Percutaneous microwave ablation of liver tumors: predicted versus actual ablation zones

Purpose To fully exploit the advantages of ablation planning software, discrepancies between predicted and actual ablation zones should be minimized. This study evaluated the volumetric and geometric correspondence between manufacturer-predicted and actual ablation zones following thermal liver ablation.Methods This multicenter retrospective study included Emprint microwave ablations with intraprocedural needle and post-ablation confirmation CT scans. The needle position was determined on the needle confirmation scan and subsequently registered to the post-ablation contrast-enhanced CT scan. Manufacturer-predicted ablation zones were projected onto segmented actual ablation zones to calculate overlap metrics: Dice similarity coefficient (DSC), average surface distance (ASD), and the under-ablated volume ratio (UAR), defined as the uncovered fraction of the predicted ablation volume. The UAR was correlated with calculated minimal ablation margins (MAM) using Spearman’s rank correlation. The Mann-Whitney U test was used to test if metrics differed for patient-specific factors.Results Forty-seven ablations were analyzed. Median predicted and actual ablation volumes were 23.3 cm3 [IQR: 18.0–27.4] and 19.0 cm3 [13.9–30.1] (p = 0.22), showing moderate correlation (r = 0.57, p < 0.01). Median DSC was 0.72 [0.64–0.79], ASD 3.1 mm [2.5–3.7], and UAR 0.26 [0.18–0.45]. None of these measures differed significantly by cirrhosis, vascular proximity or subcapsular location. UAR showed a moderate negative correlation with the MAM (r = −0.54, p = 0.0018).Conclusion Emprint-predicted ablation zones show moderate correspondence with actual ablation zones, with no significant association observed for liver cirrhosis, vascular proximity, or subcapsular location. The observed association between UAR and MAM underscores the need to refine prediction models to achieve adequate ablative margins.