Quantitative dual-energy CT for evaluating hepatocellular carcinoma after transarterial chemoembolization

We aimed to investigate the role of the quantitative parameters of dual-energy computed tomography (DECT) in evaluating patients with hepatocellular carcinoma (HCC) treated by transarterial chemoembolization (TACE). We retrospectively identified 80 HCC patients (mean age, 56 years; 61 men) treated by TACE who received contrast-enhanced DECT and were retreated by TACE within 7 days between November 2018 and December 2019. Taking digital subtraction angiography (DSA) and CT images as reference standard, two readers measured and calculated the values of normalized iodine concentration at arterial phase (NICAP), normalized iodine concentration at portal venous phase (NICPP), iodine concentration difference (ICD), arterial iodine fraction (AIF) and slope of the spectral Hounsfield unit curve (λHu) by placing matched regions of interests (ROIs) within the tumor active area (TAA), adjacent normal hepatic parenchyma (ANHP) and tumor necrotic area (TNA). Differences between the parameters were analyzed by the Kruskal–Wallis H test. Receiver operating characteristic analysis of the parameters performance in differentiating the three tissues types was performed. AIF exhibited a good performance in distinguishing TAA (0.93 ± 0.31) and ANHP (0.18 ± 0.14), the areas under the receiver operating characteristic curve (AUC) was 0.989, while the λHu exhibited an excellent performance in distinguishing TAA (3.32 ± 1.24) and TNA (0.29 ± 0.27), with an AUC of 1.000. In conclusion, quantitative DECT can be effectively used to evaluate the tumor viability in HCC patients treated by TACE.

Patient demographics. Between November 2018 and December 2019, 80 consecutive patients (61men, 19 women) with HCC who underwent conventional TACE (C-TACE) (68 patients) or drug-eluting microsphere transarterial chemoembolization (DEM-TACE) (12 patients) intervention in the preceding 1-3 months were included in this study. Patients were enrolled according to the following criteria: (a) diagnosed with HCC 15 ; (b) underwent TACE intervention in the preceding 1-3 months; (c) were to be retreated with TACE within 7 days after CT review. Among the 159 patients who were initially deemed qualified for the study, 79 were excluded for the following reasons: (a) could not achieve the required injection rate (3.5 mL/s, n = 5); (b) did not undergo retreatment with TACE within 7 days or were considered to have absolute contraindication for TACE 16 . (n = 25); (c) no TAA on CT images and DSA images(n = 10); (d) extensive lipiodol deposition, the area of non-lipiodol deposition regions was less than 0.5 cm 2 , leading to inaccurate measurement (n = 30); and (e) diffuse HCC with no ANHP to measure (n = 9). Figure 1 presents the patient enrollment workflow diagram.
CT examinations. First, 1.2 mL/kg of iodine contrast medium (iodixanol, 320 mg I/mL; Hengrui Healthcare, Jiangsu, China) was intravenously injected. Then, third-generation dual-source CT (SOMATOM Force, Siemens Healthcare, Forchheim, Germany) was performed with multiple phases. Patients were instructed to fast for four hours before the CT scan.
All patients were positioned with their arms up on the scanning table in the supine position. Scanning was performed from the top of the diaphragm to the lower edge of the liver. First, an unenhanced CT scan of the abdomen was performed with a dual-energy model. Enhanced CT acquisitions were triggered using the bolus www.nature.com/scientificreports/ tracking technique, with the regions of interests (ROIs) placed at the beginning of the abdominal aorta. Arterial phase imaging was performed using automatic image-triggering software (preMonitoring; Siemens Healthcare). The arterial phase scan began after an 8-s delay at a rate of 3.5 mL/s, when the trigger attenuation threshold reached 100 HU. Then, 40 mL of saline solution was flushed at the same flow rate. A portal venous phase scan was conducted 24-s after the end of the arterial phase. The scanning protocol was as follows: collimator width, 128 × 0.6 mm; field of view (FOV), 300 mm; and pitch, 0.6. Tube A was operated at a peak voltage of 100 kV and 180 quality reference mAs (Siemens Healthcare). Tube B was operated at 150 Sn kV and 90 quality reference mAs. The data were reconstructed with a slice thickness of 1.5 mm.
TACE retreatment. Digital subtraction angiography (DSA) examination is a part of TACE retreatment, and we only collected data from patients who required retreatment clinically without additional examinations. The Seldinger method is used to intubate the femoral artery via percutaneous puncture and insert the catheter into the abdominal trunk or common hepatic artery for DSA to acquire DSA images. Superior mesenteric artery angiography was performed, and attention was paid to finding collateral blood supply. The location, size, number and blood supply arteries of the tumor were determined by careful analysis of the angiographic findings. The chemotherapeutic drugs and embolic agents were mixed together and injected through the blood supply artery branch of the tumor 17 .
Reference standard. Taking the DSA and CT images as references standard, two abdominal radiologists with more than five years of experience independently interpreted the CT images according to LI-RADS standards but were blinded to the DSA results. LI-RADS-Treatment Response (LR-TR) Viable was considered to be TAA with nodular, mass-like, or thick irregular tissue in or along the treated lesion with any of the following: arterial phase hyperenhancement; washout appearance; or enhancement similar to pretreatment. Two interventionalists with more than ten years of experience interpreted the TACE on picture archiving and communication system (PACS) and were blinded to the CT results; TAA was considered to exist when tumor staining was observed. A senior radiologist was consulted for any disagreement. Vital tissue was considered to exist when both the TACE and CT results suggested TAA.
Postprocessing of DECT images. Images were transferred to a postprocessing software workstation (Syngo via, Siemens Healthcare). When there were multiple lesions, the three largest lesions were selected for measurement. Two types of reconstructed images were measured: color-coded iodine maps and monoenergetic images of 151 energies ranging from 40 to 190 keV. All ROIs of TAA, ANHP, TNA and abdominal aorta were manually plotted at the same location and size on arterial phase and portal venous phase to obtain the values of iodine concentration (IC) and standardized iodine concentration (NIC) on the iodine map. Spectral Hounsfield unit curves were obtained from the monoenergetic images of arterial phase. The normalized iodine concentration at arterial phase (NICAP), normalized iodine concentration at portal venous phase (NICPP), iodine concentration difference (ICD), arterial iodine fraction (AIF) and the slope of the spectral Hounsfield unit curve (λ Hu ) were then calculated ( Table 1). The λ Hu was calculated as: λ Hu = (HU 40kev − HU 70kev )/30) 18,19 on arterial phase. The HU 40keV indicates that the CT value was gauged on 40 keV images, and HU 70keV indicates that the CT value was gauged on 70 keV images. The NIC was calculated as: NIC = IC lesion /IC aorta , where IC lesion and IC aorta are the iodine level of the lesion and the aorta on the same slice, respectively; the iodine concentration in the lesion is normalized to the aortic iodine concentration to minimize the difference between patients 20 .
The ROIs of TAA, NAHP and TNA were round and were selected to avoid large blood vessels, iodized oil deposits and surrounding artifacts. For TNA, the area was selected in TACE-treated tumors that were not enhanced after repeated measurements by radiologists in three-phase enhanced images and single-energy images. ANHP was chosen to be close to the tumor and 1 cm away from it to avoid possible effects of microvascular invasion on the measured iodine concentration 17,21 . When measuring nodular lesions, tumor tissues were included as much as possible. When measuring massive lesions, multiple points were measured, and the mean values were calculated. All of the ROIs of one person were of the same size and at the same level of the scan, which depended on whichever had the smaller ROI. All data were measured at different times by two independent radiologists to avoid measurement errors, and the results were averaged.

Statistical analyses.
Continuous variables are reported as the mean ± standard deviation. All statistical analyses were performed with statistical software (SPSS, version 21.0, IBM Corporation, Armonk, NY, USA;

Results
Relevant demographic and regions of interest (ROIs) characteristics were outlined in

Intraclass correlation coefficients between two radiologists. The intraclass correlation coefficients
(ICC) of the TAA, ANHP and TNA measurements were evaluated. The agreements of the measured values were good (between 0.760 and 0.977). The results are shown in Table 3.  www.nature.com/scientificreports/ Differences in dual-energy CT parameters between different tissues. The differences in parameters between different tissues are shown in Table 4. The λ Hu , normalized iodine concentration at arterial phase (NICAP) and normalized iodine concentration at portal venous phase (NICPP) of the TAA were 3.32 ± 1.24, 16.51 ± 6.75% and 46.68 ± 15.10%, respectively, which were significantly higher than those in the ANHP(0.71 ± 0.39, 2.66 ± 2.05% and 39.48 ± 10.65%, respectively, P < 0.05) and TNA (0.29 ± 0.27, 2.28 ± 1.47%, and 9.30 ± 7.46%, respectively, P < 0.05); The iodine concentration difference(ICD) of the TAA was 0.56 ± 0.46 mg/ mL, which was lower than that of ANHP (1.49 ± 0.54 mg/mL). The AIF of the TAA was 0.93 ± 0.31, which was significantly higher than that of the ANHP (0.18 ± 0.14, P < 0.05) but not significantly different from the TNA (0.99 ± 0.86, P = 0.251). Figures 2, 3 and 4 provide representative images of the DECT parametric maps. Table 4. Comparison of various parameters in different liver tissues after TACE. λ Hu , slope of the spectral Hounsfield unit curve; NICAP, normalized iodine concentration at arterial phase; NICPP, normalized iodine concentration at portal venous phase; ICD, iodine concentration difference; AIF, arterial iodine fraction; TAA, tumor active area; ANHP, adjacent normal hepatic parenchyma; TNA, tumor necrotic area. *Indicates a significant difference between TAA and ANHP (P < 0.05). # Indicates a significant difference between TNA and ANHP (P < 0.05). & Indicates a significant difference between TAA and TNA (P < 0.05).
Parameters TAA (mean ± SD) ANHP (mean ± SD) TNA (mean ± SD) P  Figure 2. This is a patient who had a DECT examination 7 weeks after the conventional-TACE, showed that lipiodol deposited around the tumor, and TAA and TNA in the tumor after treatment. A large lesion with scattered lipiodol deposition is seen within the right lobe of the liver. The center non-enhancing area was TNA. Peripheral soft tissue nodularity is noted on axial non-enhanced CT (red arrow, A), with slight enhancement on arterial phase (B), and portal venous phase (C), which was considered by radiologists to be TAA. ROIs of TAA, ANHP, and TNA placed on arterial phase and portal phase iodine map (D and E) showed the IC and NIC of TAA to be 1. www.nature.com/scientificreports/ There were no significant differences in the parameters results of 12 patients treated by DEM-TACE and 68 patients treated by lipiodol TACE. The corresponding results were added in the supplementary materials.
Diagnostic performance of the parameters across different tissues. The diagnostic performance of the parameters across different tissues is summarized in Table 5. The receiver operating characteristic (ROC) curve showed that λ Hu , NICAP, ICD, and AIF have a high efficiency for the differential diagnosis of the TAA and ANHP (areas under the ROC curve (AUC): 0.987, 0.988, 0.909, and 0.989, respectively). All of the parameters exhibited considerable sensitivity and specificity. The NICPP exhibited low efficiency for the discrimination of the TAA and ANHP (AUC 0.686). λ Hu , NICAP, and NICPP exhibited high performance for the differential diagnosis of the TAA and TNA (AUC: 1.000,0.997, and 0.962, respectively). The ICD and AIF exhibited weak performance for the differential diagnosis of the TAA and TNA (AUC: 0.717, and 0.688, respectively) (Fig. 5).

Discussion
DECT has advantages in providing artifact reduction 22 , additional information regarding tissue composition 23,24 , and radiation dose reduction 25 . In this study, we found that the DECT parameters, especially λ Hu at the arterial phase and NICAP, showed high efficiency for discriminating the different tissues in patients with HCC, with an AUC values of 0.987 and 0.988, for the recognition of the TAA and ANHP, respectively, and an AUC values of 1.000, and 0.997 for the recognition of the TAA and TNA, respectively. The quantitative parameters of DECT may help physicians to judge the effect of treatment and make correct clinical decisions.
The linear decay coefficient of tissues decreases with increasing keV values. However, the decline rate of each tissue is different 13,26 , that is, the λ Hu can show energy attenuation and depends on the physical and chemical properties of the material itself. The CT value of the material is greatly attenuated when energy changes are maintained at a low level. The energy spectrum curve of the TAA measured in this study was steep and straight from 40 keV to 70 keV , consistent with the literature 18,19 . The λ Hu can significantly reflect the blood supply characteristics of different tissues at arterial phase. The λ Hu of different tissues in this study showed a decline with varying slopes, among which the λ Hu of the TAA was significantly larger than those of the ANHP and TNA. ROC analysis showed that the λ Hu was highly effective in discriminating various tissues. A value higher than 1.32 was more likely to indicate a TAA. Therefore, the λ Hu is of great value in the follow-up post TACE. Changes in the λ Hu in Figure 3. DECT images of a 58-year-old man with HCC undergoing DEM-TACE 6 weeks ago, showed that no lipiodol deposition in the original tumor lesions, but TAA and TNA were present. The right hepatic lobe showed a low-attenuation lesion on unenhanced CT image (A), significant enhancement at arterial phase (B), and persistent enhancement at the portal venous phase(C). TNA had no enhancement. ROIs of TAA, ANHP, and TNA placed on arterial and portal iodine map (D and E) showed IC and NIC of TAA to be 1.5 mg/ml and 20.1% at arterial phase and 1.8 mg/ml and 48.5% at portal venous phase, respectively. The IC and NIC of ANHP was 0.3 mg/ml and 4.1% at arterial phase, and 1.1 mg/ml and 30.2% at portal venous phase, respectively. The IC and NIC of TNA was 0.1 mg/ml and 0.9% at arterial phase and 0.5 mg/ml and 11.9% at portal venous phase, respectively. The λHu of TAA, ANHP and TNA were 2.86, 0.53 and 0.23, respectively (F). www.nature.com/scientificreports/ colon cancer patients with varying degrees of differentiation have also been compared 27 . The results showed that poorly differentiated or undifferentiated carcinoma had a larger λ Hu than well-differentiated carcinoma, providing a feasible basis for the evaluation of the tumor pathology. Therefore, evaluation of the degree of differentiation of tumors after TACE should be further studied. Functional information regarding the microcirculation of the normal parenchyma and liver focal tumor lesions can be provided by CT perfusion imaging of the liver, which is a promising technology for the diagnosis   29 . Our study showed that NICAP and NICPP had high sensitivity and specificity in distinguishing TAA from TNA. NICAP had high sensitivity and specificity in distinguishing TAA from ANHP, which indicates that NICAP and NICPP could reflect the perfusion information of liver tissue and distinguish TAA from TNA and ANHP in the liver. The current standard for treatment response after TACE is based on morphology, which depends on the experience of the radiologist, DECT parameters could provide quantitative information as a supplement to mRECIST and LI-RADS. Especially for atypical tumors, the quantitative parameters of DECT may help physicians to better judge the effect of treatment and make correct clinical decisions. However quantitative analysis has not been widely promoted, mostly due to the lack of uniformity of equipment and parameters, and some studies are single-center studies, making it difficult to generalize the research results.
The application of DECT in abdominal imaging has been proved to be beneficial, and our research confirms its usefulness compared with compared to conventional CT exam performed at 120 kV, the radiation dose is not increased with DECT and may even be reduced, which benefits from optimization of solutions such as iterative reconstruction technology and the tube current modulation. Thus, the wide application of DECT will not result in an additional radiation dose burden to patients 30 .
In our study, lipiodol TACE and DEM-TACE treatments were included. There were no significant differences between the two methods in the parameters of this study. The relevant content was discussed in the supplementary materials.
Our study had some limitations. First, the number of patients was small, and this was a single-center study. If additional data can be obtained, large-scale multicenter studies may more realistically and accurately confirm the results. Second, the pathological results of some patients were not obtained, because patients treated with TACE rarely undergo needle biopsy or surgical resection.
In conclusion, DECT can effectively and quantitatively distinguish different HCC tissues after TACE and may be considered supplementary to LI-RADS v2018 for assessing HCC patients post TACE.