Hepatocellular carcinoma (HCC) is the most common primary liver malignancy and the third leading cause of cancer-related mortality worldwide [1], [2]. Unlike most other malignancies, HCC can be diagnosed noninvasively in high-risk patients based on its typical imaging features, that is, arterial phase hyperenhancement (APHE) and washout on portal venous phase (PVP) or delayed phase (DP) imaging [3]. Noninvasive imaging examinations include various modalities, such as contrast-enhanced MRI (CE-MRI) or contrast-enhanced CT (CECT) and gadoxetic acid-enhanced MRI (EOB-MRI). Several previous studies have reported that EOB-MRI showed higher diagnostic performance than either CE-MRI or CECT [4], [5].

The Liver Imaging Reporting and Data System (LI-RADS), published by the American College of Radiology in 2011 and recently updated in 2018, that is, LI-RADS version 2018 (LI-RADS v2018), standardized the categorization of focal hepatic observations obtained in patients at high risk of HCC according to the probability of HCC [6]. The LI-RADS describes several major features (MFs) and ancillary features (AFs) of hepatic observations. It uses a combination of MFs to assign initial categories and then adjusts the categories according to AFs. In 2014, the rules for the use of EOB-MRI were added to the LI-RADS algorithm and continued to be refined in subsequent updated versions. However, recent studies have shown that the low sensitivity of LI-RADS is more problematic for EOB-MRI than for extracellular contrast agent-enhanced MRI (ECA-MRI) [7], [8]. This finding may have occurred because LI-RADS was originally designed for ECA-MRI and EOB-MRI was incorporated later into its algorithm [9]. Therefore, the additional benefit of EOB-MRI for LI-RADS diagnostic performance remains controversial.

To achieve the highest specificity, with LI-RADS, only PVP hypointensity on EOB-MRI is defined as washout, and the transition phase (TP) or hepatobiliary phase (HBP) hypointensity is only regarded as an AF favoring malignancy; however, it is unable to upgrade the categorization to LR-5 (definitely HCC) [6]. Strict adherence to defining washout severely degrades the sensitivity of HCC diagnosis (66.3 % to 77.9 %) [10], [11], [12]. In contrast, other guidelines in Asia suggest using TP and/or HBP hypointensity as definitions of washout [13], [14]. In Asia, the sensitivity of HCC diagnosis is often more important for guiding subsequent treatment.

Therefore, this study aimed to improve the diagnostic performance of HCC by expanding the definition of washout.

This study was approved by the Ethics Committee of Tianjin Third Central Hospital. Informed consent was waived because of the retrospective study design.

Using electronic medical records, data from patients at high risk of HCC [15] were retrospectively collected from June 2016 to June 2021 (Fig. 1). The inclusion criteria were as follows: (a) age of ≥ 18 years, (b) underwent EOB-MRI, (c) nodule number ≤ 3, and (d) definitive diagnosis by surgical resection or ultrasound-guided core needle biopsy pathology. The exclusion criteria were as follows: (a) cirrhosis due to congenital hepatic fibrosis or vascular disorders, (b) interval of > 1 month between pathologic diagnosis and EOB-MRI, (c) diffuse hepatic lesions, (d) received treatment before EOB-MRI, (e) liver function was Child–Pugh C because hepatic dysfunction can impede adequate quality of HBP images of EOB-MRI [16], or (f) LR-NC (cannot be categorized due to image degradation or omission) or LR-TIV (definite tumor in vein) as defined in LI-RADS v2018 (Fig. 1). Patient demographic, clinical, and laboratory characteristics were extracted from electronic medical records.

All examinations were performed using a 3.0-T MR system (Magnetom Verio; Siemens Healthcare, Erlangen, Germany), and an eight-channel phased-array torso coil was used for all measurements. The liver MR imaging protocol consisted of in- and out-of-phase T1-weighted imaging acquired with a gradient recalled echo (GRE) sequence, a respiratory-triggered axial T2-weighted turbo spin echo sequence with fat suppression, free-breathing single-shot echo-planar diffusion-weighted imaging with b values of 0, 50, 600, and 1,000 s/mm2, and pre and postcontrast T1-weighted three-dimensional volumetric interpolated breath-hold examination sequences acquired with a GRE sequence in the early arterial phase (AP) (10 s after aortic enhancement using the bolus tracking method), late AP (35 s), PVP (46–60 s), TP (150–180 s), and HBP (20 min after the contrast injection). Contrast-enhanced dynamic MR images of the liver were obtained after intravenous administration of gadoxetic acid (Primovist; Bayer Healthcare, Leverkusen, Germany) at 0.025 mmol/kg of body weight at a rate of 1.0 mL/s using a power injector, followed by 25 mL of 0.9 % saline as a chaser at the same rate. The detailed parameters of each acquisition sequence are shown in Table 1.

All MRI scans were independently reviewed by two board-certified radiologists with 10 (WJH) and 8 (DW) years of experience in abdominal MRI. All readers were blinded to the pathologic results and reviewed all imaging features for each liver observation according to LI-RADS v2018. The imaging features included MFs for HCC (size, APHE, nonperipheral “washout,” enhancing “capsule,” and threshold growth), targetoid features, AFs favoring malignancy in general but not HCC in particular, AFs favoring HCC in particular, and AFs favoring benignity. Features that could not be evaluated, for example, threshold growth and subthreshold growth, could not be measured because of a lack of corresponding follow-up images; fat sparing could not be assessed in a patient without steatosis and was considered not applicable.

After scoring all imaging features, the LI-RADS category was assigned to each observation according to the conventional LI-RADS v2018 (the definition of washout was confined to PVP hypointensity). Then, observations of the liver excluding the LR-1 (definitely benign), LR-2 (probably benign), and LR-M (probably or definitely malignant but not HCC specific) nodules were reassigned to the category according to the modified LI-RADS (mLI-RADS) using a different definition of washout. The definition of washout was as follows: (a) PVP or TP hypointensity, (b) PVP or HBP hypointensity, and (c) PVP or TP or HBP hypointensity. Discrepancies between the two readers were resolved by a third radiologist (R.L., with 17 years of experience in abdominal MRI) to reach a final consensus reading.

Histological diagnoses, including biopsy and surgical pathology, were all based on the formal pathologic reports of our institution by one of two pathologists who had > 10 years of experience in liver pathology.

Categorical variables are summarized as counts and percentages of patients/observations in the corresponding categories. Normally distributed continuous variables are summarized as means and standard deviations, and the median (interquartile range) was used for non-normally distributed data. Perobservation estimates of diagnostic performance, including sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and Youden index, were computed. The sensitivity, specificity, and accuracy were compared using McNemar’s test. Unless otherwise indicated, all statistical tests were conducted at the 0.05 significance level using two-tailed tests, and p values are reported. Statistical analyses were performed using the SPSS software, version 25.0 (SPSS Inc.).