New Liver MR Imaging Hallmarks for Small Hepatocellular Carcinoma Screening and Diagnosing in High-Risk Patients

Objective Early detection and diagnosis of hepatocellular carcinoma (HCC) is essential for prognosis; however, the imaging hallmarks for tumor detection and diagnosis has remained the same for years despite the use of many new immerging imaging methods. This study aimed to evaluate the detection performance of hepatic nodules in high risk patients using either hepatobiliary specific contrast (HBSC) agent or extracellular contrast agent (ECA), and further to compare the diagnostic performances for hepatocellular carcinoma (HCC) using different diagnostic criteria with the histopathological results as reference standard. Methods This prospective study included 247 nodules in 222 patients (mean age, 53.32 ± 10.84 years; range, 22–79 years). The detection performance and imaging features of each nodule were evaluated in all MR sequences by three experienced abdominal radiologists. The detection performance of each nodule on all MR sequences were compared and further the diagnostic performance of various diagnostic criteria were evaluated. Results For those patients who underwent ECA-MRI, the conventional imaging hallmark of “AP + PVP and/or DP” was recommended, as 60.19% diagnostic sensitivity, 80.95% specificity and 100% lesion detection rate. Additionally, for those patients who underwent HBSC-MRI, the diagnostic criteria of “DWI + HBP” was recommended. This diagnostic criteria demonstrated, both in all tumor size and for nodules ≤2 cm, higher sensitivity (93.07 and 90.16%, all p <0.05, respectively) and slightly lower specificity (64.71 and 87.50%, all p >0.05, respectively) than that of the European Association for the Study of the Liver (EASL) criteria. Conclusions Different abbreviated MR protocols were recommended for patients using either ECA or HBSC. These provided imaging settings demonstrated high lesion detection rate and diagnostic performance for HCC.


INTRODUCTION
Current guidelines recommend every 6 months surveillance by tumor marker measurement and ultrasonography (US) in patients with chronic liver disease (CHD) or liver cirrhosis (1)(2)(3)(4). The standard procedure in those situations with elevated level of alpha-fetoprotein (AFP) or detected nodules requires further contrast enhanced ultrasonography (CEUS), dynamic computed tomography (CT), dynamic magnetic resonance imaging (MRI) or even liver biopsy identification (5)(6)(7). Multiparametric MRI has been shown to improve the detection of clinically significant hepatic nodules especially in small and early hepatocellular carcinoma (HCC), and evidence also suggested that MRI helps to avoid the detection of clinically insignificant nodules compared with US and CT and avoid unnecessary liver biopsy identification (8)(9)(10)(11). As clinical pathways that use liver MR imaging in patients with either elevated serum AFP level or detected hepatic nodules appeared advantageous, the demand for liver MR imaging is increasing.
At present, the multiparametric liver MR imaging has been designed in several different clinical scenarios and thus to answer several different questions such as the presence or absence of liver cancer, characterization of tumor, whether the tumor emboli invaded the portal vein, or with metastatic invasion, and even the presence of microvascular invasion, the tumor histological grade or future remnant liver function estimation (12)(13)(14)(15)(16)(17)(18). Of all these clinical scenarios, the liver MR protocols are designed not only emphasized on the diagnosis but further toward a better comprehensive management of treatment and prognostic evaluation (19,20). In order to satisfy the broad range of clinical needs fulfilled by liver MR imaging, the protocols are designed comprehensively so that multi-planner high-resolution MR sequences and functional MR sequences were included (21,22). However, these multiparametric liver MR imaging protocols are time consuming and directly influence the patient throughput (23). Additionally, the duration of scanning time is also a major determinant of the direct costs of an MR examination. From the perspective of health economics, the long examination time related patient throughput and direct costs will greatly limit the patient availability of an imaging test. Furthermore, due to more frequent breath-holding times, especially performed with liver MR imaging (24,25), may further limit the patient acceptance of this imaging test.
Since the multiparametric liver MR imaging protocols are designed in different clinical scenarios and used to deal with different clinical questions, in a man with CHD or liver cirrhosis, the primary concern is whether clinically significant high-risk hepatic nodule is present or not. It has been reported that the abbreviated MRI (aMRI) protocols are applicable in breast cancer and prostate cancer screening which demonstrated high detection and diagnostic sensitivity (23,26). Thus, with the use of aMRI protocols, it is expected that the problems, namely, patient throughput and high cost caused by long examination time can be solved and further increase the availability of this imaging test. Several published studies have investigated the utilization of aMRI in liver MR imaging. However, for these studies, the nodule size was not strictly restricted (<3 cm) which may potentially increase the screening performance (27). Moreover, previous studies only investigated the use of aMRI in hepatobiliary specific contrast (HBSC) agent (28,29), but there are still large proportion of extracellular contrast agent (ECA) use. Thus, there is little known about the screening performance of aMRI protocol in the ECA group.
Therefore, the purpose of this study was to prospectively evaluate the detection performance of hepatic nodules in high risk patients using either HBSC or ECA, and to further compare the diagnostic performances for hepatocellular carcinoma (HCC) using different diagnostic criteria with the histopathological results as reference standard.

Patients
This study was approved by the institutional review board and the written informed consent was obtained from all patients. From January 2017 to June 2020, four hundred and eighty-five consecutive high-risk patients (hepatitis B or C virus infection, or liver cirrhosis) who were suspected of having focal hepatic nodules were potentially enrolled. Of those patients, they were categorized as HBSC and ECA groups according to the use of contrast agent. Patients were ineligible if they have a history of hepatectomy, transarterial chemotherapy (TACE), radiofrequency ablation (RFA), colorectal cancer liver metastasis (CRLM) and MR contraindications. Moreover, the nodule size >3 cm or nodule number >3 in a single patient were also excluded as the aim was to detect the early-stage disease that meet the Barcelona Clinic Liver Cancer (BCLC) staging system. The inclusion criteria were as follows: (1) patients were eighteen years older, (2) nodules were not typically cysts or hemangiomas, and (3) nodules were pathologically confirmed. An experienced coordinator with five years of liver MR imaging retrieved and deidentified all patient images, and hepatic observations presented on MR images were blindly matched with these hepatic nodules presented on surgical removed specimens and with determined histopathological results. Detailed information about the inclusion criteria are shown in Figure 1.

Imaging Technique
MR imaging was carried out using 3.0 T MR systems (Discovery 750w, GE Healthcare, Milwaukee, USA; Skyra 3.0 T, Siemens Healthcare, Erlangen, Germany). Sixteen-channel phased-array torso coils were used for all measurements. In each patient, they were asked to prepare to fast for 6-8 h before MR examination. In addition to localizers, the standard liver MR protocols were: 1) Coronal single shot fast spin echo (SSFSE) T2-weighted imaging, 2) Axial SSFSE T2-weighted with fat saturation (FS), 3) Axial diffusion weighted imaging (DWI), 4) Axial in and out of phase T1-weighted imaging, 5) Axial non-enhanced T1weighted imaging with FS, 6) Axial contrast-enhanced T1weighted imaging with FS at arterial phase (AP) (25-30s), portal venous phase (PVP) (60s) and delayed phase (DP) (180s), 7) Coronal contrast-enhanced T1-weighted imaging with FS at delayed phase, and 8) for the HBSC group, the axial contrast-enhanced T1 weighted with FS at transitional phase (3-5 min) and hepatobiliary phase (HBP) (20 min) was also obtained. Contrast agent was administered intravenously using a bolus injection of 0.025 mmol/kg (Primovist, Bayer Pharma AG, Berlin, Germany) at the injection rate of 1 ml/s or for a dose of 0.1 mmol/kg (Magnevist, Bayer Pharma AG, Berlin, Germany) at the injection rate of 2 ml/s, followed by a 20-ml saline flush.

Image Analysis
Three independent reviewers (readers with 6, 8, and 12 years in liver MR imaging) who were blinded to the histopathological results, clinical and prior imaging data reviewed the MR images; when the three reviewers cannot fully agree with each other, a consensual results was achieved by using the majority assessment results. Firstly, the three reviewers should independently review and determine the presented hepatic nodules on each MR sequence and the detection performance of each MR sequence was calculated. According to the detection performance, different MR sequences were combined as aMRI protocols in different clinical settings. In this study, for the HBSC-MRI group, five aMRI protocols were created, namely,: 1) a-MRI-I: DWI + T2WI; 2) a-MRI-II: DWI + HBP; 3) a-MRI-III: AP+PVP and/ or TP+ and/or HBP (Korean Guidelines 2018) (5); 4) a-MRI-IV: AP + PVP and/or +HBP (Japan Society of Hepatology Guideline 2014) (30); 5) a-MRI-V: AP + PVP only (European Association for the Study of the Liver (EASL)) (31); 6) a-MRI-VI: The Liver Imaging Reporting And Data System 2018 (LI-RADS v2018), and when the nodule was assessed as LI-RADS 4 or 5, the nodule was finally determined as HCC with the LI-RADS criteria. In addition, for the ECA-MRI group, three protocols, namely, the conventional imaging hallmark of "AP + PVP and/or DP", "DWI + T2WI"and LI-RADS v2018 were created. Of these aMRI protocols, the detection performance was calculated and the best aMRI protocol in different clinical settings was determined. Furthermore, the positive imaging features (32,33) were defined as T2WI: moderate to hyperintensity; DWI: diffusion restriction; HBP: hypointensity but exclude the targetoid sign; Arterial phase hyperenhancement (APHE): the enhancement greater in whole or in part than liver but exclude the rim-like enhancement; and the washout feature at PVP or DP. MR images were presented in a randomized manner, and review of the different aMRI protocols in the same patient was separated by a delay period of 3-4 weeks to minimize recall.

Statistical Analysis
Lesion detection was summarized by using frequencies and percentages by MR sequence. Of the detection performance by each MR protocol, the per-lesion detection performance was calculated and the detection performance in different MR sequence was compared by using McNemar's test. The diagnostic sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and diagnostic accuracy of ECA-MRI and HBA-MRI were calculated, and also their 95% confidence interval (CI). The comparison of the sensitivity and specificity was tested by using McNemar's test, and the area under curve (AUC) value for each aMRI protocol was calculated. Subgroup analyses were conducted for small (≤20 mm) and larger HCCs (21-30 mm). A two-sided p-value of less than 0.05 was considered to indicate statistical significance. All statistical analyses were performed by using a statistical software package [SPSS 23.0 (SPSS Inc., Chicago, IL, USA)].

Detection Performance of HBSC-MRI
The detailed information about the detection performance of various MR sequences with HBSC-MRI are listed in Table 1. The detection performance of HBP was significantly higher than that of DWI (p = 0.008), PVP (p <0.001) and TP (p = 0.016), but no difference was obtained from DWI vs PVP (p = 0.167), DWI vs TP (p = 1.000). For the detection performance of aMRI Regarding the performance of tumor size ≤20 mm (Table 3), the a-MRI-II achieved the highest diagnostic sensitivity and specificity  of 90.16 and 87.50%, and the Youden index was 0.78. Considering the classic imaging hallmarks (EASL Guideline), the sensitivity and specificity were 44.26 and 87.50%, respectively. The diagnostic sensitivity of the EASL Guideline was significantly lower than other imaging criteria (all p <0.05), but no significant difference was obtained from the diagnostic specificity (all p >0.05). For the tumor size of 20-30 mm (Table 3), the EASL Guideline achieved the highest diagnostic specificity, which was significant higher than a-MRI-I (p = 0.031), but no difference was obtained from other criteria (all p >0.05).

Detection Performance of ECA-MRI
The detailed information about the detection performance of various MR sequences with ECA-MRI are listed in Table 4. The detection performance of PVP and DP were significantly higher than that of T1WI (p <0.001), T2WI (p <0.001) and DWI (p = 0.049), but no difference was obtained from AP (p = 1.000). For the combination use of MR sequence, the "T2WI + DWI" detected 123 (95.35%, 123/129) lesions, and "AP + PVP and/or DP" and LI-RADS v2018 detected all the 129 (100%, 129/129) lesions, and no difference was obtained from the detection performance of these combined sequences (p = 1.000).
Regarding the tumor size ≤20 mm, the diagnostic sensitivity for imaging hallmark of "AP + PVP and/or DP" was significantly lower than LI-RADS v2018 (p = 0.031), but no difference was obtained from the comparison of specificity (all p >0.05). In addition, for the tumor size of 21-30 mm, the diagnostic sensitivity for imaging hallmark of "AP + PVP and/or DP" was significant lower than that of "DWI + T2WI" and LI-RADS v2018 (all p <0.05), but no difference was obtained from specificity (all p >0.05).

DISCUSSION
In the present study, the detection performance of various MR sequences in detecting hepatic nodules in high-risk patients were evaluated and further the diagnostic sensitivity and specificity of different clinical guidelines and aMRI protocols using either   HBSC-MRI or ECA-MRI were compared. The results showed that HBP was most sensitive in detecting hepatic nodules with HBSC-MRI, while AP was most sensitive for conventional ECA-MRI. Regarding the different clinical guidelines and aMRI protocols in diagnosing HCC, in HBSC-MRI group, the use of "DWI + HBP" as a diagnostic criteria showed the highest diagnostic sensitivity, but the EASL criteria showed the lowest diagnostic sensitivity no matter the size of tumor. Nevertheless, the EASL criteria showed the highest diagnostic specificity in diagnosing HCC. Additionally, in the ECA-MRI group, the diagnostic sensitivity of "DWI + T2WI"as diagnostic criteria was higher than that of the conventional imaging hallmark of "AP + PVP and/or DP", but the diagnostic specificity was greatly reduced. Our results found that HBP detected all the hepatic nodules in HBSC-MRI group and followed with TP and DWI, however, for the patients underwent with ECA-MRI, AP showed the highest detection ability and subsequent with PVP/DP and DWI. It is worth noting that compared with HBP, AP alone cannot detect all the hepatic lesions and only when combined with PVP/DP, all the suspected hepatic lesions can be detected with ECA-MRI. Our results were in line with Kim et al. (34) who also found that additional evaluation by MR imaging with gadoxetic acid can led to the detection of additional HCC nodules in 16% of patients. The reason why HBP can detect more hepatic nodules than AP might be explained by that the alteration of hepatic membrane function developed earlier than that of the abnormal hepatic blood supply during the process of carcinogenesis. Another noteworthy issue is that the detection performance of AP with ECA-MRI was obviously higher than that with HBSC-MRI. In this study, AP only detected 72.88% (86/118) of nodules with HBSC-MRI, but about 95.35% (123/129) of hepatic nodules were detected with ECA-MRI, this mainly caused by the high incidence rate of transient severe motion in AP with HBSC-MRI and which may directly reduce the detection performance. Thus, there is still a concern about whether AP is still necessary for these patients underwent with HBSC-MRI for nodule detection.  Previous studies (35)(36)(37) have directly compared the different diagnostic criteria for the diagnosis of HCC with HBSC-MRI, Paisant et al. (35) found that when use the histological results as the gold standard, the Korean guideline 2018 showed the highest diagnostic sensitivity of 71.6% compared with that of 67.6% with the Japan Society of Hepatology guideline 2014 and 45.1% of the EASL guideline. Another study conducted by Kim et al. (36) also found that when extended the imaging feature of washout from PVP to HBP, the diagnostic sensitivity increased from 75.3 to 95.2%. Our results also showed that when the diagnostic criteria containing TP/HBP (70.30%), the diagnostic sensitivity has significantly increased compared with the EASL criteria (49.51%). More importantly, our study proposes another more sensitive diagnostic criteria for HCC that is the combined use of "DWI + HBP", the diagnostic sensitive was 93.07, 90.16, and 97.50% for the tumor size of 0-30 mm, ≤20 mm and 21-30 mm, respectively, and which was significantly higher than that of the EASL and Asian guidelines. This can be explained by the fact that as the detection is the first step of tumor diagnosis, compared with other imaging criteria containing AP which often neglect the hepatic nodules because of the transient severe motion, the "DWI + HBP" can maximally ensure the detection of hepatic nodules and thus further improve the diagnostic performance.
Additionally, excepting for the diagnostic sensitivity, the diagnostic specificity is also essential for clinical options. The EASL criteria keeps the highest diagnostic specificity in all tumor size category with HBSC-MRI, and the specificity of "DWI + HBP" was slightly numerical lower than that of the EASL and Asian criteria, but no statistically difference was obtained. However, for the diagnostic specificity of "DWI + HBP", the further stratified analysis results showed that the highest specificity was obtained with the tumor size ≤20 mm and which was equal to the EASL criteria. The unsatisfactory diagnostic specificity was obtained with tumor range of 21-30 mm, as two AMLs, two CCs and 1 pseudo-inflammatory tumor was misdiagnosed as the imaging feature often mimics HCC and which is hard to differentiate. Moreover, the highest Youden index value was obtained by using the criteria of "DWI + HBP" which means that by using this criteria, better screening and greater authenticity was achieved.
Our results found that the diagnostic criteria of "DWI + T2WI" showed high diagnostic sensitivity compared with that of "AP + PVP and/or DP" with ECA-MRI, but the diagnostic specificity of "DWI + T2WI" lower than that of "AP + PVP and/or DP". Considering this situation, for ensuring both diagnostic sensitivity and specificity, we recommend that for patients underwent with ECA-MRI, the conventional imaging hallmark of "AP + PVP and/or DP" was strongly advocated, and for patients only underwent with plain liver MR scanning, the "DWI + T2WI"was recommended because of its high diagnostic sensitivity. In addition, another important question which merit for discussing is that whether the low diagnostic specificity of "DWI + T2WI" means that this diagnostic criteria is of little value in diagnosing HCC. On the contrary, in clinical practice, the T2WI hyperintensity nodule with diffusion restriction just goes to show the possibility of malignant hepatic nodule but not HCC specific, and which can also be used to differentiate RN and DN (8,35,36,38). Thus, the diagnostic criteria of "DWI + T2WI" have good suggestive significance as whether the malignant hepatic nodule is presented, and the further enhanced MR imaging is needed.
This study has several limitations. First, only a limited number of non-HCC hepatic nodules were included in our study, and which may potentially lead to the bias of the diagnostic specificity. However, this reflects the best current clinical practice that most nodules in high-risk patients were HCCs. Second, despite the resected liver specimens and final pathological results were taken as the reference standard; a substantial number of hepatic nodules may be missed as we did not have whole liver pathological evaluation. As one of the purposes of this study was to detect the early-stage disease that meet the BCLC staging system, the partial liver resection is the standard surgery for the enrolled patients. Therefore, the high detection rates in this study represent relative, but not absolute. However, strictly hepatic nodule detection was made by using the resected specimens, and the detected nodules on liver specimens were matched with the MR images. Thus, to a certain extent, the results of high detection rate have certain representation. Third, the head-to-head comparison of the extracellular contrast agents and hepatobiliary contrast agents in the same patient cohort was not conducted as the patients cannot underwent both MR scanning in a short period. Previous study conducted by Paisant et al. (35) has directly compared extracellular and hepatobiliary MR contrast agents for the diagnosis of small HCCs and found that hepatobiliary MR outperforms extracellular contrast agents in diagnosing small HCCs. Thus, large sample, multi-center directly comparison and validation of the two contrast agents are needed in future.

Conclusions
In conclusion, for patients with HBSC-MRI, our new diagnostic criteria, based on the evaluation of HBP and DWI, demonstrated a higher lesion detection rate, and a significantly higher diagnostic sensitivity compared with that of the EASL and Asian guidelines and a specificity slightly but not significantly lower than that of EASL and Asian guidelines for HCC. Our study also shows that for patients with ECA-MRI, the conventional diagnostic hallmark showed the highest detection and diagnostic performance, suggesting that the application of diagnostic criteria for HCC need to be differentiated in patients using HBSC-MRI and those with ECA-MRI.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the West China Hospital, Sichuan University. The patients/participants provided their written informed consent to participate in this study.