Liver Imaging Reporting and Data System Category 5: MRI Predictors of Microvascular Invasion and Recurrence After Hepatectomy for Hepatocellular Carcinoma
Abstract
OBJECTIVE. We investigated in Liver Imaging Reporting and Data System category 5 (LR-5) observations whether imaging features, including LI-RADS imaging features, could predict microvascular invasion (MVI) and posthepatectomy recurrence in high-risk adult patients with hepatocellular carcinoma (HCC).
MATERIALS AND METHODS. We retrospectively identified 149 high-risk patients who underwent 3-T MRI within 1 month before hepatectomy for HCC; 81 of 149 patients with no HCC recurrence were followed for more than 1 year. Tumors with clear surgical margins were confirmed in each hepatectomy specimen. MVI was evaluated histologically by a histopathologist. Tumor recurrence was determined by clinical and imaging follow-up. Two independent radiologists reviewed the prehepatectomy MR images and assessed LI-RADS v2018 imaging features as well as some non–LI-RADS features in all LR-5 observations in consensus. Alpha-fetoprotein level, tumor number, and imaging features were analyzed as potential predictors for MVI and posthepatectomy recurrence using multivariate logistic regression and Cox proportional hazards models.
RESULTS. One hundred forty-nine patients with pathologically confirmed HCC were included; 64 of 149 (43.0%) patients had MVI, whereas 48 of 129 (37.2%) patients had tumor recurrence within 3 years after hepatectomy. Mosaic architecture (odds ratio, 3.420; p < 0.001) and nonsmooth tumor margin (odds ratio, 2.554; p = 0.011) were independent predictors of MVI. Multifocal tumors (hazard ratio, 2.101; p = 0.034), absence of fat in mass (hazard ratio, 2.109; p = 0.015), and nonsmooth tumor margin (hazard ratio, 2.415; p = 0.005) were independent predictors of posthepatectomy recurrence.
CONCLUSION. In high-risk patients with LR-5 HCC, mosaic architecture and non-smooth tumor margin independently predicted MVI. Multifocal tumors, absence of fat in mass, and nonsmooth tumor margin independently predicted recurrence.
Hepatocellular carcinoma (HCC) is the fifth most common malignant tumor and the second most common cause of death from cancer worldwide [1]. Hepatic resection is the primary treatment modality for HCC in patients with well-preserved liver function [2, 3]. However, recurrence rates after resection can be as high as 70% within 5 years [4]. Tumor micro-vascular invasion (MVI) is a prognostic factor that predicts posthepatectomy HCC recurrence [5], but it cannot be determined until the tumor is analyzed histologically after its surgical removal. The preoperative capability to predict MVI and postsurgical recurrence would represent an advance by informing optimal selection of surgical candidates.
In recent studies, nonsmooth tumor margins, two-trait predictor for vascular invasion, peritumoral enhancement, and other imaging features have been reported as predictors in HCC for MVI or posthepatectomy recurrence HCC [5–7]. However, independent validation of these features has not yet been performed, and these features are not yet applied widely.
The Liver Imaging Reporting and Data System (LI-RADS) [8] was developed to standardized terminology, interpretation, and reporting of imaging for HCC diagnosis. The system addresses the full spectrum of liver lesions and pseudolesions with a 5-point scale reflecting the relative likelihood of HCC, from LR-1 (definitely benign) to LR-5 (definitely HCC). LI-RADS also assigns category LR-M to observations considered probably or definitely malignant but lacking criteria specific for HCC and a separate category (LR-TIV in versions 2017 and 2018) to those observations with definite tumor in vein [8]. Unlike other malignant solid tumors, for which diagnosis requires tissue sampling, HCC uniquely can be diagnosed noninvasively by imaging-based criteria without confirmatory biopsy [2, 3, 9]. In particular, although some tumors may mimic HCC on images, LR-5 conveys a high certainty of HCC, obviating routine biopsy confirmation.
Although LI-RADS was conceived as a diagnostic system, LI-RADS imaging features could possibly provide prognostic information, as supported by two recent studies showing that patients with tumors preoperatively categorized as LR-M have a worse prognosis after curative resection, even if the pathologic diagnosis is HCC [10, 11]. Although MRI features for the prediction of MVI and posthepatectomy recurrence have been reported previously, to our knowledge, no prior study has assessed standardized imaging features, such as those defined by LI-RADS, to predict MVI and recurrence. The purpose of our study was to investigate in LR-5 observations whether imaging features, including LI-RADS imaging features, in combination with serum α-fetoprotein (AFP) level, could predict MVI and posthepatectomy recurrence in high-risk adult patients with HCC.
Materials and Methods
Patients
This retrospective cohort study was approved by the institutional review board of the Third Affiliated Hospital of Sun Yat-sen University, with a waiver of the written informed consent requirement. From March 2014 to July 2017, 254 high-risk (patients with cirrhosis or hepatitis B virus infection or combined hepatitis B virus and hepatitis C virus infection) adult patients (age ≥ 18 years) who were suspected of having HCC, but who had no history of treatment, underwent hepatectomy in our hospital. All patients underwent 3-T MRI in our hospital before surgery. Patients were excluded if posthepatectomy pathologic analysis confirmed non-HCC malignancies such as cholangiocarcinoma (n = 18 patients), combined HCC and cholangiocarcinoma (n = 7), or metastasis from an extrahepatic primary tumor (n = 5); the time interval between preoperative MRI and hepatectomy exceeded 1 month (n = 1); patients had one or more LR-TIV observations (n = 17) or had no LR-5 observations (n = 26) or had observations categorized as LR-NC because of image omission or degradation according to LI-RADS v2018 (n = 2); or patients underwent contrast-enhanced MRI with gadoxetate disodium (n = 29). Patients were also excluded from the recurrence analysis if they were followed for less than 1 year after hepatectomy (n = 20). Patient selection is shown in Figure 1.
MRI Examination
Patients were scanned in the supine position on a 3-T whole-body MRI scanner (Discovery MR750, GE Healthcare) with an eight-channel phased-array coil centered over the abdomen. Unenhanced pulse sequences included breath-hold coronal balanced steady-state free precession, breath-hold coronal single-shot fast spin-echo, respiratory-triggered axial T2-weighted fast spin-echo, breath-hold 2D dual-echo T1-weighted gradient-recalled echo images at about 1.3 ms (opposed phase) and 2.6 ms (in phase), and respiratory-triggered axial DWI spin-echo echo-planar imaging with two b values (b = 0 and 800 s/mm2). Afterward, breath-hold 3D T1-weighted gradient-recalled echo imaging was performed before and at multiple time points dynamically after injection of extracellular contrast media (various formulations, 0.1 mmol/kg of gadolinium; n = 68 patients) or gadobenate dimeglumine (MultiHance, Bracco Diagnostics; 0.1 mmol/kg of gadolinium, n = 81 patients) IV by use of a power injector (Spectris Solaris EP, Medrad) at a flow rate of 2.0 mL/s (extracellular contrast media or gadobenate dimeglumine), followed by a 20-mL saline flush at a flow rate of 2.0 mL/s. A dual arterial phase sequence was initiated 15–20 seconds after the contrast media arrived at the distal thoracic aorta using bolus triggering. Dual portal venous phase and delayed phase images were acquired at 1 and 3 minutes, respectively. Optionally, 22 patients receiving gadobenate underwent hepatobiliary phase imaging at 90 minutes (14.8% of patients). The kind of contrast agent used was determined at the protocoling radiologist's discretion. Imaging examinations were in compliance with LI-RADS technical requirements. Scanning parameters are listed in Table S1, which can be viewed in the AJR electronic supplement to this article (available at www.ajronline.org).
Image Analysis
All MR images were retrieved from the PACS and reviewed by two abdominal radiologists (with 6 and 25 years of experience in liver MRI) using LI-RADS v2018. The observations were categorized as LR-5 by the two radiologists in consensus. The location of each observation was recorded according to the Couinaud classification. The largest LR-5 observation was evaluated in patients with multiple observations. Reviewers independently evaluated the following imaging features for each selected observation as defined in LI-RADS v2018 [8]: nonrim arterial phase hyperenhancement, lesion size (≥ 20 vs < 20 mm), nonperipheral washout, enhancing capsule, nonenhancing capsule, nodule-in-nodule architecture, mosaic architecture, blood products in mass, fat in mass, restricted diffusion, mild-to-moderate T2 hyperintensity, coronal enhancement, fat sparing in solid mass, and iron sparing in solid mass. Because only 22 of 149 patients (14.8%) had hepatobiliary phase images, one imaging feature assessed only on hepatobiliary phase images (hepatobiliary phase hypointensity) was not analyzed. The number of tumors was recorded. Tumor stage was scored according to Organ Procurement and Transplantation Network policy [12]. Disagreements were adjudicated by consensus.
Because it has been reported that nonsmooth tumor margin, two-trait predictor for vascular invasion, and peritumoral enhancement [7] can accurately predict MVI in HCC, we also evaluated these three non–LI-RADS features. As proposed in the study by Renzulli et al. [7], nonsmooth tumor margin was defined as nonnodular tumor in any imaging plane. Two-trait predictor for vascular invasion was defined as the presence of internal arteries visible in the arterial phase and the absence of hypointense halo in a postarterial phase. Peritumoral enhancement was defined as detectable enhancement in the arterial phase adjacent to the tumor border, later becoming isointense on MR images compared with the background liver parenchyma in the delayed phase.
Histopathologic Diagnosis
Histologic specimens were obtained from surgical resection in all patients. An experienced histopathologist (with 11 years of experience) blinded to all clinical data and MRI results reviewed the H and E–stained slides and, in every case, confirmed the histologic diagnosis and also assessed capsule formation, histologic grade, vascular invasion, and clean surgical margins according to the World Health Organization classification system [13]. Tumor grade was classified as well, moderately, or poorly differentiated. When different tumor grades coexisted, the predominant grade was assigned. The presence of tumor capsule (surrounding at least two-thirds of the tumor margin regardless of the presence of microscopic capsular or extracapsular invasion) [14], MVI (presence of tumor within a vascular space lined by endothelium and visible only on microscopy), and macrovascular invasion (presence of tumor within a vein seen on gross examination) were reported. One author reviewed each pathology report and recorded the largest tumor for its grade, as well as the presence of capsule, MVI, and macrovascular invasion. The same author also documented whether there was a single tumor (a single nodule was described in the pathology report or additional nodules adjacent to the tumor were described as satellite lesions) or multiple tumors (if two or more tumors were reported separately, each with a full description of its histopathologic features such as histologic grade, architecture, and cell type) [15].
Baseline Clinical Data
One author reviewed the electronic medical record for each patient and extracted demographics, serum AFP levels, serum hepatitis B virus DNA levels, and Child-Pugh class. Serum AFP was considered elevated if it was greater than 400 ng/mL [16].
Follow-Up Surveillance After Surgical Resection
Postoperative follow-up included clinical examination, chest radiography, biochemical liver function tests, and serum levels of AFP performed at 1 month after hepatic resection and then every 2–3 months. In addition, contrast-enhanced ultrasound, multiphasic abdominal CT, or multiphasic abdominal MRI was performed every 3 months. Whole-body PET scanning was performed when patients had increasing AFP level with uncertain findings on the other imaging modalities.
Recurrence
The presence or absence of recurrence was evaluated by another two authors in consensus, while blinded to the MRI feature analysis results. Recurrence was defined as a new lesion arising in the remnant liver after hepatectomy, including early recurrent tumors (< 2 years) most likely originating from subclinical metastasis of primary tumors, and late recurrent tumors (≤ 2 years) that might reflect multi-centric or de novo primary HCC in the remnant liver [15, 17]. Recurrence was diagnosed by dynamic CT or MRI according to typical imaging characteristics (hyperenhancement on arterial phase images and washout on portal venous or delayed phase images). Recurrence was confirmed by biopsy or surgical pathology after reresection. Otherwise, patients followed for at least 1 year were considered to be recurrence free. Patients who were recurrence free and followed less than 1 year after surgery were excluded from the recurrence analysis. Increasing AFP levels alone without imaging evidence of a new lesion were not interpreted as HCC recurrence until imaging studies became positive for recurrence, as described already [5]. Each patient's recurrence-free survival was defined as the interval from the date of surgery to that of last follow-up evaluation in patients without confirmed recurrence, or to the first imaging follow-up examination that confirmed recurrence in patients with recurrence.
Statistical Analysis
Data were summarized descriptively. Continuous variables were compared using a two-sample t test or Mann-Whitney U test. Categoric variables were compared using the chi-square test, Fisher exact test, or Kruskal-Wallis test.
Interobserver agreement was assessed by the Cohen kappa statistic. Agreement was considered excellent if kappa was more than 0.80, good if kappa ranged from 0.61 to 0.80, moderate if kappa ranged from 0.41 to 0.60, and poor if it was 0.40 or less.
Logistic regression analysis was performed to assess potential predictors for MVI. Variables with p < 0.1 in univariate logistic regression analysis were applied to a multivariate logistic regression analysis.
The Cox proportional hazards model was used for univariate analysis of potential predictors for recurrence. Variables with p < 0.1 from univariate analysis were included for multivariate analysis using a stepwise Cox hazards regression model.
Accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were calculated for the combination of predictors identified by the multivariate models for their respective outcomes (MVI or recurrence). SPSS software (version 22.0, IBM) was used for all statistical analyses. A p < 0.05 was considered statistically significant.
Results
Patient Baseline Characteristics
Demographics and baseline clinical and biologic characteristics of the MVI and recurrence analysis cohorts are summarized in Table 1 and Table 2. A total of 149 patients were included in the MVI analysis cohort, and 129 patients were included in the recurrence analysis cohort.
| Variable | Total (n = 149) | MVI | ||
|---|---|---|---|---|
| MVI Absence (n = 85) | MVI Presence (n = 64) | p | ||
| Age (y)a | 51.60 ± 11.74 | 52.01 ± 11.89 | 51.05 ± 11.62 | 0.621 |
| Sex | 0.807 | |||
| Female | 13 (8.7) | 7 (8.2) | 6 (9.4) | |
| Male | 136 (91.3) | 78 (91.8) | 58 (90.6) | |
| AFP level (ng/mL)a | 42.70 (0.61–121,000) | 20.30 (0.61–121,000) | 151.85 (0.61–121,000) | 0.214 |
| Serum hepatitis B virus DNA level (IU/mL)a | 2720 (0–1300 × 105) | 3650 (0–1300 × 105) | 2250 (0–109 × 105) | 0.282 |
| Total bilirubin level (mg/dL)a | 0.90 ± 0.69 | 0.87 ± 0.41 | 0.96 ± 0.93 | 0.418 |
| Serum creatinine level (mg/dL)a | 0.90 ± 0.18 | 0.93 ± 0.20 | 0.88 ± 0.14 | 0.078 |
| International normalized ratioa | 1.04 ± 0.11 | 1.03 ± 0.07 | 1.05 ± 0.14 | 0.368 |
| Platelet count (x103/pL)a | 180.66 ± 74.53 | 179.64 ± 74.25 | 182.03 ± 75.46 | 0.847 |
| ALT-to-AST level ratioa | 1.05 ± 0.40 | 1.09 ± 0.38 | 0.99 ± 0.41 | 0.121 |
| Neutrophil-to-lymphocyte ratioa | 1.82 (0.73–28.99) | 1.79 (0.73–7.90) | 1.86 (0.73–28.99) | 0.096 |
| Model for End-Stage Liver Disease scorea | 4.71 ± 2.77 | 4.77 ± 2.95 | 4.63 ± 2.53 | 0.765 |
| Tumor size (mm)a | 46.20 ± 25.57 | 36.38 ± 17.17 | 59.25 ± 28.98 | < 0.001 |
| Child-Pugh class | 0.256 | |||
| A | 140 (94.0) | 82 (96.5) | 58 (90.6) | |
| B | 9 (6.0) | 3 (3.5) | 6 (9.4) | |
| Tumor category | < 0.001 | |||
| T1–T2 | 97 (65.1) | 66 (77.6) | 31 (48.4) | |
| T3–T4 | 52 (34.9) | 19 (22.4) | 33 (51.6) | |
| No. of tumors | 0.756 | |||
| Single | 125 (83.9) | 72 (84.7) | 53 (82.8) | |
| Multiple | 24 (16.1) | 13 (15.3) | 11 (17.2) | |
| Histologic features | ||||
| Tumor grade | 0.019 | |||
| Well or moderately differentiated | 126 (84.6) | 77 (90.6) | 49 (76.6) | |
| Poorly differentiated | 23 (15.4) | 8 (9.4) | 15 (23.4) | |
| Histologic capsule | 0.468 | |||
| Absent | 37 (24.8) | 23 (27.1) | 14 (21.9) | |
| Present | 112 (75.2) | 62 (72.9) | 50 (78.1) | |
| Cirrhosis | 0.127 | |||
| Absent | 92 (61.7) | 48 (56.5) | 44 (68.8) | |
| Present | 57 (38.3) | 37 (43.5) | 20 (31.2) | |
Note—Except where indicated otherwise, data are number (%) of patients. Categoric variables were compared by using the chi-square test or Fisher exact test. Hepatocellular carcinoma tumor category was scored according to Organ Procurement and Transplantation Network policy [12]. AFP = α-fetoprotein, ALT = alanine transaminase, AST = aspartate transaminase.
a
Data are continuous variables, reported as mean ± SD or median (range), and were compared using the two-sample t test or nonparametric Mann-Whitney U test.
| Variable | Total (n = 129) | Recurrence | ||
|---|---|---|---|---|
| Nonrecurrence (n = 81) | Recurrence (n = 48) | p | ||
| Age (y)a | 51.08 ± 11.65 | 52.22 ± 10.54 | 49.14 ± 13.2 | 0.173 |
| Sex | 0.688 | |||
| Female | 13 (10.1) | 7 (8.6) | 6 (12.5) | |
| Male | 116 (89.9) | 74 (91.4) | 42 (87.5) | |
| AFP level (ng/mL)a | 54.60 (0.61–121,000) | 38.80 (0.61–3900.67) | 163.85 (1–121,000) | 0.320 |
| Serum hepatitis B virus DNA level (IU/mL)a | 2470 (0–1300 × 105) | 3650 (0–1300 × 105) | 1995 (0–143 × 105) | 0.394 |
| Total bilirubin level (mg/dL)a | 0.93 ± 0.72 | 0.99 ± 0.87 | 0.82 ± 0.36 | 0.199 |
| Serum creatinine level (mg/dL)a | 0.90 ± 0.18 | 0.91 ± 0.18 | 0.89 ± 0.19 | 0.590 |
| International normalized ratio)a | 1.04 ± 0.11 | 1.04 ± 0.08 | 1.04 ± 0.16 | 0.674 |
| Platelet count (× 103/pL)a | 182.66 ± 74.08 | 185.37 ± 75.20 | 178.08 ± 72.70 | 0.591 |
| A LT-to-AST level ratioa | 1.04 ± 0.38 | 1.07 ± 0.39 | 0.99 ± 0.38 | 0.255 |
| Neutrophil-to-lymphocyte ratioa | 2.21 ± 1.81 | 2.24 ± 2.06 | 2.16 ± 1.31 | 0.814 |
| Model for End-Stage Liver Disease scorea | 4.79 ± 2.84 | 5.03 ± 2.60 | 4.39 ± 3.18 | 0.220 |
| Tumor size (mm)a | 45.67 ± 25.95 | 43.44 ± 24.28 | 49.42 ± 28.41 | 0.227 |
| Child-Pugh class | 0.693 | |||
| A | 121 (93.8) | 77 (95.1) | 44 (91.7) | |
| B | 8 (6.2) | 4 (4.9) | 4 (8.3) | |
| Tumor category | 0.122 | |||
| T1–T2 | 86 (66.7) | 58 (71.6) | 28 (58.3) | |
| T3–T4 | 43 (33.3) | 23 (28.4) | 20 (41.7) | |
| No. of tumors | 0.116 | |||
| Single | 108 (83.7) | 71 (87.7) | 37 (77.1) | |
| Multiple | 21 (16.3) | 10 (12.3) | 11 (22.9) | |
| Histologic features | ||||
| Tumor grade | 0.927 | |||
| Well or moderately differentiated | 108 (83.7) | 68 (84.0) | 40 (83.3) | |
| Poorly differentiated | 21 (16.3) | 13 (16.0) | 8 (16.7) | |
| Histologic capsule | 0.377 | |||
| Absent | 32 (24.8) | 18 (22.2) | 14 (29.2) | |
| Present | 97 (75.2) | 63 (77.8) | 34 (70.8) | |
| Cirrhosis | 0.507 | |||
| Absent | 80 (62.0) | 52 (64.2) | 28 (58.3) | |
| Present | 49 (38.0) | 29 (35.8) | 20 (41.7) | |
Note—Except where indicated otherwise, data are number (%) of patients. Categoric variables were compared by using the chi-square test or Fisher exact test. Hepatocellular carcinoma tumor category was scored according to Organ Procurement and Transplantation Network policy [12]. AFP = α-fetoprotein, ALT = alanine transaminase, AST = aspartate transaminase.
a
Data are continuous variables, reported as mean ± SD or median (range), and were compared using the two-sample t test or nonparametric Mann-Whitney U test.
In the MVI analysis cohort, the groups with MVI present (n = 64) and MVI absent (n = 85) differed significantly in terms of tumor size, stage, and grade. In the recurrence analysis cohort, there was no significant different between the recurrence (n = 48) and nonrecurrence (n = 81) groups in terms of baseline clinical and biologic characteristics.
Pathologic Analysis
Of the 149 included patients, 125 (83.9%) had a single tumor, and the other 24 (16.1%) had multifocal tumors. In patients with multifocal tumors, only the largest ones were analyzed. Mean tumor size was 46.2 mm (range, 10–133 mm). The tumors were well differentiated, moderately differentiated, and poorly differentiated in 26 (17.5%), 100 (67.1%), and 23 (15.4%) cases, respectively. MVI was found in 64 (43.0%) tumors. No tumor had macrovascular invasion at pathologic examination. Histologic capsule was identified in 112 (75.2%) tumors.
Follow-Up and Recurrence
Of the 129 patients included in the recurrence analysis cohort, all surviving patients were followed until recurrence or, if recurrence free, for at least 1 year. Overall, the median follow-up period was 11 months (range, 1–36 months). Forty-eight (37.2%) of the 129 patients had tumor recurrence within 3 years after hepatectomy. The mean (± SD) time to recurrence was 296 ± 253 days (range, 36–898 days).
Interobserver Agreement
Interobserver agreement of all the assessed MRI features is shown in Table S2, which can be viewed in the AJR electronic supplement to this article (available at www.ajronline.org). Agreement between the two observers was good to excellent, with kappa values of 0.7–1.0 for all MRI features and for the LR-5 category.
Univariate and Multivariate Logistic Analyses for Microvascular Invasion Prediction
In univariate analysis, MVI was significantly associated with eight predictors: elevated AFP level (p = 0.025), tumor size greater than or equal to 20 mm (p = 0.060), mosaic architecture (p < 0.001), blood products (p = 0.008), coronal enhancement (p = 0.066), nonsmooth tumor margin (p < 0.001), two-trait predictor for vascular invasion (p = 0.002), and peritumoral enhancement (p = 0.066), but not with multifocal tumors or other LI-RADS features. In the multivariate analysis, the association remained significant for mosaic architecture (odds ratio, 3.420; 95% CI, 1.521–7.688; p < 0.001) and nonsmooth tumor margin (odds ratio, 2.554; 95% CI, 1.238–5.272; p = 0.011) but not for the other candidate predictors (Table 3 and Fig. 2).
| Variable | Univariate Analysis | Multivariate Analysis | |||||
|---|---|---|---|---|---|---|---|
| p Coefficient | SE | p | p Coefficient | SE | p | OR (95% CI) | |
| Patient characteristic | |||||||
| AFP level > 400 ng/mL | 0.799 | 0.357 | 0.025 | 0.499 | 0.403 | 0.216 | 1.647 (0.748–3.629) |
| Multifocal tumors | 0.139 | 0.448 | 0.756 | ||||
| LI-RADS major features | |||||||
| Lesion ≥ 20 mm | 2.010 | 1.068 | 0.060 | 0.800 | 1.118 | 0.474 | 2.226 (0.249–19.898) |
| Nonrim arterial phase hyperenhancementa | |||||||
| Nonperipheral washout | 0.444 | 0.636 | 0.485 | ||||
| Enhancing capsule | 0.126 | 0.928 | 0.892 | ||||
| LI-RADS ancillary features (favoring HCC in particular) | |||||||
| Nonenhancing capsulea | |||||||
| Nodule-in-nodule | −0.297 | 0.650 | 0.647 | ||||
| Mosaic architecture | 1.502 | 0.396 | < 0.001 | 1.230 | 0.413 | < 0.001 | 3.420 (1.521–7.688) |
| Blood products | 1.065 | 0.400 | 0.008 | 0.516 | 0.462 | 0.264 | 1.676 (0.678–4.144) |
| Fat in mass (absence) | −0.024 | 0.331 | 0.943 | ||||
| LI-RADS ancillary features (favoring malignancy in general) | |||||||
| Restricted diffusion | 20.931 | 40,193.049 | 1.000 | ||||
| Mild-to-moderate T2 hyperintensity | −21.518 | 28,420.777 | 0.999 | ||||
| Coronal enhancement | 0.668 | 0.363 | 0.066 | 0.553 | 0.405 | 0.172 | 1.739 (0.786–3.848) |
| Fat sparing in solid mass | −0.661 | 0.854 | 0.438 | ||||
| Iron sparing in solid mass Non-LI-RADS imaging features | 0.134 | 0.333 | 0.688 | ||||
| Nonsmooth tumor margin | 1.243 | 0.350 | < 0.001 | 0.938 | 0.370 | 0.011 | 2.554 (1.238–5.272) |
| Two-trait predictor | 1.058 | 0.348 | 0.002 | 0.455 | 0.405 | 0.261 | 1.576 (0.713–3.485) |
| Peritumoral enhancement | 0.668 | 0.363 | 0.066 | 0.553 | 0.405 | 0.172 | 1.739 (0.786–3.848) |
Note—Variables with p < 0.1 in univariate analysis were applied to a multivariate analysis. SE = standard error, OR = odds ratio, AFP = α-fetoprotein, LI-RADS = Liver Imaging Reporting and Data System, HCC = hepatocellular carcinoma.
a
All observations (149/149) had the feature of nonrim arterial phase hyperenhancement and none of the observations had the feature of nonenhancing capsule.
Fig. 2A —60-year-old man with surgically confirmed poorly differentiated hepatocellular carcinoma and α-fetoprotein level of 482.2 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis confirmed microvascular invasion. Tumor recurred 1047 days after hepatectomy.
A, MR image shows hepatic mass (49 × 35 mm) in segments VI and VII with mild-to-moderate T2 hyperintensity.
Fig. 2B —60-year-old man with surgically confirmed poorly differentiated hepatocellular carcinoma and α-fetoprotein level of 482.2 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis confirmed microvascular invasion. Tumor recurred 1047 days after hepatectomy.
B, Mass is hypointense on unenhanced T1-weighted MR image.
Fig. 2C —60-year-old man with surgically confirmed poorly differentiated hepatocellular carcinoma and α-fetoprotein level of 482.2 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis confirmed microvascular invasion. Tumor recurred 1047 days after hepatectomy.
C, Mass shows hyperenhancement, coronal enhancement (arrow), and mosaic architecture on late arterial phase MR image.
Fig. 2D —60-year-old man with surgically confirmed poorly differentiated hepatocellular carcinoma and α-fetoprotein level of 482.2 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis confirmed microvascular invasion. Tumor recurred 1047 days after hepatectomy.
D, Mass shows nonperipheral washout appearance, enhancing capsule appearance (arrowhead), and nonsmooth tumor margin on portal venous phase MR image.
Univariate and Multivariate Cox Analyses for Recurrence Prediction
Posthepatectomy recurrence was associated with five predictors: elevated AFP level (p = 0.084), multifocal tumors (p = 0.074), mosaic architecture (p = 0.073), absence of fat in mass (p = 0.025), and nonsmooth tumor margin (p = 0.014), but not other LI-RADS imaging features. In the multivariate analysis, only three predictors remained significant: multifocal tumors (hazard ratio, 2.101; 95% CI, 1.058–4.175; p = 0.034), absence of fat in mass (hazard ratio, 2.109; 95% CI, 1.155–3.851; p = 0.015), and nonsmooth tumor margin (hazard ratio, 2.415; 95% CI, 1.310–4.451; p = 0.005) (Table 4 and Fig. 3).
| Variable | Univariate Analysis | Multivariate Analysis | ||||||
|---|---|---|---|---|---|---|---|---|
| p Coefficient | SE | p | HR | β Coefficient | SE | p | HR (95% CI) | |
| Patient characteristic | ||||||||
| AFP level > 400 ng/mL | 0.504 | 0.292 | 0.084 | 1.655 | 0.299 | 0.307 | 0.329 | 1.348 (0.739–2.459) |
| Multifocal tumors | 0.616 | 0.345 | 0.074 | 1.852 | 0.743 | 0.350 | 0.034 | 2.101 (1.058–4.175) |
| LI-RADS major features | ||||||||
| Lesion ≥ 20 mm | −0.460 | 0.476 | 0.333 | 0.631 | ||||
| Nonrim arterial phase hyperenhancementa | ||||||||
| Nonperipheral washout | 0.653 | 0.723 | 0.367 | 1.921 | ||||
| Enhancing capsule LI-RADS ancillary features (favoring HCC in particular) | 0.346 | 0.730 | 0.635 | 1.414 | ||||
| LI-RADS ancillary features (favoring HCC in particular) | ||||||||
| Nonenhancing capsulea | ||||||||
| Nodule-in-nodule | −0.280 | 0.597 | 0.639 | 0.756 | ||||
| Mosaic architecture | 0.579 | 0.323 | 0.073 | 1.783 | 0.398 | 0.346 | 0.250 | 1.489 (0.756–2.936) |
| Blood products | 0.167 | 0.347 | 0.629 | 1.182 | ||||
| Fat in mass (absence) | 0.685 | 0.306 | 0.025 | 1.984 | 0.746 | 0.307 | 0.015 | 2.109 (1.155–3.851) |
| LI-RADS ancillary features (favoring malignancy in general) | ||||||||
| Restricted diffusion | 3.061 | 4.851 | 0.528 | 21.351 | ||||
| Mild-to-moderate T2 hyperintensity | 3.037 | 4.950 | 0.540 | 20.839 | ||||
| Coronal enhancement | 0.094 | 0.312 | 0.762 | 1.099 | ||||
| Fat sparing in solid mass | −0.302 | 0.723 | 0.676 | 0.739 | ||||
| Iron sparing in solid mass | 0.137 | 0.290 | 0.636 | 1.147 | ||||
| Non-LI-RADS imaging features | ||||||||
| Nonsmooth tumor margin | 0.756 | 0.309 | 0.014 | 2.129 | 0.882 | 0.312 | 0.005 | 2.415 (1.310–4.451) |
| Two-trait predictor | 0.396 | 0.294 | 0.178 | 1.485 | ||||
| Peritumoral enhancement | 0.094 | 0.312 | 0.762 | 1.099 | ||||
Note—Variables with p < 0.1 in univariate analysis were applied to a multivariate analysis. SE = standard error, HR = hazard ratio, AFP = α-fetoprotein, LI-RADS = Liver Imaging Reporting and Data System, HCC = hepatocellular carcinoma.
a
All observations (129/129) had the feature of nonrim arterial phase hyperenhancement and none of the observations had the feature of nonenhancing capsule.
Fig. 3A —32-year-old man with surgically confirmed moderately differentiated hepatocellular carcinoma and α-fetoprotein level of 9.78 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis showed no evidence of microvascular invasion. There was no recurrence during follow-up (23 months) after hepatectomy.
A, MRI shows hepatic mass (36 × 20 mm) in segments V and VI of liver parenchyma. In-phase (A) and outof-phase (B) T1-weighted images reveal fat in mass (note signal loss on out-of-phase image, B).
Fig. 3B —32-year-old man with surgically confirmed moderately differentiated hepatocellular carcinoma and α-fetoprotein level of 9.78 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis showed no evidence of microvascular invasion. There was no recurrence during follow-up (23 months) after hepatectomy.
B, MRI shows hepatic mass (36 × 20 mm) in segments V and VI of liver parenchyma. In-phase (A) and outof-phase (B) T1-weighted images reveal fat in mass (note signal loss on out-of-phase image, B).
Fig. 3C —32-year-old man with surgically confirmed moderately differentiated hepatocellular carcinoma and α-fetoprotein level of 9.78 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis showed no evidence of microvascular invasion. There was no recurrence during follow-up (23 months) after hepatectomy.
C, Contrast-enhanced MRI shows hyperenhancement in late arterial phase (C); delayed phase image (D) shows nonperipheral washout appearance, enhancing capsule appearance (arrow), and nonsmooth tumor margin.
Fig. 3D —32-year-old man with surgically confirmed moderately differentiated hepatocellular carcinoma and α-fetoprotein level of 9.78 ng/mL. Two-trait predictor for vascular invasion was absent in this tumor. Histologic analysis showed no evidence of microvascular invasion. There was no recurrence during follow-up (23 months) after hepatectomy.
D, Contrast-enhanced MRI shows hyperenhancement in late arterial phase (C); delayed phase image (D) shows nonperipheral washout appearance, enhancing capsule appearance (arrow), and nonsmooth tumor margin.
Diagnostic Performance of Exploratory Prediction Models
The diagnostic performances of the MVI and recurrence prediction models described already (combining all significant variables identified in the multivariate analysis) are summarized in Table 5. Accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of the MVI prediction model were 67.8%, 60.9%, 72.9%, 62.9%, and 71.3%, respectively. Accuracy, sensitivity, specificity, positive predictive value, and negative predictive value for recurrence prediction model were 62.8%, 8.0%, 97.5%, 66.7%, and 62.6%, respectively.
| Outcome (Selected Predictors) | Sensitivity (%) | Specificity (%) | Accuracy (%) | Positive Predictive Value (%) | Negative Predictive Value (%) | ROC AUC |
|---|---|---|---|---|---|---|
| Microvascular invasion (mosaic architecture and nonsmooth tumor margin) Value | 60.9 | 72.9 | 67.8 | 62.9 | 71.3 | 0.669 |
| No. of tumors/total | 39/64 | 62/85 | 101/149 | 39/62 | 62/87 | |
| 95% CI | 47.9–72.9 | 62.2–82.0 | 59.9–74.8 | 50.5–73.8 | 61.0–79.7 | 0.588–0.744 |
| Postoperative recurrence (multifocal HCC, absence of fat in mass, and nonsmooth tumor margin) Value | 8.0 | 97.5 | 62.8 | 66.7 | 62.6 | 0.527 |
| No. of tumors/total | 4/50 | 77/79 | 81/129 | 4/6 | 77/123 | |
| 95% CI | 2.2–19.2 | 91.2–99.7 | 54.2–70.6 | 30.0–94.1 | 53.8–70.6 | 0.438–0.616 |
Note—Sensitivity, specificity, accuracy, positive predictive value, negative predictive value, and ROC AUC of the prediction models were calculated by combination of all the selected predictors. HCC = hepatocellular carcinoma.
Discussion
In a single-center retrospective study, we found that one LI-RADS imaging feature (mosaic architecture) and one non–LI-RADS imaging feature (nonsmooth tumor margin) were significant predictors of MVI in patients with LR-5 HCC. In addition, we found that multifocal tumors and one LI-RADS imaging feature (absence of fat in mass) and one non–LI-RADS feature (non-smooth tumor margin) were significant predictors of recurrence in patients with LR-5 HCC. Thus, neither elevated AFP level nor any major LI-RADS feature was a significant independent predictor of either outcome.
Mosaic architecture is a LI-RADS ancillary feature favoring HCC in particular. It refers to the presence within a mass of randomly distributed internal nodules or compartments differing in enhancement, attenuation, intensity, shape, and size and often separated by fibrous separations [8]. Previously, Li et al. [18] found that mosaic architecture and coronal enhancement were correlated with reduced overall survival or earlier time to progression in patients undergoing liver resection. However, to our knowledge, prior studies have not assessed the correlation between mosaic architecture and MVI in HCC. Our results showed that mosaic architecture was a significant and independent factor for MVI but not recurrence. Mosaic architecture is a marker of tumor heterogeneity [19]. Differing in histologic [20] and molecular [21] features, the inner nodules may vary in biologic behavior, with some having greater propensity for vascular invasion, potentially explaining the association with MVI found in our study.
Intranodular steatosis is a characteristic histologic feature of early HCC [22], and fat in mass is applied by LI-RADS as an ancillary feature favoring but not establishing the diagnosis of HCC. Previous studies found that the presence of fat in mass at imaging predicts well-differentiated HCC tumor grade [23, 24] and that this imaging feature is unusual in poorly differentiated HCC [25]. Moreover, Siripongsakun et al. [26] reported that, compared with non–fat-containing HCCs, fat-containing HCCs have a more favorable prognosis. Our study helps to corroborate the prior results by showing that the absence of fat in mass was a significant and independent predictor of recurrence. Thus, fat in mass may provide two complementary functions: as a diagnostic feature in indeterminate lesions, it helps support the diagnosis of HCC over other differential considerations, and as a prognostic feature in lesions meeting LR-5 criteria (i.e., in lesions that can be diagnosed as HCC on the basis of other imaging characteristics), its absence suggests a worse outcome.
Tumor number remains one of the best, and most easily assessable, preoperative prognostic factors for postsurgical outcome [27]. Studies have shown that tumor recurrence is more frequent and occurs earlier in patients with multiple tumors than in those with single tumors [2, 28, 29], which was consistent with our study.
Nonsmooth tumor margin refers to a non-nodular border in any imaging plane. Lee et al. [5] found that nonsmooth tumor margin had 69% (136/197) accuracy in predicting MVI. Ariizumi et al. [30] found that nonsmooth tumor margin had 69.5% (41/59) accuracy in predicting recurrence. Our study helped confirm these prior results by showing that nonsmooth tumor margin was an independent predictor of both MVI and recurrence. Unlike prior studies [7], we did not find that two-trait predictor for vascular invasion and peritumoral enhancement were independent predictors of MVI, although they were significant predictors of MVI in univariate analysis.
As a validated biomarker of HCC, serum AFP level has been correlated with MVI, differentiation, and postsurgical recurrence [16, 29, 31–34]. Zhao et al. [16] found that serum AFP level greater than 400 ng/mL was a pre-operative predictor of MVI in patients with multifocal HCCs. Shin et al. [31] and Cucchetti et al. [29] found that tumor recurrence occurred more frequently in patients with high AFP levels. Although our univariate results were consistent with those of previous studies, AFP level was not an independent predictor for MVI or recurrence in multivariate models that incorporated MRI features.
Although imaging features were significant predictors of MVI and recurrence, the diagnostic performances of the models for their respective outcomes were only fair, with accuracies of less than 70%. This degree of accuracy does not suffice to reliably guide patient management. Thus, further research is needed to develop and validate prediction models capable of informing individualized patient management.
There were some limitations to our study. The retrospective design, together with the selection of surgical candidates, may result in an incomplete representation of all malignancies and radiologic features. We did not include patients with non-HCC LR-5 observations, which may cause selection bias, but non-HCC LR-5 observations are infrequent because LR-5 provides a high specificity for the diagnosis of HCC [35]. Our study represented a single-center experience using only 3-T scanners, and confirmation is needed in a prospective multicenter setting using a variety of scanners and field strengths. The presence of MVI was assessed by only one experienced histopathologist, and the interobserver variability of the assessment of MVI was not assessed in our study. Some LI-RADS ancillary features evaluable only in the hepatobiliary phase were not analyzed. Threshold growth, one of the major LI-RADS features, was not recorded because we focused on cross-sectional imaging features that did not depend on prior examinations. In patients with multiple tumors, we evaluated only the largest lesion, which may not represent the imaging features of smaller but potentially more aggressive tumors. Our MRI protocol used a dual arterial phase (early and late arterial phase) acquisition; although this may improve optimal arterial phase capture, dual arterial phase is not required by LI-RADS, and some centers may lack the necessary technical capability, potentially limiting study generalizability. We used different contrast agents with different T1 relaxivities and degrees of hepatocellular uptake, which probably introduced some variability in the degree of enhancement of tumor relative to background liver in arterial and postarterial phases. However, LI-RADS does not recommend any particular contrast agent, leaving the choice to radiologists and institutions, and the use of different contrast agents in our study may better represent a typical spectrum of clinical cases. Finally, in this study, recurrent HCC was not subclassified as early or late recurrence. Further research is needed to investigate the relation between LI-RADS and other imaging features with early versus late recurrence.
Conclusion
Our study suggests that, in patients with LR-5 HCC, in addition to nonsmooth tumor margin, one LI-RADS MRI feature, mosaic architecture, was an independent predictor of MVI, and another such feature, absence of fat in mass, and multifocal tumors were independent predictors of recurrence. If these findings are validated by future studies, these LI-RADS MRI features could be used in combination with nonsmooth tumor margin to predict MVI and posthepatectomy recurrence. For this information to guide surgical decision making, however, the performance of the prediction models will need improvement.
Supplemental Content
References
1.
Bertuccio P, Turati F, Carioli G, et al. Global trends and predictions in hepatocellular carcinoma mortality. J Hepatol 2017; 67:302–309
2.
Bruix J, Sherman M; American Association for the Study of Liver Diseases. Management of hepatocellular carcinoma: an update. Hepatology 2011; 53:1020–1022
3.
European Association for the Study of the Liver; European Organisation for Research and Treatment of Cancer. EASL-EORTC clinical practice guidelines: management of hepatocellular carcinoma. J Hepatol 2012; 56:908–943
4.
Bruix J, Gores GJ, Mazzaferro V. Hepatocellular carcinoma: clinical frontiers and perspectives. Gut 2014; 63:844–855
5.
Lee S, Kim SH, Lee JE, Sinn DH, Park CK. Preoperative gadoxetic acid-enhanced MRI for predicting microvascular invasion in patients with single hepatocellular carcinoma. J Hepatol 2017; 67:526–534
6.
Zhao J, Li X, Zhang K, et al. Prediction of micro-vascular invasion of hepatocellular carcinoma with preoperative diffusion-weighted imaging: a comparison of mean and minimum apparent diffusion coefficient values. Medicine (Baltimore) 2017; 96:e7754
7.
Renzulli M, Brocchi S, Cucchetti A, et al. Can current preoperative imaging be used to detect microvascular invasion of hepatocellular carcinoma? Radiology 2016; 279:432–442
8.
American College of Radiology (ACR). Liver Reporting & Data System (LI-RADS). ACR website. www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/LI-RADS. Published 2018. Accessed April 17, 2019
9.
Tang A, Cruite I, Mitchell DG, Sirlin CB. Hepatocellular carcinoma imaging systems: why they exist, how they have evolved, and how they differ. Abdom Radiol (NY) 2018; 43:3–12
10.
An C, Park S, Chung YE, et al. Curative resection of single primary hepatic malignancy: Liver Imaging Reporting and Data System category LR-M portends a worse prognosis. AJR 2017; 209:576–583
11.
Choi SH, Lee SS, Park SH, et al. LI-RADS classification and prognosis of primary liver cancers at gadoxetic acid-enhanced MRI. Radiology 2019; 290:388–397
12.
U.S. Department of Health and Human Services (DHHS). Organ procurement and transplantation network. DHHS website. optn.transplant.hrsa.gov. Accessed April 17, 2019
13.
Kleihues P, Sobin LH. World Health Organization classification of tumors. Cancer 2000; 88:2887
14.
Ishigami K, Yoshimitsu K, Nishihara Y, et al. Hepatocellular carcinoma with a pseudocapsule on gadolinium-enhanced MR images: correlation with histopathologic findings. Radiology 2009; 250:435–443
15.
An C, Kim DW, Park YN, Chung YE, Rhee H, Kim MJ. Single hepatocellular carcinoma: preoperative MR imaging to predict early recurrence after curative resection. Radiology 2015; 276:433–443
16.
Zhao WC, Fan LF, Yang N, Zhang HB, Chen BD, Yang GS. Preoperative predictors of microvascular invasion in multinodular hepatocellular carcinoma. Eur J Surg Oncol 2013; 39:858–864
17.
Imamura H, Matsuyama Y, Tanaka E, et al. Risk factors contributing to early and late phase intrahepatic recurrence of hepatocellular carcinoma after hepatectomy. J Hepatol 2003; 38:200–207
18.
Li M, Xin Y, Fu S, et al. Corona enhancement and mosaic architecture for prognosis and selection between of liver resection versus transcatheter arterial chemoembolization in single hepatocellular carcinomas >5 cm without extrahepatic metastases: an imaging-based retrospective study. Medicine (Baltimore) 2016; 95:e2458
19.
Khatri G, Merrick L, Miller FH. MR imaging of hepatocellular carcinoma. Magn Reson Imaging Clin N Am 2010; 18:421–450
20.
Theise ND, Curado MP, Frances-chi S, et al. Hepatocellular carcinoma. In: Bosman FT, Carneiro F, Hruban RH, Theise ND, eds. WHO classification of tumours of the digestive system, 4th ed. Lyon, France: IARC, 2010:205–216
21.
Villanueva A, Newell P, Chiang DY, Friedman SL, Llovet JM. Genomics and signaling pathways in hepatocellular carcinoma. Semin Liver Dis 2007; 27:55–76
22.
Sano K, Ichikawa T, Motosugi U, et al. Imaging study of early hepatocellular carcinoma: usefulness of gadoxetic acid-enhanced MR imaging. Radiology 2011; 261:834–844
23.
Kadoya M, Matsui O, Takashima T, Nonomura A. Hepatocellular carcinoma: correlation of MR imaging and histopathologic findings. Radiology 1992; 183:819–825
24.
Kajiwara M. MR imaging of small hepatocellular carcinoma (≤ 20 mm): correlation with vascularity and histological features. Kurume Med J 1997; 44:327–338
25.
Choi JY, Lee JM, Sirlin CB. CT and MR imaging diagnosis and staging of hepatocellular carcinoma. Part I. Development, growth, and spread: key pathologic and imaging aspects. Radiology 2014; 272:635–654
26.
Siripongsakun S, Lee JK, Raman SS, Tong MJ, Sayre J, Lu DS. MRI detection of intratumoral fat in hepatocellular carcinoma: potential biomarker for a more favorable prognosis. AJR 2012; 199:1018–1025
27.
Colecchia A, Schiumerini R, Cucchetti A, et al. Prognostic factors for hepatocellular carcinoma recurrence. World J Gastroenterol 2014; 20:5935–5950
28.
Lee EC, Kim SH, Park H, Lee SD, Lee SA, Park SJ. Survival analysis after liver resection for hepatocellular carcinoma: a consecutive cohort of 1002 patients. J Gastroenterol Hepatol 2017; 32:1055–1063
29.
Cucchetti A, Piscaglia F, Grigioni AD, et al. Preoperative prediction of hepatocellular carcinoma tumour grade and micro-vascular invasion by means of artificial neural network: a pilot study. J Hepatol 2010; 52:880–888
30.
Ariizumi S, Kitagawa K, Kotera Y, et al. A non-smooth tumor margin in the hepatobiliary phase of gadoxetic acid disodium (Gd-EOB-DTPA)-enhanced magnetic resonance imaging predicts microscopic portal vein invasion, intrahepatic metastasis, and early recurrence after hepatectomy in patients with hepatocellular carcinoma. J Hepatobiliary Pancreat Sci 2011; 18:575–585
31.
Shin SK, Kim YS, Shim YS, et al. Peritumoral decreased uptake area of gadoxetic acid enhanced magnetic resonance imaging and tumor recurrence after surgical resection in hepatocellular carcinoma: a STROBE-compliant article. Medicine (Baltimore) 2017; 96:e7761
32.
Zheng J, Chakraborty J, Chapman WC, et al. Preoperative prediction of microvascular invasion in hepatocellular carcinoma using quantitative image analysis. J Am Coll Surg 2017; 225:778.e1–788.e1
33.
Sakata J, Shirai Y, Wakai T, Kaneko K, Nagahashi M, Hatakeyama K. Preoperative predictors of vascular invasion in hepatocellular carcinoma. Eur J Surg Oncol 2008; 34:900–905
34.
Rodríguez-Perálvarez M, Luong TV, Andreana L, Meyer T, Dhillon AP, Burroughs AK. A systematic review of microvascular invasion in hepatocellular carcinoma: diagnostic and prognostic variability. Ann Surg Oncol 2013; 20:325–339
35.
Kim YY, An C, Kim S, Kim MJ. Diagnostic accuracy of prospective application of the Liver Imaging Reporting and Data System (LI-RADS) in gadoxetate-enhanced MRI. Eur Radiol 2018; 28:2038–2046
FOR YOUR INFORMATION
A data supplement for this article can be viewed in the online version of the article at: www.ajronline.org.
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: January 15, 2019
Accepted: April 11, 2019
First published: May 23, 2019
Keywords
Authors
Funding Information
Supported by grant 81271562 from the National Natural Science Foundation of China to J. Wang and grant 201704020016 from the Science and Technology Program of Guangzhou, China, to J. Wang.
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.