The institutional review board (IRB) of the Samsung Medical Center approved this retrospective study (IRB # 2020-08-084), and the requirement for informed consent was waived.

From January 2013 to December 2020, 386 planning US procedures were performed for patients with suspected hepatic metastasis detected on liver CT or MRI, to determine the feasibility of percutaneous RFA. Among them, 250 patients were excluded for the following reasons: multiple tumors (n=230), metastasis from other cancers (n=10), possible atypical HCC in a cirrhotic liver (n=1), benign disease on percutaneous biopsy, or follow-up CT or MRI (n=6), or a tumor size >3 cm (n=3). Eventually, 136 patients were included with a single hepatic metastasis from colorectal cancer measuring <3 cm. Hepatic metastases were diagnosed based on pathologic proof by percutaneous biopsy (n=2), subsequent hepatic resection (n=16), or typical imaging findings on liver CT or MRI (n=118). A flow diagram of patient selection is shown in Fig. 1.

The mean time interval between CT/MRI and planning US was 11.0±9.8 days. All planning US procedures were performed using US machines (LOGIQ E9 or LOGIQ E10; GE Healthcare, Waukesha, WI, USA) by one of five radiologists (H.K.L., H.R., M.W.L., K.D.S., and T.W.K.) who had experience of more than 100 cases of percutaneous RFA of liver tumors. Before planning US, the radiologists carefully reviewed the CT or MRI scans, and the patients were positioned appropriately for a US examination based on the tumor location. In general, if the tumor was located in the liver dome and the sonographic window was considered to be affected by the lung or rib shadow, the semi-erect position using a tilting table was preferred because the sonographic window is enhanced for hepatic dome lesions in this position [14]. Planning US was performed using fusion imaging (Volume Navigation, LOGIQ E9, or LOGIQ E10) of real-time US and pre-acquired CT/MRI scans [10]. The radiologists searched for metastatic nodules on B-mode US, based on fused CT/MRI scans. The radiologists evaluated and assessed lesion conspicuity and the feasibility of percutaneous RFA. If the lesion was not sufficiently conspicuous, CEUS with gaseous perflubutane (Sonazoid, GE Healthcare, Oslo, Norway) was added to the fusion imaging at the time of planning US [15]. The necessity of CEUS was determined by the attending radiologists who performed the planning US.

When percutaneous RFA was infeasible, the cause of the infeasibility of percutaneous RFA was categorized based on radiologic reports and the stored images of planning US on a picture archiving and communication system (PACS) (Centricity, GE Healthcare, Milwaukee, WI, USA) as follows: (1) an inconspicuous tumor on planning US; (2) a high risk of the heat-sink effect due to the abutment of a large vessel, defined as the portal vein, hepatic vein or inferior vena cava with vessel caliber of 3 mm or larger [16]; (3) a high risk of collateral thermal damage to an adjacent organ due to close proximity; and (4) the absence of a safe electrode path [6].

If the metastasis could not be identified even after CEUS, the lesion was regarded as an invisible tumor. The size of the tumor was based on the maximal diameter on CT/MRI scans, which was measured using an electronic caliper on the PACS. The distance between the upper margin of the tumor and the diaphragm was measured using coronal or sagittal CT/MRI scans.

The univariable associations between individual variables for the visibility of the tumor was tested using the two-sample t test or Wilcoxon rank-sum test for continuous variables (the size of the tumor, distance from the diaphragm, creatinine, age, albumin, total bilirubin, prothrombin time, platelet count, carcinoembryonic antigen, and carbohydrate antigen 19-9) and the chi-square test or Fisher exact test for categorical variables (sex, body mass index, segmental location of the tumor, prior history of hepatic resection or chemotherapy, and presence of liver disease) as appropriate. Multiple logistic regression analysis was performed to test the significance of the factors adjusted for the other variables. Variables with a P-value <0.10 in the univariate analysis were used for multiple logistic regression analysis. For both the univariate and multiple logistic regression analyses, P-values less than 0.05 were considered to indicate a statistically significant difference. R version 3.5.0 (The R Foundation for Statistical Computing, Vienna, Austria) was used for statistical analyses.

A total of 136 patients (87 male and 49 female; mean age±standard deviation, 60.3±11.6 years) constituted the study population (Table 1). The primary cancer was colon cancer in 89 patients and rectal cancer in 47 patients. Most patients (97.8%, 133/136) did not have chronic hepatitis or liver cirrhosis. The median tumor size was 1.3 cm (range, 0.3 to 2.8 cm). The median distance from the diaphragm to the tumor was 1.6 cm (range, 0 to 7.8 cm). Slightly more than half (51.5%; 70/136) of the patients underwent chemotherapy before planning US, and 11.8% (16/136) of the patients had a history of hepatic resection prior to planning US.

Percutaneous RFA was considered infeasible in 24.3% (33/136) of the patients. The causes of the infeasibility of RFA were as follows: a high risk of the heat-sink effect due to the abutment of the large vessel (n=12), inconspicuous tumors on planning US (n=11), a high risk of collateral thermal damage to the adjacent organ (n=8; heart [n=3], central bile duct [n=2], the bare area of the diaphragm [n=2], or gallbladder [n=1]), and the absence of a safe electrode path (n=2).

Of the 136 tumors in 136 patients, 13 lesions were not identified on B-mode US (Fig. 2). Two of those lesions could be visualized after CEUS (Fig. 3). CEUS was not performed in the remaining 11 patients mainly because of a deeply positioned tumor or an insufficient sonographic window. Consequently, 11 of the 136 lesions (8.1%) were considered invisible on planning US. When tumor size was categorized into three groups (≤1 cm, 1.1-2.0 cm, and 2.1-3.0 cm), the visibility of the tumor was different according to tumor size on planning US (P=0.002) (Table 2).

Table 3 compares the characteristics between the visible and invisible tumors on planning US. The distance from the diaphragm to the tumor was closer in the invisible than in the visible tumors (median, 0.7 cm vs. 1.7 cm), but this was not statistically significant in the univariate and multivariate analyses (P=0.099 and P=0.311, respectively). The tumor size was significantly different between the two groups (median, 1.3 cm vs. 0.8 cm; mean, 1.4 cm vs. 0.9 cm) in both the univariate and multivariate analyses (P=0.003 and P=0.018, respectively). The tumor size was the sole significant factor affecting the invisibility of hepatic metastasis on planning US.