Identification of the origin of tumor in vein: comparison between CEUS LI-RADS v2017 and v2016 for patients at high risk

We reviewed our database of consecutive patients who underwent both baseline ultrasound (US) and CEUS between April 2014 and December 2018 at a single hospital. The inclusion criteria were (1) High risk of HCC according to CEUS LI-RADS guideline [15, 16]. (2) All FLLs were newly detected by US and CEUS. (3) All patients met the diagnosis of intrahepatic TIV reference standard, details are described below. The exclusion criteria were (1) Cirrhosis due to congenital hepatic fibrosis or vascular disorder such as Budd-Chiari syndrome, hereditary hemorrhagic telangiectasia, cardiac congestion, chronic portal vein occlusion, or diffuse nodular regenerative hyperplasia. (2) Poor image quality could not to categorize CEUS LI-RADS category.

The reference standard for TIV was definite enhancing soft tissue in vein, regardless of visualization of a parenchymal mass and visible adjacent primary tumor in liver parenchyma (Fig. 1) [15, 16], then with the surgical pathologic confirmation or three months imaging follow-up remained the diagnosis of TIV. All FLLs were diagnosed by either histopathology (surgery or puncture biopsy pathology, the biopsy procedure were described in Additional file 1: Section S1) or composite imaging reference standard. The composite standard for HCC was FLLs with CECT or MRI scans characterized by the arterial phase hyperenhancement (APHE) and washout appearance in the portal venous phase or delayed phase in a size greater than 1 cm in diameter [20]. Other malignant lesions were diagnosed by histopathology only.

Ultrasound examinations were performed as follows: (1) Aplio 500 (Toshiba Medical Systems, Tokyo, Japan) scanner equipped with a 375BT convex transducer (frequency range, 1.9–6.0 MHz) and a Contrast Harmonic Imaging mode; and (2) Aixplorer Ultrasound system (SuperSonic Imagine, Aix-en-Provence, France) scanner equipped with the SC6-1 convex probe (frequency range, 1.0–6.0 MHz). Ultrasound examination was performed in a standardized fashion. First, all patients underwent unenhanced abdominal and hepatic sonography using the conventional US. The location, number, size and sonographic features of the FLLs and suspicious TIV were recorded. Select the most suspicious FLL for CEUS in case with more than one lesion. When the examination was done applying CEUS mode, low-mechanical index and dual screen format were used. 2.4 mL of the SonoVue was injected intravenously and immediately flushed with 5 mL 0.9% saline, then the FLL was observed for at least 5 min. The specific features appear was recorded, which could be used to differentiate the different liver masses. If a suspicious TIV was found in US or at first time of CEUS, then injecting 2.4 mL of the SonoVue again to observe if there is enhancement in the arterial phase. The enhancement phases were defined as follow: arterial phases, < 30 s after contrast injection; portal venous phase, 31 s-2 min after injection; and late phase, 2 min to approximately 4–6 min after injection.

Imaging archiving and scanning parameters of CT and MRI examinations were described in Additional file 1: Section S2.

All US and CEUS images were anonymous, and independently evaluated by two skilled radiologists (Y.H. and M.X.L.) who had more than 8 years’ experience in CEUS. They were unaware of clinical and other imaging information of the patients. If the diagnostic results were inconsistent, another radiologist (W.W.) with at least 11 years of experience in liver CEUS assessed the images to reach a consensus.

CEUS characteristics of FLL were recorded: (1) maximum diameter of the target lesion; (2) enhancement level in the arterial/portal/late phase (hyper/iso-/hypo-); (3) enhancement patterns in the arterial phase (rim/homogeneous/inhomogeneous/others); (4) washout time (early washout within 60 s or not) [21]; (5) degree of washout (no washout/mild washout/marked washout [< 2 min]) [15]. The enhancement level in the arterial phase (hyper/iso-/hypo-) of TIV was recorded. TIV was categorized as both LR-5V v2016 and LR-TIV v2017. The FLLs were classified according to CEUS LI-RADS category. The TIV originated from these contiguous FLLs were assessed according to FLLs CEUS LI-RADS category, and categorized as LR-3 subclass/ LR-4 subclass/ LR-5 subclass /LR-M subclass of LR-TIV v2017. The LR-3 subclass meant LR-TIV, etiology uncertain; the LR-4 subclass meant LR-TIV, probably due to HCC; the LR-5 subclass meant LR-TIV, definitely due to HCC; the LR-M subclass meant LR-TIV, may be due to non-HCC malignancy.

Statistics were calculated using SPSS Statistics software v22.0 program and Microsoft Excel 2019. LR-5V v2016 considers the originated tumor of TIV as HCC. LR-TIV v2017 considers the originated tumor of TIV based on the contiguous FLL CEUS LI-RADS category. LR-5 subclass and LR-4/5 subclass of LR-TIV v2017 were used as diagnosis criteria for HCC separately. Diagnostic performance of CEUS LI-RADS v2016 and v2017 in identifying the originated tumor of TIV was assessed with sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV). McNemar’s test and Fisher's test was applied for comparison the diagnostic performance between CEUS LI-RADS v2016 and v2017 in identifying the originated tumor of TIV.

A flow diagram of our study is showed in Fig. 2. A total of 273 FLLs (mean size, 9.5 ± 3.71 cm) in 273 patients (median age, 50.0 years; 249 male) were final included. 168 (61.5%) of 273 FLLs developed in patients with chronic hepatitis B virus (HBV) infection only, 81 (29.7%) with chronic HBV infection and cirrhosis, 2 (0.7%) with chronic HCV infection and cirrhosis, 2 (0.7%) with chronic HBV and HCV infection and cirrhosis, and 20 (7.3%) with cirrhosis only. Patients and FLLs clinical characteristics are shown in Table 1. As for the reference standard, there were 266 HCC (97.4%) and 7 (2.6%) non-HCC malignancies. In 266 HCCs, 146 (54.9%) were diagnosed as HCC by pathology, and 120 (45.1%) were diagnosed as HCC by CECT or MRI images. In 7 non-HCC malignancies, including 3 ICCs (1.1%), 1 combined hepatocellular-cholangiocarcinoma (CHC) (0.4%), 1 hepatosarcoma (HSC) (0.4%), 1 primary neuroendocrine tumor (0.4%), and 1 other malignant tumor (0.4%), were confirmed by pathologic diagnosis. Other characters including US and CEUS characteristics, are presented in Additional file 1: Table S1.

When applying CEUS LI-RADS v2017 (Table 2), in the 273 TIV cases, there were 2 (0.7%) LR-3 subclass of LR-TIV v2017 cases confirmed as HCC; 6 (2.2%) cases were LR-4 subclass and confirmed as HCC; 164 (60.1%) were LR-5 subclass including 163 HCCs and 1 HSC; 101 (37.0%) were LR-M subclass including 95 HCCs, 3 ICCs, 1 CHC, 1 primary neuroendocrine tumor, and 1 other malignant tumor. As the level of CEUS LI-RADS rose, the proportion of HCC in each level to the overall HCC increased gradually (LR-3 subclass, 0.8% [2/266]; LR-4 subclass, 2.3% [6/266]; LR-5 subclass, 61.3% [163/266]). In the 266 HCCs, 95 (35.7%) were categorized as LR-M subclass.

In Table 3, when regarding LR-5 subclass of LR-TIV v2017 as a criteria of HCC diagnosis, the sensitivity, specificity, accuracy, PPV and NPV were 61.3%, 85.7%, 61.9%, 99.4% and 5.5%, respectively. when regarding LR-4/5 subclass of LR-TIV v2017 as a criteria of HCC diagnosis, the sensitivity, specificity, accuracy, PPV and NPV were 63.5%, 85.7%, 64.1%, 99.4% and 5.8%, respectively.

When applying CEUS LI-RADS v2016 (Table 3), all cases were categorized as LR-5V v2016 and considered the originated tumor as HCC, the accuracy and PPV were both 97.4%. Statistical analysis revealed that the PPV of LR-5V v2016 and LR-TIV v2017 were both pretty high without significant differences (p > 0.05, both LR-5 subclass and LR-4/5 subclass of LR-TIV v2017). The accuracy of LR-5V v2016 was higher than that of LR-TIV v2017 with remarkable differences (p < 0.001, both LR-5 subclass and LR-4/5 subclass of LR-TIV v2017).

In Table 2, 164 (60.1%) TIV were categorized as LR-5 subclass of LR-TIV v2017 including 163 (99.4%) HCCs (Fig. 3) and 1 (0.6%) HSC. The FLL of HSC showed inhomogeneous hyperenhancement in the arterial phase, no marked washout (< 2 min), and washout time > 60 s.

Of the 101 LR-M subclass of TIV v2017, 95 (94.1%) were confirmed as HCCs, 3 were ICCs (3.0%), 1 was CHC (1.0%), 1 was primary neuroendocrine tumor (1.0%), and 1 was unspecified malignant tumor (1.0%). Among the 95 HCC categorized as LR-M, one (1.1%) FLL exhibited rim APHE, 91 (95.8%) FLLs exhibited washout time < 60 s, and 19 (20.0%) FLLs exhibited marked washout (< 2 min). Of those, there were 14 (14.7%) FLLs had two CEUS features and one (1.1%) FLL had all three CEUS features mentioned above (Fig. 4).