Objective: To retrospectively evaluate the diagnostic performance of contrast-enhanced ultrasound (CEUS) LI-RADS in liver nodules < 20 mm at high risk of hepatocellular carcinoma (HCC) and their correlation with clinic-pathological features. Methods: A total of 432 pathologically proved liver nodules < 20 mm were included from January 2019 to June 2022. Each nodule was categorized as LI-RADS grade (LR)-1 to LR-5 through LR-M according to CEUS LI-RADS. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC) of CEUS LI-RADS were evaluated using pathological reference standard. Correlations between clinic-pathological features and CEUS LI-RADS categorization, together with major CEUS features, were further explored. Results: With LR-5 to diagnose HCC, the sensitivity, specificity, PPV, NPV, and AUC were 50.3%, 70.0%, 91.2%, 18.5%, and 0.601, respectively. The proportion of LR-5 in primary HCCs was significantly higher than that in recurrent ones (p = 0.014). HCC 10–19 mm showed significantly more frequent arterial phase hyper-enhancement (APHE) and late washout (p < 0.05) and less no-washout (p = 0.003) compared with those in HCC < 10 mm. Well-differentiated HCCs showed more frequent non-APHE and no-washout than moderate- and poor-differentiated HCCs (p < 0.05). Upgrading "APHE without washout" LR-4 nodules 10–19 mm with HCC history and "APHE with late mild washout" LR-4 nodules < 10 mm to LR-5 could improve the diagnostic performance of LR-5. The corresponding sensitivity, specificity, PPV, NPV, and AUC are 60.2%, 70.0%, 92.6%, 22.1%, and 0.651, respectively. Conclusions: CEUS LI-RADS is valuable in the diagnosis of HCC < 20 mm and performance can be improved with the combination of clinic-pathological features. Critical relevance statement: CEUS LI-RADS was valuable in the diagnosis of HCC < 20 mm and its diagnostic performance can be improved by combining clinic-pathological features. Further research is needed to define its value in this set of lesions. Key Points: Contrast-enhanced ultrasound can detect small liver lesions where LI-RADS accuracy is uncertain. Many LI-RADS Grade-4 nodules were upgraded to Grade-5 by combining imaging with clinic-pathological factors. The reclassification of LI-RADS Grade-5 can improve sensitivity without decreasing positive predictive value.
Keywords: Hepatocellular carcinoma; Ultrasound; Contrast media; Diagnosis
Daohui Yang, Xuejun Chen and Linjin Huang contributed equally to this study. As a result, they are listed as joint first authors of this article.
Graph
Hepatocellular carcinoma (HCC), which constitutes the most common primary liver malignancy, accounting for approximately 80%, is the sixth most diagnosed cancer and the fourth leading cause of cancer death worldwide [[
During routine screening for patients at high risk of HCC, together with the improvements in image modalities, liver nodules are prone to be detected at an earlier stage with relatively small size. However, the imaging of small HCC, especially those < 20 mm, is challenging because of the complexity of blood supply during multistage hepatic carcinogenesis, which leads to overlapping microvascular perfusion of regenerative and dysplastic nodules and HCC. Depending on the balance of remnant portal venules and new dysplastic arteries, small nodules may not demonstrate the classic HCC enhancement characteristics on CT or MRI, that is, arterial hyper-enhancement followed by washout in the portal-venous and delayed phase. The diagnostic performance of CEUS LI-RADS in characterizing small liver nodules (< 20 mm), to our best knowledge, has not been fully evaluated. The purpose of our study was to retrospectively evaluate the diagnostic performance of CEUS LI-RADS in liver nodules < 20 mm at high risk of HCC and to explore the correlation between clinic-pathological features and the CEUS LI-RADS categorization for further improving its diagnostic performance.
The implementation of this retrospective observational study was approved by our institutional ethics committee (No. B2022-347R) and written informed consent was waived for its retrospective nature. Fig. 1 shows the flow chart of patient recruitment.
Graph: Fig. 1 The flow chart of patient recruitment
From January 2019 to June 2022, a total of 2248 consecutive patients with liver nodules underwent CEUS examination. Inclusion criteria were (
Conventional ultrasound and CEUS examinations were performed by one of three board-certified radiologists with more than 10 years' experience in abdominal US and CEUS (X.W., H.H., and Q.L., respectively) with GE Logiq E9 (GE Healthcare, Berlin, Germany), Phillip Epiq 7 (Philips Medical Solutions; Mountain View, CA, United States), or Hitachi-Aloka Preirus Arietta 70 (Tokyo, Japan). As a routine procedure of liver nodule examination, on gray-scale ultrasound, the location, number, and maximum diameter of nodules were recorded. The pulse inversion harmonic imaging was applied to CEUS examinations with a mechanical index < 0.1. A bolus injection of 1.2–2.4 mL sulfur hexafluoride-filled contrast agent (SonoVue; Bracco, Milan, Italy) was injected via an antecubital 20-gauge catheter followed by a 5-mL flush of 0.9% sodium chloride solution. A timer was started immediately at the end of the contrast agent injection, and each CEUS examination lasted at least 2 min, with recording arterial phase (10–30 s), portal-venous phase (30–120 s), and delayed phase (> 120 s) in digital video format for further evaluation.
The CEUS images were assessed by two experienced radiologists with 5 years (D.Y.) and 8 years' experience (X.C.) in liver CEUS interpretation in consensus, they were both blinded to the final pathology and clinical data. According to the American College of Radiology (ACR) CEUS LI-RADS v2017, the categorization was based on the arterial enhancement, onset time of washout, and washout degree at 120 s after contrast agent injection. In the arterial phase, the enhancement pattern was classified as follows: homogeneous or heterogeneous hyper-enhancement, rim-like hyper-enhancement, iso-enhancement, and hypo-enhancement. Both homogeneous and heterogeneous hyper-enhancement were referred to as arterial phase hyper-enhancement (APHE), and rim-like hyper-enhancement was referred to rim-APHE. As for washout, it was applied only in those with hyper-enhancement or iso-enhancement during the arterial phase. The degree of washout was categorized into three types: no washout, mild washout (a reduction in enhancement degree but continues to show some microbubbles within the nodule), and marked washout, compared with the surrounding liver parenchyma. According to the onset time of washout, early and late washout were defined with 60 s as the cut-off value. The above-mentioned CEUS features, together with lesion size, were used to give each lesion a CEUS LI-RADS categorization [[
The reference standard for the final diagnosis was histopathological findings, either by surgical resection (424/432) or by US-guided percutaneous biopsy (8/432). Surgical or biopsy specimens were assessed for hepatic background using the Scheuer fibrosis stage and for HCC grades using Edmondson-Steiner grades, which were routinely recorded on the pathological reports.
Statistical analysis was performed with IBM SPSS Statistics version 22.0 (Armonk, NY, USA; IBM Corp) and MedCalc version 19.0.4 (MedCalc Software, Ltd, Ostend, Belgium). Qualitative data were presented as numbers and percentages, and quantitative data were presented as mean ± standard deviation. The chi-square test and Fisher's exact test were used for the comparisons of categorical data. Student's t-test for unpaired samples was used for comparison of mean values. For the performance of CEUS LR-5 and LR-M in the diagnosis of HCC and non-HCC malignancies, the sensitivity, specificity, positive predictive values (PPV), negative predictive values (NPV), and area under the curve (AUC) were calculated, respectively. Since CEUS was an imaging modality associated with certain observer experience and subjectivity, the Kappa value was calculated to measure the inter-observer agreement of CEUS LI-RADS categorization. The strength of agreement was interpreted as follows: 0–0.20, poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81–1.00, almost perfect agreement. Differences were considered statistically significant for p < 0.05. To better understand the clinic-pathological features that might influence CEUS LI-RADS diagnostic performance on HCCs, the correlations between HCC grade, fibrotic stage of liver parenchyma, CEUS LI-RADS categorization, and major CEUS features were explored, respectively.
In total, 432 small solitary liver nodules < 20 mm in 432 patients were included in this retrospective study with a mean size 14.7 ± 3.5 mm (range 5–19 mm). Table 1 lists the demographic features of patients and liver nodules.
Table 1 Subject demographic characteristics
Variable Value (%) No. of men/women 350 (81.0%)/82 (19.0%) Median age (years)* 58 (57.4 ± 9.7) Size < 10 mm 50 (11.6%) 10–20 mm 382 (88.4%) Recurrence Yes 153 (35.4%) No 279 (64.6%) Etiology HBV 431 (99.8%) HCV 10 (2.3%) Parenchymal background Fibrosis 401 (92.8%) Cirrhosis 264 (61.1%) Pathological diagnosis HCC 372 (86.1%) ICC 8 (1.9%) CHC 20 (4.6%) MLT 6 (1.4%) Other malignancies 4 (0.9%) DN 3 (0.7%) RN 1 (0.2%) FNH 5 (1.2%) Other benign nodules 13 (3.0%) Differentiation of HCC Well-differentiated 6 (1.6%) Moderate-differentiated 290 (78.0%) Poor-differentiated 76 (20.4%)
Except where indicated, the data are the number of nodules, and the number in brackets is the percentage, *data are mean. ICC intrahepatic cholangiocarcinoma, MTL metastatic liver tumors, CHC combined hepatocellular cholangiocarcinoma, DN dysplastic nodule, RN regenerative nodule, FNH focal nodular hyperplasia
Inter-observer agreement according to Cohen's Kappa was excellent on CEUS LI-RADS with the Kappa value of 0.830 (confidence interval [CI]: 0.788, 0.872). The results are summarized in Table 2.
Table 2 Inter-observer agreement in CEUS LI-RADS categorization
Reader 1 Reader 2 Total LR-2 LR-3 LR-4 LR-5 LR-M 5 1 0 0 0 6 0 30 5 3 0 38 0 3 65 2 4 74 0 1 3 191 39 234 0 0 0 1 79 80 5 35 73 197 122 432
The distribution of the CEUS LI-RADS categorizations and pathological results are summarized in Table 3. The distribution of HCC was 0.0%, 1.1%, 8.1%, 16.4%, 50.2%, and 24.2% in CEUS LR-1, 2, 3, 4, 5, and M, respectively.
Table 3 The distribution of CEUS LI-RADS categorization and pathological diagnosis
Pathological diagnosis LR-2 LR-3 LR-4 LR-5 LR-M Total HCC 4 30 61 187 90 372 ICC 0 1 1 3 3 8 MLT 0 0 1 0 5 6 CHC 0 0 3 6 11 20 DN 0 1 1 1 0 3 RN 0 1 0 0 0 1 FNH 0 1 4 0 0 5 Other malignant nodules 0 0 0 1 3 4 Other benign nodules 1 1 1 7 3 13 Total 5 35 72 205 115 432
CEUS LI-RADS, contrast-enhanced ultrasound Liver Imaging Reporting and Data System
LR-5 category included 187 (91.2%) HCCs (Fig. 2), 10 non-HCC malignancies (3 intrahepatic cholangiocarcinoma, 6 combined hepatocellular cholangiocarcinoma, 1 primary hepatic lymphoma) (4.9%), and 8 benign nodules (1 dysplastic nodule, 3 epithelioid angiomyolipomas, 2 focal inflammatory lesions, 1 hepatocellular adenoma and 1 intrahepatic bile duct adenoma) (3.9%). With LR-5 to diagnose HCC, the sensitivity, specificity, PPV, NPV, and AUC were 50.3% (95%CI: 45.1%, 55.5%), 70.0% (95%CI: 56.8%, 81.2%), 91.2% (95%CI: 87.4%, 93.9%), 18.5% (95%CI: 15.7%, 21.6%), and 0.601 (95%CI: 0.553, 0.648), respectively.
Graph: Fig. 2 Images showed a nodule classified as LR-5 in a 62-year-old male with chronic hepatitis B virus-related cirrhosis and pathologically confirmed HCC. A 13 mm nodule located in the right lobe of the liver showed homogeneous hyperenhancement (arrow) in the arterial phase on contrast-enhanced ultrasound (a). A mild washout (arrow) was observed in the portal-venous phase (> 60 seconds; timer, 01:36) (b). The lesion was classified as LR-5 according to the CEUS LI-RADS.
LR-M included 22 (19.1%) non-HCC malignancies, 90 (78.3%) HCCs, and 3 (2.6%) benign nodules (focal inflammatory lesions). With LR-M for the diagnosis of non-HCC malignancy, the sensitivity, specificity, PPV, NPV, and AUC were 57.9% (95%CI: 40.8%, 73.7%), 76.4% (95%CI: 71.9%, 80.5%), 19.1% (95%CI: 14.6%, 24.6%), 95.0% (95%CI: 92.8%, 96.5%), and 0.671 (95%CI: 0.625, 0.716), respectively.
The correlations between HCC CEUS LI-RADS categorization and clinic-pathological features are summarized in Table 4.
Table 4 Correlations between HCC CEUS LI-RADS categorization with clinic-pathological features
Category HCC LR-2 LR-3 LR-4 LR-5 LR-M Size < 10 mm ( 4 13 16 0 7 10–19 mm ( 0 17 45 187 83 Recurrence Yes ( 2 15 30 57 32 No ( 2 15 31 130 58 Cirrhosis Presence ( 1 21 42 125 60 Absence ( 3 9 19 62 30 Differentiation Well ( 0 4 1 1 0 Moderate ( 3 22 47 149 69 Poor ( 1 4 13 37 21 Serum AFP level ≥ 20 ng/mL ( 2 9 22 74 38 < 20 ng/mL ( 2 21 39 113 52 Total 4 30 61 187 90
The proportion of LR-5 categorization in primary HCCs was significantly higher than that in recurrent ones (55.1% vs. 41.9%, p = 0.014), and the proportion of LR-4 categorization was significantly higher in recurrent HCCs than that in primary ones (22.1% vs. 13.1%, p = 0.029). LR-M categorization in primary and recurrent HCCs made no significant difference (24.6% vs. 23.5%, p = 0.820).
The distribution of HCC in LR-5 and LR-M was not influenced by the hepatic background, HCC grade, and serum alpha-fetoprotein (AFP) level (all p > 0.05). According to CEUS LI-RADS, tumor size is a subsequent influence factor of categorization. Since the threshold of nodule size for LR-5 was 10 mm, no nodules < 10 mm could be categorized into LR-5. The distribution of HCC < 10 mm in CEUS LR-1, 2, 3, 4, and M was 0.0%, 10.0%, 32.5%, 40.0%, 0.0%, and 17.5%, respectively; and the distribution of HCC 10–19 mm was 0.0%, 0.0%, 5.1%, 13.6%, 56.3%, and 25.0%, respectively.
The correlations between HCC major CEUS LI-RADS features and clinic-pathological features are illustrated in Table 5. HCC 10–19 mm showed a significantly higher frequency of APHE than HCC < 10 mm (92.5% vs. 82.5%, p = 0.034), HCC < 10 mm showed more frequent no-washout and less frequent late washout (37.5% vs. 17.8%, p = 0.003; 45.0% vs. 55.4%, p = 0.036) compared with HCC 10–19 mm. With respect to the influence of HCC grade on CEUS features, the prevalence of non-APHE and no-washout were substantially greater in well-differentiated HCCs than those in moderate- and poor-differentiated HCCs (50.0% vs. 7.9%, p = 0.010; 66.7% vs. 19.1%, p = 0.016). The other clinical-pathological features, including hepatic background, AFP, and recurrence or primary HCC did not correlate with the CEUS LI-RADS major features of HCCs.
Table 5 Correlation between major CEUS LI-RADS features and clinic-pathological features of HCC
Arterial phase Portal and delayed phase Non-APHE APHE Non-WO Early WO Late WO Size < 10 mm ( 7 (17.5%) 33 (82.5%) 15 (37.5%) 7 (17.5%) 18 (45.0%) 10–19 mm ( 25 (7.5%) 307 (92.5%) 59 (17.8%) 89 (26.8%) 184 (55.4%) 0.034 0.012 Recurrence Yes ( 12 (8.9%) 123 (91.1%) 35 (25.9%) 32 (23.7%) 68 (50.4%) No ( 20 (8.4%) 217 (91.6%) 40 (16.9%) 64 (27.0%) 133 (56.1%) 0.882 0.154 Differentiation Well ( 3 (50.0%) 3 (50.0%) 4 (66.7%) 0 (0.0%) 2 (33.3%) Moderate ( 20 (6.9%) 270 (93.1%) 55 (19.0%) 75 (25.8%) 160 (55.2%) Poor ( 9 (11.8%) 67 (88.2%) 15 (19.8%) 21 (27.6%) 40 (52.6%) 0.004 0.126 Cirrhosis Yes ( 19 (7.6%) 230 (92.4%) 49 (19.7%) 65 (26.1%) 135 (54.2%) No ( 13 (10.6%) 110 (89.4%) 25 (20.3%) 31 (25.2%) 67 (54.5%) 0.342 0.978 Serum AFP level ≥ 20 ng/mL ( 11 (7.4%) 137 (92.6%) 22 (14.9%) 41 (27.7%) 85 (57.4%) < 20 ng/mL ( 21 (9.4%) 203 (90.6%) 52 (23.2%) 55 (24.6%) 117 (52.2%) 0.513 0.142
APHE arterial phase hyper-enhancement, WO washout
With upgrading LR-4 nodules < 10 mm with "APHE and late mild washout" to LR-5 (n = 19) (Fig. 3), 16/19 HCC might be correctly categorized into LR-5. CEUS LI-RADS categorization with upgrading LR-4 nodules 10–19 mm with HCC history and "APHE with no washout" to LR-5 (n = 22) (Fig. 4), 22/22 HCCs might be correctly categorized into LR-5. Then the diagnostic performance of LR-5 in the diagnosis of HCC would be improved, with sensitivity, specificity, PPV, NPV, and AUC being 60.2% (95%CI: 55.3%, 65.5%), 70.0% (95%CI: 56.8%, 81.2%), 92.6% (95%CI: 89.4%, 94.9), 22.1% (95%CI: 18.8%, 26.0%), and 0.651 (95%CI: 0.596, 0.698), respectively. The AUC (0.651) of LR-5 in modified CEUS LI-RADS was significantly higher than that (0.601) in CEUS LI-RADS (p < 0.001) (Fig. 5).
Graph: Fig. 3 Images in a 61-year-old male with chronic hepatitis B virus-related cirrhosis and HCC history. A 9 mm nodule in the right lobe of the liver showed homogeneous hyperenhancement in the arterial phase on CEUS (a). A mild washout was observed in the portal-venous phase (> 60 seconds; timer, 01:33) (b). The lesion was classified as LR-4 according to the CEUS LI-RADS. While this lesion was recategorized to LR-5 according to the features of "APHE and late mild washout" and it was finally confirmed as hepatocellular carcinoma by histopathology
Graph: Fig. 4 A nodule in a 42-year-old male with an HCC history was classified as LR-4 according to the CEUS LI-RADS. A 19 mm nodule located in the right lobe of the liver was homogeneous hyperenhanced in the arterial phase on CEUS (a). Washout was not observed during the portal-venous phase to the late phase (> 120 seconds) (b). We upgraded this lesion that showed "APHE with no washout" with HCC history from LR-4 to LR-5, and this lesion was finally confirmed as hepatocellular carcinoma by histopathology.
Graph: Fig. 5 The compassion of area under the receiver operating characteristic (ROC) curve of CEUS LI-RADS (blue line) and modified CEUS LI-RADS (green line) for diagnosing HCC < 20 mm. DeLong's tests show that the p value between the ROC curve of CEUS LI-RADS and modified CEUS LI-RADS was less than 0.001.
With the decrease in liver nodule diameter, the diagnostic sensitivity of imaging techniques would correspondingly reduce from 79–97% in HCC ≥ 20 mm [[
In present study, the sensitivity and specificity of LR-5 for HCC diagnosis were 50.3% and 70.0%, respectively, which was significantly lower than that in Peng's meta-analysis for nodules < 20 mm (sensitivity 0.70 [95%CI 0.64–0.76], specificity 0.91 [95%CI 0.87–0.95]) [[
In the study by Terzi et al [[
LR-M was designed to preserve the specificity for the diagnosis of HCC while not losing sensitivity for the diagnosis of malignancy. In this study, 115 nodules were categorized in CEUS LR-M, which had a high proportion of HCC (78.3%). With LR-M to diagnose non-HCC malignancies, the sensitivity and specificity were 57.9 and 75.8%, respectively. Other studies showed that 25–63% of LR-M lesions were HCC with various nodule size range and percentages of ICC and CHC in each study [[
Our study also had some limitations. First, due to the topic focusing on small liver nodules (< 20 mm) and its retrospective nature with single-center design, our result is placed at a lower level in the hierarchy of evidence-based medicine, and the relatively small sample size of nodules that upgraded from LR-4 to LR-5 can introduce a bias and limited the generalization of our results. Second, using pathology as a reference standard exclusively might introduce verification bias because lesions receiving pathological assessment are more inclined to be malignant and be assigned to higher CEUS LI-RADS categories. However, choosing a pathological reference standard seems the best way to enable an objective assessment of diagnostic accuracy compared with CCRS. Third, chronic HBV infection accounted for the major etiology of HCC in the present study. Thus, our results may not be reproducible in patients with other etiologic causes (non-alcoholic steatohepatitis) or the Western population (HCV infection). Further prospective multicenter studies should be investigated in the future. Fourth, there were seven patients with poor CEUS imaging quality were excluded from the present study, which would introduce a bias and a strong limitation to clinical practice.
In conclusion, the present data supported the value of CEUS LR-5 for HCC < 20 mm. With the combination of clinic-pathological features, the diagnostic performance of CEUS LI-RADS in HCC < 20 mm can be improved.
The authors really appreciate all staff members and colleagues in the Ultrasound department at Zhongshan Hospital (Xiamen), Fudan University and Zhongshan Hospital, Fudan University.
DY: investigation, methodology, project administration, conceptualization; data curation, formal analysis, writing—original draft. XC: conceptualization, formal analysis, investigation, project administration. LH: investigation, methodology, formal analysis. XW: software, resources, data curation. LM: formal analysis, investigation, project administration. LL: resources, investigation, pathologic consultation. HH: supervision, validation. QL: conceptualization, supervision, validation, project administration, writing—review & editing. All authors read and approved the final manuscript.
This work was supported by the Natural Science Foundation of Fujian Province, China (Grant No. 2022J011420); the Xiamen Medical and Health Guidance Project, China (Grant No. 3502Z20224ZD1090).
The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.
This study was approved by the Institutional Review Board of Zhongshan Hospital, Fudan University (No. B2022-347R), and written informed consent was waived for its retrospective nature.
We obtained consent from that person or legal guardian whose person's data is contained in our manuscript.
The authors declare that they have no competing interests.
• AUC
- Area under the curve
• APHE
- Arterial phase hyper-enhancement
• CCRS
- Composite clinical reference standard
• CEUS
- Contrast-enhanced ultrasound
• HBV
- Hepatitis B viral infection
• HCV
- Hepatitis C viral infection
• HCC
- Hepatocellular carcinoma
• LR
- LI-RADS category
• LI-RADS
- Liver Imaging Reporting and Data System
• NPV
- Negative predictive value
• PPV
- Positive predictive value
Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
By Daohui Yang; Xuejun Chen; Linjin Huang; Xi Wang; Lijuan Mao; Lewu Lin; Hong Han and Qing Lu
Reported by Author; Author; Author; Author; Author; Author; Author; Author