Modeling the involution of microwave liver ablation zones

Abstract

Background: The post-ablation involution of microwave liver ablation zones (AZs) remains poorly understood. This study develops mathematical models to characterize AZ involution and identifies key predictors influencing its dynamics.

Materials and methods: Fifty-four patients (mean age 61 ± 10 years (standard deviation), 33 men) underwent microwave liver ablation (MWA) of 76 liver tumors and follow-up contrast enhanced CT (CECT) imaging in this retrospective single-center cohort study. AZs were segmented on intraprocedural post-ablation portal-venous phase CECT and all available subsequent postprocedural follow-up scans, or until local tumor progression (LTP). Volumetric AZ involution was modeled using non-linear regression methods and correlated with initial tumor and ablation parameters.

Results: In total, 366 AZ segmentations were performed over median 304 days CECT-follow-up (range 21-741). Involution was best modeled by mono-exponential decay (SSE = 4.64, RMSE = 0.11). AZs shrank to one-third of baseline volume within a year, with a half-life of 158 days. At 6 weeks, relative volume was 0.81 of baseline (95% prediction interval 0.59-1.04, 95% confidence interval 0.80-0.83). Variables with a significant effect on involution included initial tumor diameter (p = 0.03), initial AZ volume (p < 0.01), and tumor:AZ volume ratio (p = 0.04).

Conclusions: Microwave ablation zones rapidly involute and stabilize at approximately one-third of their baseline volume within a year. The involution process is best modeled by mono-exponential decay and influenced by the type of tissue ablated. These findings highlight the potential need for predictive models to adjust for involution for follow-up imaging-based margin assessment to optimize accuracy and ablation outcomes.

Keywords: CT; Interventional Radiology; Neoplasms; ablation techniques; liver.