This study was approved by the Institutional Review Board of Zhongshan Hospital, Fudan University (approval number B2018-236) in accordance with the ethical guidelines of the Declaration of Helsinki. The committee waived the requirement for informed consent because it is a retrospective study.
Patients selection
According to the 2018 practice guidance by the AASLD [22], patients with preexisting cirrhosis were at high risk for developing HCC and the surveillance program of ultrasound (US) and/or alpha-fetoprotein (AFP) every 6 months were recommended. In our institution, between February 2016 and March 2017, 222 patients underwent gadoxetic acid-enhanced MRI for further evaluation of suspicious HCCs detected during surveillance. Patients with probable benign nodules (i.e., cysts, hemangiomas, arterioportal shunts) screened by US or HCC patients that having macrovascular invasion or receiving any previous treatments such as transcatheter arterial chemoembolization (TACE) and radiofrequency ablation (RFA) were initially excluded from the study. The inclusive criteria (Figure 1) of the patient selection were: (a) single HCC diagnosed by histology with preoperative gadoxetic acid-enhanced MRI; (b) the interval time between the MRI and the operation in less than 2 weeks; (c) patients in Child-Pugh A-B; (d) having qualified MR images.
MR Imaging Protocol
All patients enrolled in our study underwent gadoxetic acid-enhanced MRI in a single 1.5-T MR system (MAGNETOM Aera, Siemens Healthcare, Erlangen, Germany), with a 8-channel phased-array receiver coil. Single-spin echo plane DWI for free breathing (3200/56 milliseconds repetition time (TR) / echo time (TE), 84 × 128 matrix, 380-400 × 300-324 mm field of view (FOV), 5.5 mm slice thickness) was performed, and corresponding ADC maps were automatically generated with b values of 0 and 500 s/mm2. Dynamic contrast-enhanced T1-weighted 3D gradient-recalled echo images (3.47 / 1.36 TR / TE, 320 × 195 matrix, 10° flip angle, 308 × 380 mm FOV, 3 mm slice thickness) were obtained after intravenously injection of contrast agent. A dual flip-angle (Flip angle, 2° and 12°) before and at 20min after injection of gadoxetic acid based on a voxel-by-voxel basis was applied for automatically generating T1 maps with syngo MapIt (Syngo Offline and mono-exponential fit, Siemens Medical Solutions, Erlangen, Germany). The precontrast phase was obtained before a hand bolus injection of 0.025 mmol/kg of gadoxetic acid (Primovist, Bayer Schering Pharma, Berlin, Germany) at a rate of 1ml/sec with a subsequent 20ml saline flush. Subsequent MR images during the arterial phase (automatically triggered when the ascending aorta reached peak enhancement), the portal vein phase (about 14 seconds), the transition phase (about 3 minutes), and the hepatobiliary phase (HBP; 20 minutes) were obtained.
MR images analysis
Two abdominal radiologists (C.Y.R. with 5 years of experience and X.Q.W. with 10 years of experience) independently reviewed the MR images. Both the two radiologists knew that all the patients were having HCC but they were blinded to the histology of MVI. The region of interest (ROI) in a lesion was outlined around the edge of tumor on each slice on the precontrast T1 maps, hepatobiliary phase T1 maps and the ADC maps by the two radiologists. The ROIs were copied from the same ROIs that drawn on the high flip angle (12°) T1-weighted images and high b value (500 s/mm2) images, respectively (Figure 2). Precontrast and postcontrast T1 relaxation time were obtained from the precontrast T1 maps and hepatobiliary phase T1 maps, respectively. The ADC values of HCCs were also calculated by the two observers. The reduction rate (Δ%) of T1 relaxation time were calculated by using the following formula: Δ%= 100% × (pre T1 value − post T1 value) / pre T1 value [14, 20], in which the pre T1 and post T1 values representing the T1 relaxation time before and after the injection of gadoxetic acid.
Reference standard for MVI
Pathological data including presence of cirrhosis, Edmondson-Steiner grade of I, II, III and IV according to the nuclear grading scheme [23] or microvascular invasion was according to surgical pathologic reports generated by our institutional pathologists specialized in liver histology (each individual with more than 20 years of experience). Microvascular invasion was defined as presence of tumor invasive in any portal vein, hepatic vein, or a large capsular vessels lined by endothelium that visible only microscopically.
Statistical Analysis
Frequencies of categorical variables for differentiating MVI were compared by using Fisher exact test. Difference of quantitative variables including precontrast/postcontrast T1 relaxation time and ADC values between MVI-positive and MVI-negative groups was compared by using independent sample t test. The interclass correlation coefficient (ICC) of quantitative data between the two observers was calculated (poor: <0.40; fair: 0.40-059; good: 0.60-0.74; excellent: 0.75-1.00). Area under receiver operating characteristic curve (AUC) with 95% confidence interval (95% CI) based on receiver operating characteristic curve (ROC) analysis was generated for evaluating the utility of variables to discriminate the status of MVI. Sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV) and likelihood ratio (LR) of appropriate cutoff value corresponding to maximal Youden index by using ROC analysis were calculated with 95% CI. All the statistical tests were performed by using statistical software (SPSS version 21, SPSS, Chicago, III) and a two-side P value less than 0.05 indicating significance level.