gms | German Medical Science

Introduction: Radiofrequency ablation (RFA) is a clinically established minimally invasive therapy option for liver metastases. However, a high local recurrence rate due to the size of the tumor, the vascular cooling effect of nearby liver vessels and the individual treatment experience limits its curative application. In preliminary studies a numeric simulation software for the calculation of the expected ablation zones was established. The planning system aids the decision for choosing suitable applicators, electric parameters and positioning of the applicators for achieving complete tumor necrosis.

Aim of this study was to validate a simulation software in consideration of vascular cooling effect in multipolar radiofrequency ablation in ex situ experiments.

Material and methods: 36 multipolar radiofrequency ablations were performed in porcine livers ex vivo with three bipolar, internally cooled applicators. The ex vivo ablation zones were compared with the calculated lesions of the planning software. In order to simulate a perfused vessel, a glass tube (3 mm inner diameter, 100 ml/min flow volume, H₂0) was placed in the liver. Five different applicator-vessel-geometries, as well as a dry run without flow, were analyzed. Thermolesions were digitally and planimetrically measured along the cross-section area at the level of the largest lesion diameter, orthogonally to the applicators. The ex vivo lesions (A) and the lesions calculated by the planning software (M) were compared on the basis of the Dice Coefficient (2*(A∩M)/(A+M)), the Volumetric Overlap ((A∩M)/(AUM)), as well as the inner (false positive) and the outer surface distance (false negative) in relation to the simulated surface.

Results: A comparison between the ex situ ablation zones and the predicted lesions resulted in an average congruency of 0.85 (Dice coefficient) respectively 0.92 (Volumetric Overlap). The average nonconformance was 13 %. The simulation software overestimated the ex situ ablation zone by 5 % (false positive). The average surface distance was 0.5 mm, whereas the maximum false positive distance was 1.6 mm. The false negative error, in which the software underestimated the ablation zone, was 12 %, whereas the mean surface distance was 1.0 mm and the maximum distance was 2.7 mm.

Conclusion: The software is able to predict a radiofrequency ablation in consideration of the vascular cooling effect. The simulation underestimates the lesions slightly. However, the average calculation error, measured as interfacial distance, stays in the range of the physical resolution of a conventional CT image. The pre-therapeutic calculation of the expected ablation zone might be a valuable tool for planning a multipolar RFA in consideration of the vascular cooling effect of liver vessels.