energy CT in patients suspected of having renal masses: can virtual nonenhanced images replace true nonenhanced images? Radiology 2009;252(2):433e40. 20. Ascenti G, Krauss B, Mazziotti S, et al. Dual-energy computed tomography (DECT) in renal masses: nonlinear versus linear blending. Acad Radiol 2012;19(10):1186e93. 21. Wilcox RR, Keselman HJ. Modern robust data analysis methods: measures of central tendency. Psychol Methods 2003;8(3):254e74. 22. Francis IR, Cohan RH, McNulty NJ, et al. Multidetector CT of the liver and hepatic neoplasms: effect of multiphasic imaging on tumor conspicuity and vascular enhancement. AJR Am J Roentgenol 2003;180(5):1217e24.
e9
23. Kaza RK, Caoili EM, Cohan RH, et al. Distinguishing enhancing from nonenhancing renal lesions with fast kilovoltage-switching dual-energy CT. AJR Am J Roentgenol 2011;197(6):1375e81. 24. Ascenti G, Mazziotti S, Mileto A, et al. Dual-source dual-energy CT evaluation of complex cystic renal masses. AJR Am J Roentgenol 2012;199(5):1026e34. 25. Mileto A, Marin D, Ramirez-Giraldo JC, et al. Accuracy of contrastenhanced dual-energy MDCT for the assessment of iodine uptake in renal lesions. AJR Am J Roentgenol 2014;202(5):W466e74. 26. Ayuso C, Luburich P, Vilana R, et al. Calcifications in the portal venous system: comparison of plain films, sonography and CT. AJR Am J Roentgenol 1992;159(2):321e3. 27. Ochodo EA, de Haan MC, Reitsma JB, et al. Overinterpretation and misreporting of diagnostic accuracy studies: evidence of “spin”. Radiology 2013;267(2):581e8.
Please cite this article in press as: Ascenti G, et al., Dual-energy CT with iodine quantification in distinguishing between bland and neoplastic portal vein thrombosis in patients with hepatocellular carcinoma, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.05.002