Ablative margins in percutaneous thermal ablation of hepatic tumors: a systematic review

ABSTRACT

Introduction

This study aims to systematically review current evidence on ablative margins and correlation to local tumor progression (LTP) after thermal ablation of hepatocellular carcinoma (HCC) and colorectal liver metastases (CRLM).

Methods

A systematic search was performed in PubMed (MEDLINE) and Web of Science to identify all studies that reported on ablative margins (AM) and related LTP rates. Studies were assessed for risk of bias and synthesized separately per tumor type. Where possible, results were pooled to calculate risk differences (RD) as function of AM.

Results

In total, 2910 articles were identified of which 43 articles were eligible for final analysis. There was high variability in AM measurement methodology across studies in terms of measurement technique, imaging modalities, and timing. Most common margin stratification was < 5 mm and > 5 mm, for which data were available in 25/43 studies (58%). Of these, all studies favored AM > 5 mm to reduce the risk of LTP, with absolute RD of 16% points for HCC and 47% points for CRLM as compared to AM < 5 mm.

Conclusions

Current evidence supports AM > 5 mm to reduce the risk of LTP after thermal ablation of HCC and CRLM. However, standardization of AM measurement and reporting is critical to allow future meta-analyses and improved identification of optimal threshold value for clinical use.

KEYWORDS:

• Thermal ablation
• ablative margins
• local tumor progression
• hepatocellular carcinoma
• colorectal liver metastases

1. Introduction

Liver tumors constitute a substantial and growing proportion of all malignancies worldwide. Hepatocellular carcinoma (HCC) is the most common primary liver cancer with a global incidence of 0.9 million cases in 2020 [1]. Secondary liver metastases comprise approximately 70% of all liver tumors, with colorectal cancer (CRC) being the most common primary origin [2]. Up to 35% of CRC patients will either be diagnosed synchronously or develop colorectal liver metastases (CRLM) during follow-up [3–7]. Survival rates for patients diagnosed with HCC or CRLM have increased over the past decades, but the prognosis for these patient groups remains poor with a median 5-year survival rate of 20–25% [8–10].

Preferential curative treatments for HCC and CRLM are liver transplantation and surgical tumor resection in combination with (adjuvant) chemotherapy, as these have been regarded to provide best long-term survival rates [11]. Although the number of liver resections increased substantially in recent years, less than half of all liver tumors are deemed resectable [12].

In recent years, thermal ablation has been established as a widely adopted alternative to surgical resection in early-stage HCC and low-volume irresectable CRLM. Compared to surgery, thermal ablation has been associated with shorter hospital stay, shorter time to recovery and a lower rate of complications [13–17] and has been shown to offer comparable 1-, 3- and 5-year overall survival rates [16–20]. Previous less reliable methods of percutaneous treatments of liver tumors, like percutaneous ethanol injection, have been phased out in recent years [18]. Additionally, ablation techniques and knowledge have advanced, allowing for more patients to be treated curatively or as bridging to liver resection or transplantation. This can consequently be attributed as one of the reasons for improved survival of patients with liver tumors [19].

Despite these advantages, local tumor recurrence after thermal ablation tend to be higher, with reported rates between 2.9% and 21.7% and even ranging up to 50% in some smaller studies [20–27]. Multiple predictors for local recurrence after ablation have been suggested, including tumor size, number of tumors, location, and ablation technique. In recent years, the ablative margin (AM) or safety margin, defined as the minimal margin distance that is achieved between the area of tissue necrosis and tumor boundary, has been investigated as a prognostic factor. Current standard practice is to target complete ablation of the tumor including a surrounding margin of at least 5 mm for HCC and 10 mm for CRLM to achieve technical success [28]. However, there is currently no evidence-based consensus on the optimal threshold to ensure complete ablation and different techniques have been used to assess the AM in clinical practice.

Therefore, the aim of this study is to systematically review currently available data on correlation between ablative margin thresholds and local recurrence rates after percutaneous thermal ablation for HCC and CRLM. Additionally, this review provides an overview of concurrent methods that have been studied for ablative margin measurement.

2. Methods

2.1. Literature search

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed to conduct and report this systematic literature review [29]. No protocol for this study has been prospectively registered. The search strategy consisted of three elements: liver tumors, image guidance, and percutaneous ablation, and was performed on PubMed and Web of Science. The search strategy was developed in collaboration with an information specialist of the medical library at our institution. The following search terms and synonyms were used: liver tumors, ablation, magnetic resonance imaging, computer tomography, ultrasound, image guided, and guidance. The complete search queries for each search index are available in the Appendix. The date of publication was restricted to between January 2012 up until December 2022, as methods to assess the ablative margin using volumetric analysis and image registration techniques have changed significantly over the past decade.

2.2. Article selection

After removal of duplicates using Mendeley (Mendeley v1.19.8, Mendeley Ltd.) according to the method described by Kwon et al., remaining articles were imported in a browser-based review tool (The Systematic Review Accelerator) to facilitate the screening process [30,31]. Articles were scanned on title and abstract for inclusion criteria: (1) thermal ablation using RFA or MWA in a patient cohort, (2) ablative margins, and (3) local tumor progression (LTP). Articles were excluded if they described (1) combinations of thermal ablation with other treatments, (2) laparoscopic or other non-percutaneous approaches, (3) case studies, (4) animal studies, and (5) the use of nuclear imaging. Furthermore, articles had to be written in English and not be categorized as review, meta-analysis, commentary, or conference proceedings.

Two readers (K.V. and S.A.) independently reviewed a subgroup of 100 articles for the inclusion and exclusion criteria. Inclusions were compared between the two readers and conflicts were discussed and resolved. After agreement was reached for all articles, the first reader (K.V.) continued including articles and adjusted his inclusion based on the agreements reached with the second reader (S.A.). After inclusion, full-text assessment was performed by K.V. Articles were further excluded if no details were given on ablative margin measurements, LTP rates or ablative treatments other than RFA or MWA were used.

2.3. Quality assessment

Risk of bias assessment was performed using the Newcastle-Ottawa Scale (NOS) for nonrandomized cohort studies [32]. Studies could receive stars for subquestions of each category (selection, comparability, and outcome) with a maximum of nine stars. For comparability, studies received one or two stars if the study controlled for tumor size and any additional factor. Overall risk of bias was rated low for studies with 7–9 stars, moderate for 4–6 stars, and high for studies with less than 4 stars.

2.4. Data extraction

From each included article the following data was collected: study design, patient population and characteristics, age, tumor type and size, imaging techniques before, during and after thermal ablation procedure, ablation techniques, image registration techniques, LTP rate, ablative margin thresholds, patients per ablative margin threshold, technical parameters for ablative margin measurement methodology, and significant factors influencing LTP. In studies that did not directly provide numbers on cohort sizes, cohort sizes between different margin thresholds were calculated from indirectly provided data, such as survival graph data and tables.

2.5. Data synthesis and analysis

Data was synthesized and stratified for analysis using Microsoft Office Excel (Microsoft Corp., Redmond, WA, USA). Where possible, data were pooled for subgroups of patients with ablative margins less than 5 mm (<5 mm) and greater than 5 mm (>5 mm), since this was the most common subgrouping available in the included articles. In studies giving multiple thresholds, threshold groups were merged in order to fit the margin stratification for pooled analysis. Absolute risk differences (RD) were calculated for the risk of developing LTP between patients with AM <5 mm and patients with AM >5 mm. There was no common way to categorize AM of 5 mm; therefore, this margin was not assigned to either one of the subgroups, but the categorization of the individual studies was followed. Analysis of the data was performed in RStudio (RStudio Team, PBC, Boston, MA, USA).

3. Results

3.1. Literature search

From the research repositories PubMed and Web of Science, 3779 articles were identified. Totally, 869 duplicates were removed from the list. Of the remaining 2910 unique articles, the title and abstracts were read and screened. Based on the pre-determined in- and exclusion criteria, 2843 articles were further excluded. For three of the remaining 67 articles, the full-text articles could not be retrieved and were further excluded. Of the 64 articles included for full-text reading, another 21 were excluded for not adhering to in- and exclusion criteria. A total of 43 articles were selected for final inclusion in this review. A flowchart of the literature search results is available in Figure 1.

Figure 1. Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) flow diagram describing the results of the systematic literature search and additional selection based on exclusion criteria. LTP = local tumor progression.

3.2. Study characteristics

Detailed characteristics of the studies included in this review are available in Table 1. The majority of the articles originated from Asia (n = 28), with China (n = 14) and Japan (n = 11) as the most common countries of origin of the articles. The remainder of the articles came from The Netherlands (n = 4), the USA (n = 4), South Korea (n = 4), France (n = 2), Austria (n = 2), Romania (n = 1), and Italy (n = 1).

Table 1. Characteristics of articles included in this review. NR = not reported; PS = prospective; RS = retrospective; RFA = radiofrequency ablation; MWA = microwave ablation; LTP = local tumor progression; HCC = hepatocellular carcinoma; CRLM = colorectal liver metastases; AM = ablative margin; 3D-QAM = three-dimensional quantitative ablative margin; NLR = neutrophil-lymphocyte ratio, 3D-CEUS-FI = three-dimensional contrast enhanced ultrasound fusion imaging.

Reference	Year	Country	Study design	Ablation	Patients (lesions)	Tumor type	Tumor size (cm)	Factor influencing LTP
An et al. [33]	2020	China	PS	MWA	141	HCC	2.3 ± 0.9	Older age (HR = 2.463; 95% CI: 1.028, 6.152; p = .032) and an AM ≤5 mm (HR = 3.195; 95% CI: 1.324–7.752; p = .010)
An et al. [34]	2019	China	RS	MWA	68	HCC	3.4 ± 0.6	Maximal tumor diameter (HR = .923; p = .006) and tumor map (HR = .716; p = .003)
Bo et al. [35]	2017	China	RS	RFA	34 (47)	Mixed	2.0 ± 1.0	NR
Du et al. [36]	2022	China	PS	MWA	31 (42)	HCC	NR	NR
Faber et al. [37]	2022	The Netherlands	RS	RFA	21 (29)	CRLM	2.05	Uncovered tumor volume (622 mm3 (IQR 471–2246)) vs. (123 mm3 (IQR 11–230); p = .010);
Fukuda et al. [38]	2019	Japan	PS	RFA	76 (85)	HCC	1.5 (0.8–3.0)	NR
Hendriks et al. [39]	2019	The Netherlands	RS	RFA	18	HCC	2.2 (1.2–1.7)	AM size (p = .001)
Hocquelet et al. [40]	2017	France	RS	RFA	16	HCC	2.9	Margins >425 mm2 (HR = 6.71,95% CI: 1.27–35.5; p = .014)
Inoue et al. [41]	2013	Japan	PS	RFA	70 (86)	HCC	1.7 ± 0.7	NR
Izaaryene et al. [42]	2021	France	RS	MWA	84	CRLM	1.5 ± 0.8	Presence of a colonic obstruction or a perforation at the time of initial diagnosis (p < .05), the number of metastases at the time of diagnosis (p < .05), the number of surgical resections of liver metastases before MWA (p < .05), the type of protocol (p < .05), and ablative margins (p < .001)
Jiang et al. [43]	2018	China	RS	RFA	134 (159)	HCC	2.0 ± 0.9	min-AM <5 mm (HR = 4.587, 95% CI: 1.045–22.296; p = .044).
Joo et al. [44]	2022	USA	RS	MWA	172 (334)	Mixed	1.85 ± 0.89	Insufficient ablative margin on post-MWA MRI (HR = 14.39, 95% CI: 4.61–44.92: p < .001), perivascular tumor location (HR = 6.03, 95% CI: 1.83–19.87: p = .003), and tumor size (HR = 1.0, 95% CI: 1.01–1.11: p = .011)
Kaye et al. [45]	2019	USA	RS	RFA	72 (93)	CRLM	1.8	5-mm metric highest 2-year LTP discrimination power, AUC = .923 (p = .004)
Ke et al. [46]	2014	China	PS	RFA	281	HCC	4.1	AM >1.0 cm (HR = 1.484, 95% CI: .101–1.812; p = .001)
Kim et al. [47]	2018	South Korea	PS	RFA	33 (42)	HCC	1.58 ± 0.59	NR
Laimer et al. [48]	2021	Austria	RS	RFA	45 (76)	CRLM	2.42	Peri-ablational 3D safety margin (p < .001)
Laimer et al. [49]	2020	Austria	RS	RFA	110 (176)	HCC	2.52 ± 1.49	Minimum AM (p = .036)
Li et al. [50]	2021	China	RS	RFA/MWA	444	HCC	1.9	Tumor size (HR = 2.16; 95% CI: 1.25–3.72; p = .003) and AM (HR = .72; 95% CI: .61–.85; p < .001)
Li et al. [51]	2016	China	PS	RFA	98 (126)	HCC	1.96 ± 0.77	AM (HR = 9.167, 95% CI: 1.070–78.571; p = .043)
Makino et al. [52]	2013	Japan	RS	RFA	85 (94)	HCC	1.4 ± 0.52	Protrusion from the ablation zone (HR = 7.09, 95% CI: 2.26–22.3; p < .001)
Makino et al. [53]	2016	Japan	RS	RFA	69 (85)	HCC	1.19	NR
Matsuki et al. [54]	2021	Japan	PS	RFA	42 (48)	HCC	1.32	NR
Minami et al. [55]	2020	Japan	RS	RFA	101 (121)	HCC	1.8 ± 0.7	NR
Park et al. [56]	2018	South Korea	RS	RFA	178 (178)	HCC	1.73 ± 0.61	NR
Ruiter et al. [57]	2021	The Netherlands	RS	MWA	47 (65)	CRLM	1.3	3D-QAM (OR = .52, CI .29–.95; p = .03)
Sakakibara et al. [58]	2014	Japan	RS	RFA	84 (133)	HCC	1.38 ± 0.46	NR
Sanga et al. [59]	2020	Japan	PS	RFA	43 (5)	HCC	1.53	NR
Shady et al. [60]	2018	USA	RS	RFA/MWA	110 (145)	CRLM	NR	RFA: ablative margin <5 mm (p < .001) and peri-vascular location (p < .001)MWA: ablative margin <5 mm (p < .001) and no prior liver resection (p = .002) MWA: ablative margin <5 mm (p < 0.001) and no prior liver resection (p = 0.002)
Shin et al. [61]	2014	South Korea	PS	RFA	240	HCC	1.95 ± 0.79	Visual vs. CT margin assessment (p = .0101)
Sibinga Mulder et al. [27]	2019	The Netherlands	RS	RFA	18	CRLM	2.2	NR
Solbiati et al. [62]	2019	Italy	RS	MWA	50 (90)	HCC	2.7 ± 2	NR
Sparchez et al. [63]	2019	Romania	RS	RFA/MWA	61 (82)	Mixed	2.5	Ablative margin ≤5 mm (p = .002) and tumor number (p = .049)
Takeyama et al. [64]	2019	Japan	RS	RFA	29 (59)	HCC	1.12 ± 0.44	Ablative margin status (p = .028)
Tang et al. [65]	2015	China	PS	RFA	75	HCC	2.4 ± 0.7	Pre-operative NLR (HR: 2.34, 95% CI: 1.11–5.12; P = .026), AM change (HR = 2.15, 95% CI: 1.01–4.60; p = .048)
Tokunaga et al. [66]	2012	Japan	PS	RFA	45 (60)	HCC	1.68 ± 0.59	NR
Toshikuni et al. [67]	2017	Japan	PS	RFA	45 (45)	HCC	1.9	NR
Vasiniotis Kamarinos et al. [68]	2022	USA	RS	RFA	68 (104)	CRLM	1.9	NR
Wang et al. [69]	2016	China	PS	RFA	51 (61)	HCC	2.0 ± 1.0	NR
Wang et al. [70]	2013	USA	RS	RFA	94	CRLM	1.75	Tumor size, for each 5 mm increase (HR = 1.22, 95% CI: 1.02–1.49; p = .028)Ablative margin, for each 5 mm increase (HR = .54, 95% CI .36–.79; p = .002) Ablative margin, for each 5 mm increase (HR = 0.54, 95% CI 0.36–0.79; p = 0.002)
Ye et al. [71]	2018	China	PS	RFA	55 (55)	HCC	2.0 ± 0.7	NR
Yoon et al. [72]	2019	South Korea	PS	RFA	68 (88)	HCC	1.6 ± 0.6	Locally progressed tumor (OR = 8.67, 95% CI: 1.43–52.49; p = .019)Insufficient ablative margin (OR = 7.12, 95% CI: 1.49–34.12; p = .014) Insufficient ablative margin (OR = 7.12, 95% CI: 1.49–34.12; p = 0.014)
Zhang et al. [73]	2019	China	PS	RFA	84	HCC	2.1 ± 0.7	AM ≥5 mm on 3D-CEUS-FI (p = .024)
Zhong-Zhen et al. [74]	2013	China	PS	RFA	41 (49)	HCC	1.8 ± 0.86	NR

None of the studies were randomized trials. In total, 18 of the 43 articles (42%) were set up as prospective research. The remaining 25 articles used retrospectively selected data. One study used multicenter data, and all other articles were single center investigations.

3.3. Quality assessment

Risk of bias assessment for studies included in meta-analyses is shown in Figure 2. Almost all studies scored low risk of bias on selection (23/23) and outcome (22/23) domains. Only in comparability domainwere high risk (10/23) and moderate risk of bias (7/23) noted. Main reason for worse scores on the comparability domain was lack of control for confounders. All studies scored ≥ 7 out of 9 stars and overall risk of bias was rated as low.

Figure 2. Risk of bias according to Newcastle-Ottawa Scale for selection, comparability, outcome and overall risk of bias.

3.4. Ablative margin and local tumor progression

3.4.1. HCC

Of the 43 articles, 31 (72%) investigated the relation between AM and local recurrence for HCC. Detailed characteristics of each study are available in Table 2. Five of the 31 articles (16%) did not employ a registration of pre- and post-ablation image volumes and relied on visual methods to assess the ablation margins; with three articles using side-by-side visualization of pre- and post-ablation images to assess ablation margin, the other two articles using contrast agents to visualize residual tumor and ablative margins on post-ablation imaging.

Table 2. Ablative margin and measurement characteristics of studies investigating hepatocellular carcinoma. AM = ablative margin; LTP = local tumor progression; CT = computer tomography; CECT = contrast enhanced CT; MRI = magnetic resonance imaging; CEMRI = contrast enhanced MRI; US = ultrasound; CEUS = contrast enhanced US. *Results from rigid registered analysis is given; **Results from non-rigid registered analysis is given.

Reference	Pre-ablation imaging	Time from treatment	Post-ablation imaging	Time from treatment	Registration	AM method	Median follow-up	Ablation margins	Lesions/ patients	LTP rate
An et al. [33]	CEMRI	<1 month	CEMRI	<3 months	Non-rigid	2D	28.9 months	≤5 mm	61	26.2%
								>5 mm	80	3.8%
An et al. [34]	CEMRI	NR	CEMRI	<1 week	Rigid	3D	21 months	≤5 mm	24	50.0%
								>5 mm	44	9.1%
Du et al. [36]	CECT	NR	CECT	NR	Rigid	2D	30 months	5-8 mm	16	43.8%
								8-10 mm	16	6.3%
								≥10 mm	10	10.0%
Fukuda et al. [38]	CEMRI	<8 hours	CEMRI	4–7 days	Visual	2D	40 months	AM(-)	11
								AM(0)	24	34.0%
								AM(+)	50	2.0%
Hendriks et al. [39]	CECT	Intraprocedural	CECT	Intraprocedural	Non-rigid	2D	9.5 months	<0 mm	12	66.7%
								>0 mm	6	0.0%
Hocquelet et al. [40]	CEMRI	NR	CEMRI	1 month	Non-rigid	3D	26 months	976 ± 1064 mm^2 <5 mm	8	100.0%
								319 ± 385 mm^2 <5 mm	8	0.0%
Inoue et al. [41] *	CECT	<1 month	CEUS/CECT	1–3 days	Visual	2D	28.7 ± 14 months	<0	3
								0–5	49	18.0%
								>5	33	0.0%
Jiang et al. [43]	CECT	1 week	CECT	1 month	Rigid	3D	26 months	<5 mm	101	35.1%
								≥5 mm	58	4.1%
Ke et al. [46]	CECT	Intraprocedural	CECT	Intraprocedural	Visual	2D	38.2 months	5-10 mm	158	70.9%
								>10 mm	123	44.7%
Laimer et al. [49]	CECT	Intraprocedural	CECT	Intraprocedural	Rigid	2D	26.0 ± 10.3 months	<5 mm	110	9.1%
								>5 mm	66	0.0%
Kim et al. [47]	CT	NR	MRI	4–7 hours	Visual	2D	26.1 ± 5.7 months	AM(-)	1	40%
								AM(0)	9
								AM(+)	32	3.1%
Li et al. [50]	MRI	<1 month	MRI	<1 month	Rigid	2D	19.9 months	Continuous	444	44.0%
Li et al. [51]	CT/MRI	NR	CEUS	Intraprocedural	NR	2D	23 months	<5 mm	43	11.6%
								>5 mm	82	1.2%
Makino et al. [52]	CECT	NR	CECT	<7 days	Rigid	2D	21 months	<0 mm	23	44.6%
								0-5 mm	61	13.1%
								≥5 mm	10	0.0%
Makino et al. [53]	CT/MRI	<2 months	US	Intraprocedural	NR	2D	12 months	<0 mm	0
								0-5 mm	73	8.90%
								≥5 mm	6	0%
Matsuki et al. [54]	None	None	CECT/CEMRI	<7 hours	NR	2D	26.3 ± 12.7 months	AM(-)	6	50.0%
								AM(+)	42	4.8%
Minami et al. [55]	US/CECT/MRI	US: Intraprocedural CECT/MRI: NR	US/CECT/MRI	US: Intraprocedural CECT/MRI: NR	Visual	2D	19 months	<5 mm	253	5.5%
								>5 mm	321	0%
Park et al. [56]	MRI	NR	CECT	Immediate post-ablation	Non-rigid	2D	32 months	0-2 mm	34	32.4%
								2-5 mm	32	9.4%
								≥5 mm	112	2.7%
Sakakibara et al. [58]	CECT/CEMRI	NR	CECT/CEMRI	NR	Rigid	2D	NR	<0 mm	39	34.5%
								0 mm	41	12.5%
								5 mm	41	8.5%
								≥5 mm	12	0.0%
Sanga et al. [59]	CEUS	Intraprocedural	CEUS	Intraprocedural	Rigid	2D	21.4 months	Incomplete	1
								Grade C (0 mm)	0
								Grade B (0-5 mm)	25	4.0%
								Grade A (≥5 mm)	24	0.0%
Shin et al. [61]	CECT	<6 weeks	CECT	Intraprocedural	Non-rigid	2D	29.2 ± 10.8 months	<2 mm	14	14.3%
								≥2 mm	136	10.3%
Solbiati et al. [62]	CECT	NR	CECT	<1 day	Non-rigid	3D	12 months	100% >5 mm	24	0.0%
								90–99% >5 mm	17	5.9%
								50–89% >5 mm	30	16.7%
								0–49% >5 mm	2	100.0%
								Incomplete	17	76.5%
Takeyama et al. [64]	CEMRI	<3 months	CEMRI/MRI	<3 days	Rigid	2D	37.9 months	AM(+)	30	3.3%
								AM(0)	25	32.0%
								AM(-)	3	100.0%
Tang et al. [65]	CECT	NR	CECT	1 month	Rigid	3D		0-5 mm	39
								≥5 mm	36
Tokunaga et al. [66]	None	None	CEMRI	7 days	Visual	2D	20 ± 5 months	AM(+)	17	0.0%
								AM(0)	35	11.8%
								AM(-)	3
Toshikuni et al. [67]	CECT/MRI/CEUS	NR	US	Intraprocedural	Rigid	2D	27 months	incomplete	12%	8.3%
								Incomplete	40%	12.6%
Wang et al. [69]	CEMRI	NR	CEMRI	1 month	Rigid	2D	14.2 ± 5.4 months	0-5 mm	29	13.8%
								>5 mm	32	0.0%
Ye et al. [71]	3D-CEUS	Intraprocedural	3D-CEUS	Intraprocedural	Rigid	2D	13.6 months	<5 mm	28	28.0%
								≥5 mm	25	0.0%
Yoon et al. [72] **	CEMRI	<30 days	CECT	Immediate post-ablation	Non-rigid	2D	48 months	<3 mm	10	66.7%
								>3 mm	57	24.6%
Zhang et al. [73]	3D-CEUS	Intraprocedural	3D-CEUS	Intraprocedural	Rigid	2D	12.5 ± 4.8 months	<5 mm	49	16.3%
								≥5 mm	28	0.0%
ZhongZhen et al. [74]	CEMRI/CECT	NR	3D-CEUS	Intraprocedural	Rigid	2D	38 months	<5 mm	13	30.8%
								≥5 mm	27	0.0%

Various methods of AM categorization were applied in the studies. Most studies (17/31; 55%) categorized the lesions according to the AM reaching >5 mm or not. Five articles (16%) graded the ablative margins based on the continuity of the ablative zone around the tumor; AM(+) having an uninterrupted ablative rim around the tumor, in AM(0) the tumor is completely encapsulated but ablative rim is partially discontinued, AM(-) tumor protrudes from ablation zone. Two studies investigated surface coverage of the tumor by a 5 mm margin, with both studies showing a larger coverage of the margin reducing the risk of LTP [40,62]. The other articles used unique categorization of the ablative margins. Where possible, results were pooled to analyze the effect of AM <5 mm vs. >5 mm. Seventeen articles could be included, resulting in a pooled population of 782 lesions with AM <5 mm, and 769 lesions with AM >5 mm. All studies favored AM >5 mm to reduce the risk of LTP. Pooled analysis shows a reduction in the absolute risk for an HCC treated with AM >5 mm of developing LTP by 16% points (95% CI: 0.11–0.21), compared to AM <5 mm (Figure 3).

Figure 3. Forest plot and risk difference calculation for studies investigating HCC. LTP = local tumor progression; AM = ablative margin.

3.4.2. CRLM

Nine studies (22%) assessed the LTP rate for ablative margins reached in the treatment for CRLM, detailed characteristics on AM measurements and LTP rates are available in Table 3.

Table 3. Ablative margin and measurement characteristics of studies investigating colorectal liver metastases. AM = ablative margin; LTP = local tumor progression; CT = computer tomography; CECT = contrast enhanced CT; MRI = magnetic resonance imaging; CEMRI = contrast enhanced MRI. *Results from rigid registered analysis is given.

Reference	Pre-ablation imaging	Time from treatment	Post-ablation imaging	Time from treatment	Registration	AM method	Median follow-up	Ablation margins	Lesions/ patients	LTP rate
Faber et al. [37]	MRI/CECT	NR	CECT	Intraprocedural	Rigid	3D	24 months	Positive (2.4 ± 2.5 mm)	17
								Negative (−4.7 ± 2.7 mm)	12
								−6.6 mm median		0,0%
								0.5 mm median		100,0%
Izaaryene et al. [42]	MRI/CECT	<30 days	MRI	1 month	Visual	2D	13.3 months	<1 mm	5	100,0%
								2-5 mm	15	73,3%
								6-10 mm	12	8,3%
								>10 mm	7	0,0%
Kaye et al. [45]	CECT	NR	CECT/MRI/PET-CT	4–8 weeks	Rigid	3D	24 months	Incomplete	31	90,3%
								0 mm	6	100,0%
								1-5 mm	29	51,7%
								6-10 mm	21	4,8%
								>10 mm	6	0,0%
Laimer et al. [48]	CECT	Intraprocedural	CECT	Intraprocedural	Non-rigid	3D	36.1 ± 18.5 months	≥3 mm > 90%	74	9,4%
								≥3 mm > 95%	68	6,9%
								≥3 mm 100%	55	0,0%
Ruiter et al. [57]	CECT	Intraprocedural	CECT	Intraprocedural	NR	3D	1 year	−2.4 mm median	10	100,0%
								0.0 mm median	55	0,0%
Shady et al. [60]	CECT	NR	CECT	6 weeks	Visual	2D	RFA: 56 months MWA: 29 months	<5 mm	77	70.8%
								5-10 mm	46	13%
								>10 mm	22	0%
Sibinga Mulder et al. [27]	CECT	<2 months	CECT	Intraprocedural	Non-rigid	2D	44.7 ± 20.5 months	−2.6 mm average	9	100,0%
								2.1 mm average	9	0,0%
								Incomplete (≤0 mm)		88,9%
								Complete (≥1 mm)		11,1%
Vasiniotis Kamarinos et al. [68] *	CECT	Intraprocedural	CECT	Intraprocedural	Rigid	3D	21 months	<5 mm	74	50,0%
								>5 mm	30	10,0%
Wang et al. [70]	CECT	<6 weeks	CECT	4–8 weeks	Visual	2D	20 months	0 mm	30	74%
								1-5 mm	41	54%
								6-10 mm	15	26%
								11-15 mm	8	20%

One study categorized the AM according to reaching AM <5 mm or >5 mm. Four other studies used multiple threshold categorizations. Three studies analyzed the mean or median MAM for lesions with and without LTP, all concluding that lesions with LTP were associated with a mean negative MAM, while the lesions without LTP had completely ablated tumors with a mean positive margin: −6.6 mm vs. 0.5 mm, −2.4 mm vs. 0.0 mm, and −2.6 mm vs. 2.1 mm, respectively [27,37,57].

Where possible, results were pooled to analyze the effect of AM <5 mm vs. AM >5 mm (Figure 4) Six studies could be included, with a pooled population of 316 lesions vs. 201 lesions for, respectively, AM <5 mm and AM >5 mm. Again, all studies favored AM >5 mm to reduce the risk of LTP. Pooled analysis shows an absolute RD between CRLM treated with AM <5 mm vs. AM >5 mm of 47% points (95% CI: 0.30–0.64).

Figure 4. Forest plot and risk difference calculation for studies investigating CRLM. LTP = local tumor progression; AM = ablative margin.

3.4.3. Mixed

Three studies reported on mixed populations of both CRLM and HCC, without specifying results per tumor type. Again, these studies show a decrease in LTP with increasing AM. Due to the limited number of studies available, RD was not calculated for these studies.

3.5. Ablative margin assessment

3.5.1. Pre-ablation imaging

Contrast enhanced computed tomography (CECT) scans were acquired prior to thermal ablation in 17 out of 41 (41%) studies. Seven studies (17%) used contrast enhanced magnetic resonance imaging (CEMRI) as pre-treatment imaging. Two studies did not require pre-ablation imaging for AM assessment and used only post-ablation imaging with contrast agents for measuring the AM [54,66]. The other studies used multiple different modalities of pre-ablation imaging for their AM measurements, details are available in Table 2, Tables 3 and 4. Time from imaging to treatment session was often not reported (17/41 studies; 41%). Reported periods from imaging to treatment ranged from intraprocedural imaging up to 3 months prior to ablation. Intraprocedural imaging was used in 11 of the 41 studies (27%).

Table 4. Ablative margin and measurement characteristics of studies investigating mixed populations of HCC and CRLM. AM = ablative margin; LTP = local tumor progression; CT = computer tomography; CECT = contrast enhanced CT; MRI = magnetic resonance imaging; CEMRI = contrast enhanced MRI; CEUS = contrast enhanced ultrasound. * sufficient ablative margin is categorized as >5 mm for HCC, > 10 mm for CRLM.

Reference	Pre-ablation imaging	Time from treatment	Post-ablation imaging	Time from treatment	Registration	AM method	Median follow-up	Ablative margins	Lesions/ patients	LTP rate
Bo et al. [35]	CECT/CEMRI	NR	CEUS	1–3 days	Rigid	2D	2–34 months	<0 mm	4	25,0%
								0-5 mm	15	2,3%
								≥5 mm	28
Joo et al. [44] *	CT/MRI	<2 months	MRI	1 week	NR	2D	11 months	Insufficient	66	33,8%
								Sufficient	205	6,90%
Sparchez et al. [63]	CEUS	NR	CEUS	4 weeks	Rigid	2D	NR	≤5 mm	13	76,9%
								>5 mm	69	18,8%

3.5.2. Post-ablation imaging

Post-ablation imaging was most often performed by CECT scans (19/43; 44%), with CEMRI and (3D-) contrast-enhanced ultrasonography (CEUS) following with both six studies (24%). The other studies used various combinations of post-ablation imaging for their AM measurements, details are available in Tables 2, 3, and Table 4. Nineteen studies (44%) performed the AM measurements on imaging acquired intraprocedural or immediate post-ablation. Eleven other articles (26%) acquired their imaging within a week of ablation treatment. In 11 articles (26%), the interval from treatment to image acquisition ranged from 1 month up to 3 months. In the remaining two articles, the interval from treatment to follow-up imaging was not reported. Several studies investigated the agreeability in measured margins between multiple imaging modalities [38,41,47,53,59,71].

3.5.3. Image registration

Multiple methods for image registration were employed in the included studies. The various methods were simplified into two methods: rigid registration vs. non-rigid registration. Nine articles (21%) used a non-rigid registration method to deformably align the pre-and post-ablation image volumes. In 20 studies (47%), a rigid registration was used. Five studies (12%) did not report their method for image registration. Eight studies did not perform a registration of the images, instead assessing the ablative margins on side-by-side visualization of the pre- and post-imaging.

One study compared two different 3D registration methods (rigid with tissue contraction simulation and rigid without tissue contraction simulation) and 2D side-by-side visualization. The study found assessed ablative margin depending heavily on the registration method used, with 30/104 patients reaching >5 mm AM (10% LTP) on 3D registration without tissue contraction versus 70/104 patients (28.6% LTP) when using registration with tissue contraction versus 87/104 patients (36.8% LTP) on 2D side-by-side assessment [68].

3.5.4. 2D vs 3D margin assessment

Margin assessment can be achieved on 2D images, using the coronal, sagittal and axial imaging to measure or rate the ablative margin, or in 3D by segmenting the tumor and ablation zone and calculating distances or volumetric distance maps between the two segmented volumes. Of the 43 articles, 10 studies (23%) measured ablative margins in 3D. One article investigated 3D AM assessment versus 2D assessment and found 3D assessment to be more discriminative with an AUC of 0.893 vs. 0.790, respectively (p = 0.01) [45]. The 3D assessment resulted in fewer patients being categorized as having an AM greater than 5 mm: 27/93 (29%) with 3D measurement vs. 34/93 (37%) with 2D measurement.

3.6. Ablation modality

All studies reported ablation techniques used. Thirty-three out of the 43 (77%) articles employed RFA as their sole technique for thermal ablation. 7 out of 43 studies (16%) used solely MWA, while in three studies (7%) both modalities were used. These three studies further differentiated the groups by the use of either RFA or MWA. One study found patients treated with RFA having a higher chance of not achieving an AM greater than 5 mm, 79% vs. 100% of the patients treated with RFA vs. MWA [63]. This consequently resulted in 1-, 2-, and 3-year LTP rates of 27 vs 7%, 40% vs 15% and 47% vs 32% for RFA vs. MWA. However, in multivariate analysis the ablation modality was not an independent predictor for LTP. Another study found 51% and 42% of patients having an AM greater than 5 mm for RFA and MWA, respectively [60]. This was however not represented in the LTP rate after 1 and 2 years, being 31% and 39% for RFA vs. 25 and 40% for MWA, respectively. The third study found the ablation technique to have no statistically significant impact on LTP, whereas the measured ablative margin did [50].

3.7. Image guidance

Ultrasound (US) for image guidance of the ablation was used in 25 of the 42 included articles (60%). Computed tomography (CT) guidance was employed in 10 out of 42 articles (24%). In four studies (10%), both CT and US were used for image guidance. The remaining three articles did not mention the method of image guidance used during the ablation procedure.

3.8. Intraprocedural assessment

Fifteen studies (35%) performed an intraprocedural assessment of ablation completeness using pre- and post-ablation imaging. Three articles gave data on the effect of intraprocedural ablation verification with an additional study investigating the added value of intraprocedural CEUS image fusion with pre-ablation CECT/CEMRI/CEUS imaging [44,51,59,67]. In this study, the cohort treated with the aid of image fusion achieved a lower rate of insufficient ablations in the first RFA session compared to a control group (88% vs. 60%, p = 0.041). The image fusion cohort also saw a lower LTP rate of 8.3% vs. 12.6%, but these results were not significant (p = 0.98) [67].

4. Discussion

This review provides an overview of the current literature available on ablative margins and its relation to LTP after thermal ablation of HCC and CRLM. Present evidence underlines the current recommendation to aim for an ablative margin of at least 5 mm to significantly reduce the risk of LTP. At the same time, a multitude of methods is currently used for ablative margin measurement. Reported AMs may largely depend on various factors, such as imaging modality, timing of pre- and post-treatment imaging, image co-registration technique, and dimensionality of the measurements.

All studies included in this review found an association between an increased ablative margin and a reduction in risk of LTP. The RD between the pooled patient populations with AM less than 5 mm and greater than 5 mm was 16% points for HCC, while for CRLM, the RD was even larger with 47% points. Considering reported LTP rates in current literature are on the order of 3–50%, with an average of 13.6% in a previous meta-analysis, this indicates local recurrences may to a large degree be attributable to achieving insufficient ablative margins. This corresponds to several studies reporting that upon retrospective analysis of patients previously treated with liver thermal ablation the intended AM of >5 mm was not achieved in large proportions (35% − 92%) [49,53].

However, heterogeneity in pooled analysis for both HCC and CRLM was relatively high, meaning that the found results cannot be completely attributed to the ablative margin but other factors may also be of influence. Although overall risk of bias was low, multiple studies showed moderate or high-risk of bias for comparability due to absence of control for confounders. Studies that did perform multivariate analysis of factors influencing LTP rate all concluded AM to be one of or the only independent predictor of LTP. Other identified predictive factors were tumor diameter, perivascular location and number of tumors or metastases [34,42,44,51,60,63,70]. For both tumor diameter and perivascular location, the effect can be rationalized as both factors make it increasingly more difficult to achieve complete tumor coverage including an adequate circumferential ablative margin. Large tumors will often require multiple overlapping ablations to achieve complete ablation coverage, increasing the risk of not achieving a spherical ablation zone completely encapsulating the tumor. Thermal ablation in perivascular locations is affected by the cooling effect of the blood flow, i.e. the heatsink effect. In these areas, achieved ablative margins are restricted to reaching the vessel wall or fully encircling the vessel with larger vessels often remaining patent. As a result, it is more difficult to achieve complete ablation, which is conceived to be the cause of increased risk of LTP. A recent meta-analyses concluded MWA to provide comparable or better results to RFA in terms of local recurrence rates and tumor control [75–77]. As MWA is known to be less sensitive to the heat sink effect and impedance differences between tissue layers and capable to achieve larger ablation zones, theuse of MWA may improve the ability of achieving sufficient ablative margins [78]. However, none of the included articles found the ablation technique to be a prognostic factor for LTP, suggesting that both techniques can achieve similar control rates as long as sufficient ablative margins are reached. Further analysis is, therefore, needed to determine whether MWA also leads to improved ablative margins and local outcomes at perivascular locations and in larger tumors [78].

Across studies, large variation was observed in methodology for ablative margin measurement in terms of used imaging modality, timing of pre- and post-treatment imaging, image co-registration technique and 2D vs. 3D measurement. In AM assessment, CT and MRI have a theoretical advantage over US, as they can provide volumetric imaging of the tumor and the ablation zone and are not impeded by gas formation hampering full ablation zone assessment. Post-ablation imaging was most often performed with CECT, as it can facilitate treatment of deeper situated tumors and provide direct imaging, enabling intraprocedural confirmation of tumor coverage by the ablation zone. Nevertheless, the use of ionizing radiation has to be considered by the treating physician. While MRI also provides near real-time imaging, the higher costs associated with MRI and the limited compatibility of RFA and MWA equipment can hamper implementation compared to CT guidance. US is a fast and often used tool for procedural guidance, as evidenced by the majority of articles included in this review, mostly limited by the ability of guidance for deeper situated tumors. There have been few studies performed investigating the influence of different imaging modalities on AM assessment. Of the included studies, Kim et al. found MRI to be superior to CECT in differentiating the AM based on subsequent LTP, while Inoue et al., Sanga et al. and Ye et al. found (3D-)CEUS to have good agreeability in the measured AM compared with measurements on CECT or CEMRI scans [41,47,59,71]. As it is currently not known if either of the imaging modalities under- or overestimates the ablative margin, further research in this direction is imperative to find the optimal imaging method of ablative margin measurements in the workflow for thermal ablation.

Time interval between ablation procedure and image acquisition varied between studies. Many studies did not mention the timing of pre-ablation imaging, with two studies including imaging made up to 2 months before the treatment. Any time interval may introduce errors to the AM assessment, since both HCC and CLRM are fast growing tumors that can double in volume within months [79–81]. Similarly, a delay in post-ablation imaging may bias the AM assessment due to morphological changes resulting from tissue responses occurring in the first days to weeks after ablation, followed by shrinkage [82,83]. Nine articles reported to have used both intraprocedural pre- and post-ablation imaging, but due to poor reporting of the interval for pre-ablation imaging this number is uncertain.

Image registration of pre- and post-ablation imaging is paramount to accurately assessing ablative margins. Side-by-side assessment of the ablative margin on pre- and post-ablation imaging has poor intrareader agreeability, with an intraclass correlation coefficient of 0.357 (95% CI: 0.194–0.522) in a study by Schaible et al [84]. This is backed up by studies from Yoon et al., Vasiniotis Kamarinos et al., Shin et al., and a recent article by Zhou et al. All found a relatively large proportion of ablations having an insufficient ablative margin after margin assessment using registered images, compared to initial side-by-side assessment [61,68,72,85]. In many articles conditions for successful registration were given. Oftentimes, registration was considered successful if target registration error between a chosen landmark was less than 3 mm or a similar measurement [52]. When measuring margins of 5 mm, a target registration error of up to 3 mm can have significant impact on the measured AM.

Most studies measured AM in 2D on coronal, axial and sagittal images. 3D measurement based on tumor and ablation volume surfaces potentially allows for a more accurate representation of the true minimal ablative margin in any given direction. The ten studies that performed measurements on 3D segmentations of tumor and ablation zone also were enabled to calculate subpercentages of the tumor that were treated with a certain ablative margin [40,48,57,60,62]. Nearly, all studies categorized AM according to predefined thresholds. In future studies it would be imperative to have studies providing data on continuous ablative margins. This allows for more detailed analysis of all ablative margins to find the optimal threshold to reduce risk of LTP.

Intraprocedural ablation verification is still relatively rarely used, with 16 included articles reporting the verification of ablation zone being part of their ablation work process. Articles and data on the effects of intraprocedural image fusion are even more scarce. Three articles provided numbers on the amount of re-ablations within the first session, after intraoperative imaging showed insufficient ablation margins. Joo et al., Li et al., and Sanga et al. performed additional ablations in 14% and 34% of patients, after intraprocedural AM assessment revealed insufficient ablative margins [44,51,59].

This review has a several limitations. Many different categorizations of ablative margins were used in the included studies, complicating homogenization of the data. A complete meta-analysis of the studies was therefore not possible. No consensus on reporting of ablative margins exists. The most common method of categorization was differentiation between AM <5 mm vs. AM >5 mm. However, this method is relatively crude and methodology of measurements varied greatly between different studies, introducing uncertainty to the found margins and the comparability. Additionally, the margin of 5 mm was not consistently grouped in either the group of AM <5 mm or AM >5 mm between different studies and remains ambiguous. Despite these uncertainties, margins measured to be greater than 5 mm were found to have a decreased risk of LTP. Standardization of AM measurement and reporting is critical to allow future meta-analyses and identification of optimal threshold value for clinical use.

Another limitation of this study is the inclusion of only studies related to HCC or CRLM. While these are the most common hepatic malignancies, nearly all tumors can metastasize to the liver. We can therefore not conclude whether these margins are also applicable to thermal ablation of other hepatic tumors. Some data were not systematically reported in all studies, such as imaging interval prior to treatment and the use of intraprocedural imaging to assess the ablative margin immediately post-ablation. These are all factors that can impact the achieved ablation zone and accuracy of measurements.

For future studies, it is recommended to report the complete work process and timing from pre-operative imaging to ablation treatment to post-operative imaging. Ideally, the imaging should all be performed intraprocedurally, to provide the most accurate and recent imaging to measure tumor, ablation zone and ablative margin. It is also recommended to perform measurements on segmented 3D models of the tumor and ablative zones, to provide more accurate and detailed results and the acquired ablative margins.

5. Conclusion

Current evidence supports the aim of achieving an ablative margin >5 mm to reduce the risk of LTP after percutaneous thermal ablation of HCC and CRLM. However, heterogeneity in AM measurement methods and a lack of continuous data on ablative margin thresholds hampers a detailed analysis of optimal margin for the minimization of LTP. Standardization of ablative margin measurement and reporting is critical to allow future meta-analyses and identification of optimal threshold value for clinical use.

6. Expert opinion

This review underlines the critical importance of achieving sufficient ablative margins for effective liver thermal ablation and summarizes the currently used methods for AM measurement. Even in the presence of substantial heterogeneity in AM measurement methodology, all studies confirm the influence of AM to reduce the risk of LTP. Therefore, to the author’s opinion, assessment of ablative margins should become a mandatory requirement to confirm treatment adequacy and ensure optimal and reliable outcomes of liver thermal ablation in daily practice.

As the importance of ablative margins has been increasingly recognized in recent years, multiple software tools to assess ablative margins are now entering the clinic. However, most of these tools have not been extensively validated in large cohorts and given the large dependency of AM measurement on imaging modality, timing of pre-ablation and post-ablation imaging and image co-registration technique, predictive value for LTP and the optimal threshold value may vary across different software tools and methodologies. The field would greatly benefit from standardization in AM measurement and reporting, allowing improved comparison of results across studies. At minimum, studies should report details on AM measurement methodology, provide appropriate 2 × 2 tables as well as absolute margin measurements and report related LTP rates. Ideally, continuous data of AM measurement studies should become available to facilitate data pooling and allow more accurate determination of an optimal cutoff value for clinical use. In addition, it would be recommendable to establish a large multicenter dataset with known outcome data to validate AM software tools and methodologies against and serve as a benchmark to assess predictive performance before entering the clinic.

Finally, optimal strategies to incorporate margin analysis in the clinical process need to be established. Ideally, we recommend the use of intra-interventional pre- and post-ablation CECT acquired in apnea to yield best accuracy in margin determination and allow for immediate re-ablation if required, thereby preventing the need for repeated treatment. Diagnostic pre-treatment imaging or post-treatment follow-up (e.g. day 1 or 7 post-ablation) imaging or other modalities (e.g. MRI or CEUS) may be useful surrogates when thoroughly validated but are not able to reduce the need for an additional treatment session. Thereby, to the author’s opinion, intra-interventional AM assessment should become the standard. In the coming years, larger prospective studies are needed to evaluate the effects of ablative margin analysis on primary treatment efficacy and safety and firmly establish a clinically adopted methodology leading to optimized outcomes of liver thermal ablation. Ultimately, these strategies and their clinical consequences need to be integrated to come to well-evidenced recommendations on intra-interventional margin evaluation, indications for immediate or repeat re-ablation or intensified follow-up in high-risk patients based on the achieved ablative margin.

Article highlights

• Thermal ablation is an established treatment in early stage hepatocellular carcinoma (HCC) and low-volume colorectal liver metastases (CRLM), however reported local recurrence rates remain high.

• In recent years, the ablative margin (AM) has been shown a main predictor of local tumor progression (LTP), but there is no consensus on the optimal margin threshold to avoid LTP.

• Current evidence shows achieving an AM >5 mm to reduce the risk of LTP with 16% points for HCC and 47% points for CRLM as compared to AM < 5mm.

• Included studies however showed high variability in AM measurement methodology in terms of measurement technique, imaging modalities, and timing.

• Standardization of AM measurement and reporting is critical to allow future meta-analyses and improved identification of optimal threshold value for clinical use.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper received no funding.

Appendix

Search queries

PubMed

(‘Liver Neoplasms’[Mesh] OR Hepatocellular Carcinoma*[tiab] OR Liver Cell Carcinoma*[tiab] OR Hepatoma*[tiab] OR Colorectal liver metastas*[tiab] OR metastatic liver cancer*[tiab] OR liver tumor[tiab] OR liver tumor[tiab])

AND

(‘Ablation Techniques’[Mesh] OR ablation[tiab])

AND

(image guided [tiab] OR image-guided [tiab] OR magnetic resonance [tiab] OR ‘Magnetic Resonance Imaging’[Mesh] OR ‘Tomography, X-Ray Computed’[Mesh] OR x-ray[tiab] OR x-rays[tiab] OR CECT[tiab] OR CT-guided[tiab] OR ‘ultrasonography, interventional’[MeSH Terms] OR US[tiab] OR ultrasonography[tiab] OR ultrasound[tiab])

Web of Science

(‘Liver Neoplasms’ OR ‘Hepatocellular Carcinoma*’ OR ‘Liver Cell Carcinoma*’ OR Hepatoma* OR ‘Colorectal liver metastas*’ OR ‘metastatic liver cancer*’ OR ‘liver tumor’ OR ‘liver tumor’)

AND

(‘Ablation Techniques’ OR ablation)

AND

(‘image guided’ OR image-guided OR ‘magnetic resonance’ OR ‘Magnetic Resonance Imaging’ OR ‘Tomography, X-Ray Computed’ OR x-ray OR x-rays OR CECT OR CT-guided OR ‘ultrasonography, interventional’ OR US OR ultrasonography OR ultrasound)