Automated Segmentation of CT-guided Ablation Margins in Colorectal Liver Metastases: How We Move Forward to a Standardization of Ablation Zone Assessment

See also the article by Lin et al in this issue.

Dr Sofocleous is a professor of interventional radiology at Weill-Cornell Medical College and practices interventional oncology at Memorial Sloan Kettering Cancer Center in New York City. He has extensive experience in image-guided therapies in cancer and is a clinical scientist with a focus on thelocal and local-regional treatment of colorectal metastatic disease, especially in the liver and lungs. His work is dedicated to the development, validation, and implementation of tissue and image biomarkers for assessment of treated colorectal hepatic metastases with thermal ablation and radioembolization. He has developed methods to assess the ablation zone and margins with the use of stereotactic software, real-time metabolic imaging, and dedicated immunohistochemical and fluorescent assays. Dr Sofocleous is the lead investigator in ongoing clinical trials assessing treatment effects of thermal ablation and radioembolization aiming to improve oncologic outcomes after image-guided therapies for colorectal hepatic metastases.

I read with great interest the work of Lin et al (1) in this issue of Radiology. Their study describes assessment of the ablation zone and minimal ablative margin (MAM) in colorectal liver metastases (CLMs) using deformable CT image registration and artificial intelligence (AI)–based autosegmentation. In their single-center retrospective study of 124 patients with 213 ablated CLMs treated with CT-guided microwave or radiofrequency ablation, the authors applied a biomechanical deformable CT image registration method with AI-based autosegmentation. Their results showed no local disease progression with an MAM greater than or equal to 5 mm. An MAM of 0 mm (subdistribution hazard ratio, 23.3; P < .001) was the dominant independent factor associated with local disease progression in the multivariable analysis. The majority (83%) of local disease progression was spatially localized within areas considered as tissue at risk for postablation tumor progression (MAM <5 mm) using a three-dimensional (3D) ray-tracing method.

The authors correctly identified the MAM as the most critical factor impacting local tumor control of CLMs treated with thermal ablation (2–4). The study validated prior data indicating that an MAM of at least 5 mm all around the target tumor is an absolute minimum requirement for ablation of cancerous liver metastases with local curative intent (4). Similar to prior publications using CT anatomic landmarks (2–4) and 3D software (5–7) for assessment of the ablation zone and MAM, there was concordance of local tumor progression (LTP) with ablative margins of less than 5 mm. This is an area with insufficient coverage that is at risk for LTP. The Lin et al (1) study provides additional support for the creation of a 5-mm MAM as a minimum requirement and a critical end point for acceptable local tumor control by thermal ablation, regardless of the energy source used (radiofrequency or microwave) (3,4). Yet, lacking the actual proportion of margins, not only greater than 5 mm but greater than 10 mm, limits the value of the finding that a 5-mm MAM is a sufficient end point for optimal local tumor control.

Prior studies have indicated no LTP for tumors ablated with an MAM greater than 10 mm regardless of the energy used (3,4). Specifically, for both radiofrequency and microwave ablation, prior studies showed optimal local tumor control without any LTP when the ablation zone covered the target CLMs with an MAM greater than 10 mm. Prior 3D assessments support that the MAM required to achieve local tumor control is perhaps less than what has been reported with the anatomic-landmarks method (10 mm). However, the exact cutoff remains unclear as most of these studies failed to identify how many of the MAMs greater than 5 mm were greater than 10 mm (6,7). This limitation may also explain the discrepancy of MAM association with KRAS mutation in the current study.

Unlike prior series where KRAS mutant CLM had a higher rate of LTP even with a 10-mm MAM (8), the study by Lin et al (1) did not show any association between KRAS status and the incidence of LTP. It is possible that this lack of association is the result of larger ablation zones with MAMs more than 10 mm that have not been captured. The 10-mm MAM in two-dimensional assessments has been shown to provide absolute tumor control both for radiofrequency and microwave ablation (3,4). But this ablative margin of 10 mm was also associated with an increased risk of biliary complications in patients at high risk; specifically, those previously treated with hepatic arterial chemotherapy, prior bevacizumab, and preexisting biliary dilation in the preablation scans (4). Therefore, it is very desirable to determine the absolute minimum size of the MAM that can offer local disease control with minimal impact on the surrounding hepatic parenchyma. Such strategies include intraoperative assessments of the ablation zone using 3D software (5–7), the interrogation of the ablation zone center, and MAM measurement with biopsy (9,10) to confirm complete tumor eradication and the absence of residual tumor cells at the end of ablation. These approaches have shown that a 5-mm margin accompanied by a tumor-negative ablation zone center and margins (ie, no detected viable cancer cells in the ablation zone) provides sustained local tumor control with a local progression-free survival rate of 93% 12 months after thermal ablation (9). The use of intraoperative real-time tissue examinations was associated with local tumor control (10). These real-time examinations can offer added certainty of complete tumor eradication or detect residual tumor that can be reablated on the spot, similar to the use of frozen-section pathologic examinations during resections (10).

Intraoperative use of 3D software assessments of the ablation zone is lacking in the literature. Thus, the results of the current study and other similar work indicate that 3D assessments have much better discrimination power for detecting areas at risk for subsequent LTP within the ablation zone (5–7). All available software platforms depend on the accuracy of liver and tumor registration, which may require manual adjustments prior to assessment (1,6). In a prior work evaluating ablation confirmation software (NeuWave; Ethicon), manual adjustments to improve registration were needed in 24% of cases (6). The deformable CT registration method with autosegmentation presented by Lin et al also required manual adjustments. However, the exact number of cases needing manual adjustments is not reported. In addition, all software platforms carry a standard error that remains unknown and probably depends on the input CT quality and section thickness. For example, an MAM of 2–3 mm is unreliable if it is based on 5-mm thickness input CT scans, whereas it may be very accurate with a 1-mm thickness. In addition, when we measure a 5-mm margin, we are often referring to a categorical range rather than an absolute number because we do not know the standard error. It is for this reason that all prior papers using anatomic landmarks, as well as subsequent work using 3D software platforms, have reported the margins in categories of 0 mm, 1–5 mm, greater than 5 mm and less than 10 mm, and greater than or equal to 10 mm (2–6).

In conclusion, the Lin et al study (1) has correctly identified the MAM (at least 5 mm) as the most critical factor impacting local tumor control of CLMs treated with thermal ablation. But more standardization of pre- and postablation imaging used for assessment of ablation zones and MAM localization is necessary. Further validation of measurements is also needed in subsequent prospective trials. A reproducible technical end point for curative intent tumor ablation is absolutely needed and attainable. A required MAM of 5 mm all around the target CLM is strongly supported by the existing level of evidence. More data are expected from the recently announced ACCLAIM clinical trial (Ablation with Confirmation of Colorectal Liver Metastases; ClinicalTrials.gov registration no. NCT05265169). The ACCLAIM trial will assess microwave ablation as a local cure for CLM. This prospective multicenter international trial by the Society of Interventional Oncology will validate the impact of the MAM on LTP-free survival through the mandatory use of intraoperative real-time 3D ablation zone and MAM assessment with independent readers.