This study prospectively enrolled consecutive adult participants who were suspected of having iCCA based on previous CT or ultrasonography examinations and were referred to our hospital for liver surgery between January 2021 and April 2022. The study was approved by the institutional review board and written informed consent was obtained from each participant. Participants with no contraindications for gadopentetate dimeglumine underwent contrast-enhanced MRI examination with MRE sequence in our institute, all patients should fast 4–6 h prior to the examination. All patients included underwent liver surgery. Exclusion criteria were (1) previous history of local or systemic oncologic treatment; (2) lesions pathologically diagnosed as other tumors rather than iCCA; (3) small tumors less than 1 cm (to avoid incorrect tumor stiffness measurements); (4) lesions located subcapsular (areas within 1 cm from the liver capsule); (5) time interval between MR scan and surgery more than 1 month; and (6) difficult to measure stiffness values because of poor image quality. The flowchart of the inclusion and exclusion criteria is presented in Fig. 1.

Flowchart of the inclusion and exclusion criteria. MRE, magnetic resonance elastography; TACE, transcatheter arterial chemoembolization; HCC, hepatocellular carcinoma; cHCC-CC, combined hepatocellular carcinoma-cholangiocarcinoma; iCCA, intrahepatic cholangiocarcinoma

The two-dimensional MRE examinations were acquired with a standard commercial equipment at a 3.0-T MRI scanner (uMR 770, United Imaging Healthcare, Shanghai, China). The shear wave frequency was set as 60 Hz. A passive pneumatic driver located on the right lobe of the liver and centered at the level of the xiphisternum was utilized to generate mechanical vibrations and produce propagating shear waves in the imaging region. The 2D-MRE scanning protocol was based on an axial spin-echo echo-planner-imaging sequence and the detailed scanning parameters are as follows: TR: 1000.2 ms, TE: 44.6 ms, flip angle: 90°, field of view: 420 × 420 mm2, slice thickness: 8 mm, matrix: 256 × 256, bandwidth: 1500 Hz, scanning time: 10 s. MRE sequence was performed during one breath-hold at end-expiration. The magnitude, phase, wave, and elastographic images were transferred offline and the parametric maps were generated with a custom software package (MRE Quant, Mayo Clinic, Rochester, MN).

Other conventional sequences consisted of a breath-hold T2-weighted fat-suppressed fast spin-echo sequence, T1-weighted in-phase, and opposed-phase gradient echo sequence, respiratory-triggering single-shot spin-echo echoplanar diffusion-weighted imaging (DWI) with b values of 0, 50, and 500 s/mm2. Dynamic imaging was performed with a 3D breath-hold T1-weighted fat-suppressed gradient-echo sequence, prior to and after intravenous administration of gadopentetate dimeglumine (Magnevist; Bayer HealthCare, Berlin, Germany) at a rate of 2 mL/s and at a dose of 0.1 mmol/kg. The arterial phase acquisitions were automatically triggered when contrast media reached the ascending aorta. Subsequent acquisitions were performed at 60–70 s for the portal venous phase and 180 s for the delay phase. Detailed parameters are shown in Table S1.

MRE images were independently evaluated by two radiologists (with 5 and 10 years of experience in liver MRI, respectively). The reviewers were aware that the patients had iCCA, but were blinded to all other information, including the Ki-67 labeling index, clinical history, and laboratory results. Regions of interest (ROIs) were drawn for the lesion and unaffected hepatic parenchyma at the same level and segment, the tumors were guaranteed to be located in their entirety in the 95% confidence area of the map. By using T2-weighted and contrast-enhanced images as references, ROIs were manually drawn on magnitude images, including the solid tumor area as large as possible, and then copied to the stiffness maps, the most peripheral portions of tumors were avoided to exclude partial volume effects. Great care was also taken to avoid areas of significant wave interference and necrosis. For each case, 3 adjacent slices with a maximum cross-section of the tumor were chosen. Stiffness values of the lesion and liver were measured on each slice, and the average value of the 3 measurements was used. Mean values measured by two observers were averaged for final analysis.

The conventional MRI images were assessed by another two radiologists (with 7 and 13 years of experience in liver MRI, respectively), who were blinded to the results of the Ki-67 labeling index and MRE analysis. When disagreement occurred in the qualitative analysis between the two observers, a consensus review was made by a third senior radiologist (with 35 years of experience in abdominal MRI) for the final decision.

The qualitative imaging parameters were evaluated as follows: (a) location (left/right/left and right/caudal lobe); (b) tumor margin (well-defined: well-defined tumor with distinct contour/ill-defined: ill-defined tumor with indistinct contour); (c) signal homogeneity (homogeneous: the entire tumor was uniform with homogeneous signal inside/heterogeneous: the entire tumor was nonuniform with heterogeneous portion compared with the main body of tumor). Signal homogeneity was evaluated on T2-weighted imaging, a cut-off of ≥ 10% heterogeneous regions of the entire tumor was regarded as heterogeneous, and heterogeneous regions < 10% was defined as homogeneous; (d) arterial phase enhancement (global hyperenhancement: increased signal relative to the liver parenchyma, involving the totality of lesion/partial hyperenhancement: increased signal involving ≥ 25% of the lesion, except the central area/peripheral enhancement: increased signal limited to the periphery of the lesion, involving < 25% of the lesion); (e) enhancement pattern (progressive: increasing enhancement over time/persistent: invariable enhancement over time/degressive-washout: decreasing enhancement over time); (f) arterial peritumoral hyperenhancement (defined as fuzzy-marginated hyperenhancement outside the tumor borders that becomes isointense with normal liver parenchyma in later dynamic phases); (g) enhancing tumor capsule (smooth, uniform, sharp border around most or all of tumor, and visible as an enhancing rim in portal venous or delayed phases); (h) targetoid appearance (rim arterial phase hyperenhancement, peripheral washout, delayed central enhancement, or targetoid restriction on DWI); (i) bile duct dilation with diameter ≥ 5mm; (j) liver capsule retraction; (k) hemorrhage in mass (defined as high-signal foci on T1-weighted images with variable signal intensity on T2-weighted images); (l) necrotic or cystic portion in mass (defined as bright signal foci on T2-weighted images without contrast enhancement); (m) central scar (central or eccentric area within a tumor with stellate appearance and radiating septa); (n) central darkness on T2-weighted imaging (central signal darker than liver signal); (o) vessel invasion (defined when vessels cannot be separated from the mass, with the rough change of the wall or narrowing and occlusion of the lumen); (p) lymphadenectasis; (q) distant metastasis. In cases of multiple tumors with satellite nodules, the major tumor with the largest size was analyzed.

Tumor apparent diffusion coefficient (ADC) values were measured by ROIs manually drawn in ADC maps. Slice locations of ROIs were selected in consistent with stiffness measurement as much as possible, avoiding large vessels, necrosis, hemorrhage, and artifacts. Similarly, 3 ROIs were drawn for each case and the average value was used. Tumor size (the largest diameter) was measured in the delay phase by the senior reviewer.

Pathologic characteristics were evaluated by an experienced pathologist with 30 years of experience in liver pathology. The Ki-67 level was determined by using the percentiles of immunoreactive cells from 1000 malignant cells (× 400), and scoring was performed in the areas with the highest number of positive nuclei (hot spot) within the tumor. Then, we classified iCCAs into the “high Ki-67 group” (positive staining ratio ≥ 50%) and “low Ki-67 group” (positive staining ratio < 50%), referring to prior researchers [14, 21, 22].

Statistical analysis was performed using SPSS 26.0 (SPSS, Armonk, NY, USA) and MedCalc software (www.medcalc.org). Continuous variables were compared with the Student’s t-test or Mann–Whitney U-test; categorical variables were compared using Pearson’s chi-squared test or Fisher’s exact test. The interobserver agreement on qualitative imaging findings was determined using kappa statistics: poor, 0–0.2; fair, 0.2–0.4; moderate, 0.4–0.6; good, 0.6–0.8; and excellent, 0.8–1.0. The interobserver agreement on the quantitative findings was determined using intraclass correlation coefficient (ICC) (two-way random, absolute agreement, single measurements): poor, < 0.5; moderate, 0.5–0.75; good, 0.75–0.9; and excellent, > 0.9. Variables showing p < 0.05 in the univariate logistic regression analysis were applied to multivariate logistic regression analysis. Receiver operating characteristic analysis was performed to assess the ability of tumor stiffness in distinguishing the high Ki-67 iCCAs from the low Ki-67 iCCAs, and the specificity, sensitivity, and accuracy were calculated for the corresponding area under the curve (AUC). In addition, the threshold values for tumor stiffness and ADC were evaluated based on the best Youden’s index on the receiving operating characteristic curve. Spearman correlation analysis was performed to analyze the correlations between tumor stiffness and the Ki-67 labeling index. The correlation was very weak for absolute value of correlation coefficient |r|= 0.0–0.2, weak for |r|= 0.2–0.4, moderate for |r|= 0.4–0.7, strong for |r|= 0.7–0.9, very strong for |r|= 0.9–1.0.