Since the caudate lobe is centrally located in the liver and adjacent to the bare area, hepatocellular carcinoma (HCC) arising in the caudate lobe can receive complex arterial blood supply including from the extrahepatic arteries.^1–8 Therefore, transarterial chemoembolization (TACE) can frequently fail to embolize the target HCCs in the caudate lobe completely, mainly because of the difficulty of identifying its feeding arteries and unsuccessful catheterization due to frequent severe angulation of its orifice.^1,3,5,6,9

Recently, cone-beam computed tomography (CBCT) and TACE guidance software including automated tumor-feeder detection (AFD) functionality have been introduced to TACE for HCC.^10–19 CBCT has a pooled sensitivity of 93%, specificity of 80%–97%, and positive predictive value of 89% for detecting arteries supplying the tumor.²⁰ It can also improve the technical success and early tumor response of TACE.^17–19,21 However, the detectability of tumor feeders of HCC in the caudate lobe by AFD and outcomes of TACE are still unknown. The purpose of this study was to evaluate outcomes of TACE using guidance software for HCC in the caudate lobe.

Our institutional review board approved this retrospective study (approval No. 2021-21) and individual patient consent was waived. Written informed consent related to TACE treatment was also obtained from each patient before the procedure.

Patient selection

Between September 2012 and November 2021, transarterial treatment using TACE guidance software was performed for 77 patients with 85 treatment-naïve caudate lobe HCCs. Among them, two tumors in one patient embolized non-selectively without effort to catheterize into the tumor-feeding caudate artery, three ruptured tumors in three patients embolized with gelatin sponge (GS) particles alone, and one huge tumor > 10 cm were excluded. In total, 72 patients with 79 tumors were eligible. The patients’ characteristics are summarized in Table 1.

Table 1 . Patient Characteristics.

Variable	Value
No. of patients	72
Sex
Male	47
Female	25
Mean age (yr)	73.6 ± 8.3 (53–85)
Child-Pugh Class
A	51
B	20
C	1
Etiology
Hepatitis C	43
Hepatitis B	13
Hepatitis B + C	3
Alcohol	3
Unknown	10
ECOG performance status
0	71
1	1
Number of tumors in the caudate lobe
Single	68
2	3
3	0
4	1
Mean tumor size (mm)	18.6 ± 9.9 (6–53)
Subsegment of the caudate lobe
Spiegel lobe	30
Paracaval portion	38
Caudate process	11
Number of tumors in other sites
None	24
1–5	33
5–10	7
≥ 11	8
Prior treatment
None	27
Conventional TACE alone	37
Hepatectomy plus TACE	6
Hepatectomy plus HAIC plus TACE	1
RFA plus TACE	1
Mean AFP level (ng/mL)	527.0 ± 1,866.1 (2–8,240)
Mean PIVKA-II level (mAU/mL)	412.3 ± 1,217.0 (10–7,685)

Values are presented as number only or mean ± standard deviation (range).

ECOG, European Cooperative Oncology Group; HAIC, hepatic arterial infusion chemotherapy; RFA, radiofrequency ablation; TACE, transarterial chemoembolization; AFP, alpha fetoprotein; PIVKA-II, protein induced by absence of vitamin K or antagonist-II.

The diagnosis of HCC was established by imaging findings of CT and/or magnetic resonance imaging (MRI) referencing the serum levels of tumor markers. Histological confirmation was not obtained in any patient. Sixty-eight patients had a single HCC, three had two, and one had four in the caudate lobe.

Twenty-seven (37.5%) patients had naïve HCCs and 45 (62.5%) had a history of TACE (1–11 times, mean: 3.3 ± 2.4) plus other treatments for HCCs. Twenty-four (33.3%) patients only had a single HCC in the caudate lobe. Forty-eight (66.7%) had HCC(s) other than in the caudate lobe.

CBCT and angiography protocol

All CBCT images were obtained with a C-arm angiographic unit (Allura Xper FD20 or AlluraClarity Xper FD20; Philips Healthcare, Best, The Netherlands) with the technique described previously.^{8,10,11,16–19,21}

After CBCT during arterioportography (CBCTAP) and digital subtraction angiography (DSA) of superior mesenteric and celiac arteries, DSA of a common/proper hepatic artery was conducted using a 4-Fr twist catheter (Hanako Medical, Saitama, Japan). Subsequently, dual-phase CBCT during hepatic arteriography (CBCTHA) was performed by injection of 24 mL of contrast material at a rate of 2 mL/s. The 1st scan began 7 seconds after the beginning of the injection of contrast material, and the 2nd scan began 30 seconds after the end of the 1st scan. If DSA of the common/proper hepatic artery using a 4-Fr catheter was not performed because of anatomical difficulties, dual-phase CBCTHA at the proper or left/right hepatic artery was carried out using a microcatheter (Progreat ∑ [1.9-Fr tip]; Terumo Clinical Supply, Kakamigahara, Japan; Asahi Veloute [1.7-Fr tip]; Asahi Intecc, Seto, Japan; or Asahi Veloute Ultra [1.5-Fr tip]; Asahi Intecc) by injecting 12 mL of contrast material at a rate of 1 mL/s.

Identification of tumor feeders of HCC in the caudate lobe

TACE guidance software (EmboGuide; Philips Healthcare) has been routinely used at a workstation (XtraVision Interventional Workstation; Philips Healthcare) with techniques described previously.^8,16–18 First, a virtual target lesion including a safety margin (approximately 5-mm wide for tumors < 25 mm, and 10-mm wide for tumors ≥ 25 mm around the tumor) was created on HCC on the 1st-phase CBCTHA images. After deciding on the start position of vessel tracking using the “Set Catheter” functionality, the “Auto Detect” button was pressed. Then, all potential tumor feeders of all tumors were automatically highlighted within a few seconds on CBCTHA images and a 3D arteriogram. Prototype TACE guidance software (EmboGuide App; Philips Healthcare) has also been used since May 2018.^19,21 In this software, the start position was automatically set on the catheter tip and tumor feeders were highlighted instantaneously during creation of the target lesion. The AFD process was usually completed within 2 minutes.

If AFD failed to detect the tumor-feeder, reanalysis was performed after changing the target lesion size. When the target lesion was attached to the large arterial branch, vessel tracking was ended at the attached portion. In such a situation, the target lesion was reduced in size to separate it from the large arterial branches. When AFD could not detect appropriate tumor feeders despite several efforts, the branch in the vicinity of the target lesion was determined as a tumor-feeder and it was clicked on a 3D arteriogram or CBCTHA image. Then, the vessel route was highlighted on the display (manual tumor-feeder detection). This process was the same for both forms of software.

TACE procedure

When a microcatheter was advanced into the tumor-feeder, 0.5 mL of 2% lidocaine (Terumo, Tokyo, Japan) was injected through the microcatheter to prevent pain and vasospasm. Then, a mixture of 2–10 mL of iodized oil (Lipiodol 480; Guerbet Japan, Tokyo, Japan) and chemotherapeutics (10–30 mg of epirubicin; [Farmorbicin; Pfizer, Tokyo, Japan] plus 2–6 mg of mitomycin C [Mitomycin; Kyowa Hakko Kirin, Tokyo, Japan]), 10–30 mg of epirubicin, or 50–100 mg of cisplatin (IA Call; Nippon Kayaku, Tokyo, Japan) was slowly injected, followed by an approximately 0.2–0.5-mm diameter GS slurry created from 1-mm-diameter GS particles (Gelpart; Nippon Kayaku) by a pumping method. The ratio of chemotherapeutic solution to iodized oil was 1:3 for epirubicin ± mitomycin C and 1:2 for cisplatin. The total amount of iodized oil was almost equal to the sum of the diameters of the target tumors.

When the branch detected by AFD was determined to be a true tumor-feeder, it was embolized without selective DSA or CBCTHA. The right inferior phrenic artery (RIPA) was also embolized, if necessary, without CBCT and AFD analysis. TACE was finished when all tumor stains disappeared and the tumor feeders were occluded on DSA. CBCT after TACE (LipCBCT) was not performed until May 2018 when accumulation of iodized oil in the target tumor was identified on fluoroscopy. Since May 2018, LipCBCT has been routinely performed at the end of the procedure. When LipCBCT showed incomplete TACE, TACE was added if additional tumor feeders could be identified by manual feeder detection or DSA.

Follow-up

Unenhanced CT was performed 1 week after TACE to check for iodized oil distribution. Dynamic CT and/or MRI was performed every 2–4 months after TACE. When the tumor recurred, additional treatment was performed according to the patient and tumor conditions, if possible.

Assessment

The tumor location was divided into three subsegments according to Kumon’s classification²² and the tumor diameter of each subsegment was compared by one-factor analysis of variance. When iodized oil accumulation in the tumor and surrounding liver was demonstrated on fluoroscopy and/or LipCBCT, the branch was determined as a tumor-feeder. Although manual feeder detection was used during TACE, the results of AFD were only analyzed at the subsubsegmental hepatic artery level. When the tracing of the tumor-feeder was stopped at a more proximal level, the feeder was judged as “not detected.” When AFD mis-traced the origins of the feeder, but it was easily identified on a 3D arteriogram without additional DSA, the feeder was judged as “detected”. If additional DSA was required to identify the correct origin, the feeder was judged as “not detected”.

The origins of tumor feeders were classified into right proximal (R₁), right distal (R₂), left proximal (L₁), left distal (L₂), anterior (A), posterior (P), middle (M), proper hepatic (Ph), common hepatic (Ch), and extrahepatic (Ex) according to the previous definitions (Fig. 1).⁵

Figure 1. The definition of the origins of tumor-feeding caudate arteries. R₁ (right proximal), arising from the right hepatic artery (RHA) between its origin and the middle hepatic artery (MHA) bifurcation, or the cystic artery bifurcation if MHA was not present; R₂ (right distal), arising from RHA between MHA (or cystic artery) bifurcation and the bifurcation of the anterior and posterior segmental artery of RHA. If both MHA and the cystic artery were not present, RHA was divided into the two equal parts; L₁ (left proximal), arising from the left hepatic artery (LHA) between its origin and the medial segmental artery (A4) bifurcation; L₂ (left distal), arising from LHA between A4 bifurcation and the umbilical portion of LHA. If A4 did not arise from LHA, LHA was divided into the two equal parts; A, arising from the anterior segmental artery of RHA; P, arising from the posterior segmental artery of RHA; M, arising from MHA or A4; Ph, arising from the proper hepatic artery; Ch, arising from the common hepatic artery; Ex, arising from the extrahepatic vessels; CHA, common hepatic artery; GDA, gastroduodenal artery.

The embolized area was defined as the area where iodized oil was retained on unenhanced CT performed 1 week after TACE. The images were evaluated in 3 dimensions (axial, coronal, and sagittal views) on the image viewer (ShadeQuest/Report; Fujifilm, Tokyo, Japan). The technical success of TACE was classified into three grades: 1) grade A, the entire tumor was embolized with a safety margin (at least 5 mm for tumors < 25 mm and 10 mm for tumors ≥ 25 mm); 2) grade B, the entire tumor was embolized but the safety margin was lacking in parts; and 3) grade C, the entire tumor was not embolized.¹¹

Early tumor response was evaluated on the first follow-up dynamic CT or MRI after TACE using the modified Response Evaluation Criteria in Solid Tumors (mRECIST). Local tumor progression (LTP) was also judged by follow-up CT or MRI. The rates of tumor-feeder detection by AFD, grade A success of TACE, complete response (CR) at 2–4 months, and durable CR were compared by the Fisher’s exact test and post-hoc analysis using the Bonferroni's multiple comparison test, and the cumulative LTP rates were calculated by the Kaplan-Meier method and compared by the log-rank test in each subsegment. LTP rates between the grade A and B tumors were also compared by the Fisher’s exact test. Statistical calculations were performed using software (R version 4.1.2, 2021; R Foundation for Statistical Computing, Vienna, Austria) and a P-value less than 0.05 was considered to indicate a significant difference.

The results are summarized in Table 2.

Table 2 . Tumor Characteristics, Tumor-Feeder Detectability, and Technical Success of TACE in Each Subsegment.

Characteristic	Spiegel lobe (n = 30)	Paracaval portion (n = 38)	Caudate process (n = 11)
Mean diameter (mm) (range)	16.8 ± 9.5 (6–48)	19.6 ± 9.7 (6–53)	21.8 ± 9.9 (9–36)
Origins of tumor-feeders	R1, 6; R2, 3; L1, 3; L2, 5; A, 6; P, 3; M, 5; A2, 2; and Ex, 2 (the 3 o’clock artery and RGA)	R1, 3; R2, 6; L1, 2; L2, 6; A, 27; P, 13; M, 11; A2, 1; A3, 1; and A8, 5	R1, 1; R2, 2; A, 5; P, 7; A7, 1; and A6, 1
No. of embolized artery	35	77	17
No. of embolized artery/per tumor	1.3 ± 0.9/tumor	2.0 ± 0.9/tumor	1.5 ± 0.9/tumor
Detectability of tumor-feeders (%)	71.4	94.8	76.5
Technical success of TACE (%)
Grade A	93.3	65.8	63.6
Grade B	0	21.1	18.2
Grade C	6.7	10.5	18.2
CR rate at 2–4 months*	93.1	59.4	80
Durable CR rate*	79.3	34.4	30

Values are presented as mean ± standard deviation or number only.

RGA, right gastric artery; A2, the superior lateral subsegmental artery of the left hepatic artery; A3, the inferior lateral subsegmental artery of the left hepatic artery; A8, the anterior superior subsegmental artery of the right hepatic artery; A7, the posterior superior subsegmental artery of the right hepatic artery; A6, the posterior inferior subsegmental artery of the right hepatic artery; CR, complete response.

*CR rates were evaluated in 29 tumors in the Spiegel lobe, 32 in the paracaval portion, and 10 in the caudate process.

Tumor characteristics

The mean tumor diameter in the caudate lobe was 18.6 ± 9.9 mm (range, 6–53 mm). Thirty (38.0%) tumors were located in the Spiegel lobe (SP) (Fig. 2A, 3A), 38 (48.1%) in the paracaval portion (PC) (Fig. 4A), and 11 (13.9%) in the caudate process (CP). When a tumor was located between two subsegments, it was classified based on the subsegment in which it was dominantly located. There were no significant differences in the tumor diameter among the three subsegments (P = 0.6934). Fifty-seven (72.2%) tumors were depicted on DSA (Fig. 3B, 4B), but 22 (27.8%) were not (Fig. 2B). All tumors could be depicted on CBCT (Fig. 2A).

Figure 2. Successful identification of tumor feeders of hepatocellular carcinoma (HCC) in the Spiegel lobe (SP). (A) Cone-beam computed tomography (CBCT) studies showed well-differentiated HCC of 23 mm in diameter (arrows) in SP. Approximately 70% of tumor portions showed hypervascularity. (B) Common hepatic arteriogram showed no tumor stains. The caudate artery (A1) was also unclear. The superior lateral subsegmental artery of the left hepatic artery arose from the left gastric artery (not shown). (C) Automated tumor-feeder detection (AFD) identified two tumor feeders arising from the medial subsegmental artery (A4). (D) Selective arteriogram of A4 showed A1 (arrow). (E) Selective arteriogram of A1 showed equivocal tumor staining (arrow). (F) Then two branches were subsequently embolized. The arrows indicate the tumor. Note that AFD mis-traced one tumor-feeder (pink in Fig. 1C). (G) Unenhanced CT performed 1 week after TACE showed dense iodized oil accumulation throughout almost the entire tumor with a sufficient safety margin. Arterial-phase CT performed 3 months after TACE showed the disappearance of SP, and the tumor has remained well-controlled for 5 years and 7 months after TACE. Reprinted from the article of Miyayama et al (Cancers [Basel]. 2021;13:6370).²¹ CBCTAP, CBCT during arterioportography; CBCTHA, CBCT during hepatic arteriography.

Figure 3. Unsuccessful identification of a tumor-feeder of hepatocellular carcinoma (HCC) in the Spiegel lobe (SP). (A) Hepatobiliary phase of gadoxetate disodium-enhanced magnetic resonance imaging showed HCC of 27 mm in diameter in SP (arrow). (B) Celiac arteriogram showed tumor staining (arrow); however, the tumor-feeder could not be identified. (C) Automated tumor-feeder detection identified the tumor-feeder arising from the common hepatic artery; however, it was considered a pseudo-feeder that was created by artifacts from contrast material. (D) Right hepatic arteriogram showed the tumor staining (arrow) and tumor-feeding caudate artery arising from the distal right hepatic artery (arrowhead). (E) The feeder was selectively embolized until the portal vein was opacified with iodized oil. The arrow indicates the tumor, and the arrowhead indicates the previously embolized tumor. (F) Unenhanced CT performed 1 week after TACE showed dense iodized oil accumulation in the tumor. (G) Arterial-phase CT performed 7 months after TACE showed that the tumor has remained well-controlled.

Figure 4. Grade B technical success of TACE despite successful identification of tumor feeders of hepatocellular carcinoma (HCC) in the paracaval portion (PC). (A) Arterial-phase computed tomography (CT) showed HCC of 20 mm in diameter in PC (arrow). A small tumor was also revealed in segment 5 (S5) (not shown). (B) Common hepatic arteriogram showed tumor staining (arrow). (C) Automated tumor-feeder detection could identify two tumor feeders arising from the anterior segmental artery of the right hepatic artery and middle hepatic artery. The feeder supplying the small tumor in S5 was also identified. (D) The tumor feeders were subsequently embolized. The arrows indicate the tumor. (E) Unenhanced CT performed 1 week after TACE showed that the entire tumor was embolized but the safety margin was not obtained in some tumor parts. A small amount of right pleural effusion was also seen. (F) Arterial-phase CT performed 2 months after TACE showed complete tumor response. (G) However, arterial-phase CT performed 17 months after TACE showed the recurrent tumor (arrow) adjacent to the iodized oil accumulated tumor (arrowhead).

Detectability of tumor feeders by AFD

In total, 129 arterial branches arising from the common hepatic artery were embolized for HCCs in the caudate lobe. Among them, AFD missed 16 tumor feeders. Additionally, the origins were mis-traced in four tumor feeders, and two feeders were judged as “detected” (Fig. 2C–2E) and the remaining two were judged as “not detected”. In total, 111 (86.0%) tumor feeders were detected by AFD. Moreover, 16 additional branches were misdiagnosed as tumor feeders.

Of 30 SP tumors, three small tumors were supplied by the same feeder of the main tumor; therefore, the detectability and origins of tumor feeders were evaluated in 27 HCCs. Among 35 embolized branches, 25 (71.4%) could be detected by AFD (Fig. 2C), but 10 were misdiagnosed (Fig. 3C). Each tumor-feeder arose from various sites, including two from Ex (the 3 o’clock and right gastric arteries) that could not be detected by AFD, although these branches were opacified on CBCTHA. The ratio of the right to left hepatic artery was 18:15 (6:5). Technically, 34 feeders could be catheterized by a microcatheter, but the remaining feeder arising from L₁ could not be catheterized. So, TACE was performed through a microcatheter placed proximal to the orifice of the feeder under blocking the distal flow of the left hepatic artery by a microballoon catheter (Logos; Piolax, Yokohama, Japan) via the additional access route (balloon-assisted TACE⁷).

For 38 PC tumors, 73 of 77 (94.8%) embolized branches could be corrected by AFD (Fig. 4C). Each tumor-feeder arose from various sites and the ratio of the right to left hepatic artery was 54:21 (≒5:2). Technically, three feeders arising from A could not be catheterized; therefore, two were non-selectively embolized and one was embolized by balloon-assisted TACE.

For 11 CP tumors, 13 of 17 (76.5%) embolized branches were correctly detected by AFD. All branches arose from the right hepatic artery at various sites. Technically, one feeder arising from P was embolized non-selectively due to unsuccessful catheterization.

The tumor-feeder detectability of SP tumors was significantly lower than that of PC tumors (P = 0.0035). There were no significant differences in the tumor-feeder detectability between SP and CP tumors (P = 1.0000) or PC and CP tumors (P = 0.1000).

Technical success of TACE

TACE was performed through the feeders that were detected not only by AFD but also by manual feeder detection and/or selective DSA (Fig. 2F, 3D, 3E, 4D). The RIPA was also embolized in seven tumors, four SP and three CP tumors. Of 79 tumors, 60 (75.9%) were classified into grade A (Fig. 2G, 3F), 11 (13.9%) into grade B (Fig. 4E), and eight (10.1%) into grade C. Regarding the tumor location, grade A was achieved in 28 of 30 (93.3%) tumors in SP, 25 of 38 (65.8%) in PC, and seven of 11 (63.6%) in CP. The rate of grade A technical success of TACE of SP tumors was significantly higher than that of PC tumors (P = 0.0240), but it was not significant between PC and CP tumors (P = 1.0000) or SP and CP tumors (P = 0.1650).

In six of eight grade C tumors, additional TACE performed 6–34 months (mean, 12.7 ± 10.6 months) later, showing that three residual tumors were supplied by the neighboring hepatic arteries, and two by the previously embolized arteries, including one that had been embolized non-selectively. The remaining tumor was supplied by the RIPA.

Tumor response

Early tumor response was evaluated in 71 tumors using dynamic CT (n = 57) or MRI (n = 14) performed 2–4 months (mean, 2.6 ± 1.6 months) after TACE, excluding a tumor receiving additional radiofrequency ablation (n = 3), lost to follow-up (n = 4), and sudden death 12 days after TACE due to cardiac infarction (n = 1). CR was noted for 54 (76.1%) tumors, partial response for 12 (16.9%), stable disease for four (5.6%), and progressive disease for one (1.4%). CR was achieved in 27 of 29 (93.1%) tumors in SP (Fig. 2G, 3G), 19 of 32 (59.4%) in PC (Fig. 4F), and eight of 10 (80.0%) in CP. The CR rate in SP tumors was significantly higher than that in PC tumors (P = 0.0081), but there were no significant differences between SP and CP tumors (P = 0.8010) or between PC and CP tumors (P = 0.8581). During a mean follow-up period of 25.6 ± 24.2 months (range, 2–105 months), durable CR was achieved in 37 (52.1%) tumors, in 23 of 29 (79.3%) SP tumors (Fig. 2G, 3G), 11 of 32 (34.4%) PC tumors, and three of 10 (30.0%) CP tumors. The durable CR rate of SP tumors was significantly higher than that of SP and CP tumors (P = 0.0020 and P = 0.0230, respectively), but there were no significant differences between PC and CP tumors (P = 1.0000). Cumulative LTP rates at 6, 12, and 24 months of SP, PC, and CP tumors were 10.3%, 41.5%, and 21.3%; 16.3%, 45.2%, and 43.7%; and 42.6%, 71.9%, and 71.9%, respectively. LTP rates of SP tumors were significantly lower than those of PC tumors (P = 0.0044), but there were no significant differences between SP and CP tumors (P = 0.4580) or PC and CP tumors (P = 0.7454) (Fig. 5).

Figure 5. Graphs of LTP of SP, PC, and CP tumors. Cumulative LTP rates at 6, 12, and 24 months in SP, PC, and CP tumors were 10.3%, 41.5%, and 21.3%; 16.3%, 45.2%, and 43.7%; and 42.6%, 71.9%, and 71.9%, respectively. LTP rates of SP tumors were significantly lower than those of PC tumors (P = 0.0044). There were no significant differences in the LTP rates between SP and CP tumors (P = 0.4580) or PC and CP tumors (P = 0.7454). LTP, local tumor progression; CP, caudate process; PC, paracaval portion; SP, Spiegel lobe.

Regarding the relationship between technical success of TACE and LTP, 18 of 54 (33.3%) grade A tumors locally progressed 3–51 months (mean, 16.2 ± 39.7 months) after TACE, and 36 (66.7%) showed durable CR 2–105 months (mean, 26.2 ± 22.7 months) after TACE. In contrast, nine of 10 (90.0%) grade B tumors locally progressed 2–22 months (7.1 ± 16.8) after TACE, and only one tumor (10%) showed CR 2 months after TACE. LTP rates of the grade A tumors were significantly lower than those of the grade B tumors (P = 0.0012).

Surgical resection for HCC in the caudate lobe is associated with a high mortality rate because of marked blood loss and postoperative complications, in addition to a high tumor recurrence rate.²³ Percutaneous ablation therapy might also be technically difficult because of the deep tumor location and adjacent large vessels.^24–26 Although TACE for HCC in the caudate lobe is also difficult due to complex arterial supply, it has been reported that selective catheterization into the tumor-feeding caudate artery is the most important prognostic factor for a solitary tumor.^4,12 Therefore, identification of a tumor-feeding caudate artery is very important to obtain favorable outcomes.

According to previous reports, AFD can detect 85% to 96% of tumor feeders,^13–19 and 81% to 87% of tumors can be embolized with a sufficient safety margin.^16–19 In the present study, AFD could detect 86% of tumor feeders of HCC in the caudate lobe; however, the detectability differed among the three subsegments. SP is a small area protruding from the right hepatic lobe; therefore, it is difficult to create a virtual target lesion with a sufficient safety margin while separating from the large arteries. Additionally, artifacts from contrast material in the large artery infrequently create “a pseudo-feeder (Fig. 3C).” Moreover, the tumor is infrequently fed by the extrahepatic artery.^3,5 These may reduce the detectability of tumor feeders of SP tumors. CP is also a small area and the target lesion is likely to be attached to the right hepatic artery and/or its major branches; therefore, tumor-feeder detectability may also be reduced.

There is a tendency among origins of tumor-feeding caudate arteries according to the tumor location. SP, PC, and CP tumors are fed by the caudate arteries arising from the hepatic artery, with right to left ratios of 3:2, 3:1, and 3:0, respectively.⁵ This study also showed that ratios of right to left hepatic artery were 6:5 and 5:2 in SP and PC tumors, respectively, and CP tumors were dominantly supplied by the right hepatic artery. This information is helpful to identify a tumor-feeder on DSA when AFD cannot detect it.

In the present study, the outcomes of TACE for HCC in the caudate lobe using guidance software were worse than those of previous reports that included tumors originating at other sites.^16–19 Additionally, there was no relationship between the rates of tumor-feeder detection and outcomes of TACE among the three subsegments. The feeder detectability was the highest in PC tumors, but the outcomes of TACE were reduced. In contrast, the feeder detectability was the lowest in SP tumors, but the outcomes of TACE were the best. These suggest that PC tumors might have a potential to be supplied by tiny feeders that could not be detected by guidance software because PC is a watershed area between the bilateral hepatic lobes. Although the feeder detectability in CP tumors was higher than that in tumors in SP, the outcomes were the lowest. This also suggests that CP tumor might be supplied by tiny branches arising from the right hepatic artery and/or its major branches that could not be detected by guidance software. Moreover, complex arterial communications in the hilar plate can promote tumor recurrence.^27–30 In contrast, SP tumors might be fed by various branches including extrahepatic arteries;^3,5,8 however, SP is “the periphery of the liver.” As a result, favorable outcomes can be achieved when the tumor-feeder is selectively embolized.

There are several limitations of this study. First, selective DSA and/or CBCT was not performed in almost all embolized branches. Additionally, a safety margin was added to the target lesions of most tumors. These might overestimate the number of the tumor-feeder. Second, the anatomical border between PC and the posterior aspect of segment 4, which is frequently supplied by the caudate artery.²⁹ as well as that of CP and segment 6, are unclear on images;³¹ therefore, some tumors might be extended outside the caudate lobe. Additionally, a tumor located between two subsegments was classified as the subsegment in which it was dominant. These might influence the results of this study. Finally, the prognosis was not evaluated because many patients had multiple tumors outside the caudate lobe, and HCCs in the caudate lobe mainly developed at the end of the treatment course.

In conclusion, AFD could detect 86% of tumor feeders of HCCs in the caudate lobe. However, the detectability of tumor feeders and outcomes differed among the three subsegments. The anatomical location and presence of undetectable tiny tumor feeders may lead to unfavorable outcomes.

None.

The corresponding author received lecture fees from Guerbet Japan, Asahi Intecc, and Philips Healthcare, but other authors have no conflict of interest to disclose with respect to this paper.