Optimized Microwave Ablation With a Novel Applicator: Integration of Taguchi Neural Networks for Enhanced Predictive Accuracy of Ablation Zone

Abstract

This study examines the computational challenges associated with modeling liver tumors using microwave ablation (MWA), while highlighting the limitations of conventional methods and advocating for the use of MWA in conjunction with artificial intelligence as a more promising approach. The proposed innovative antenna design, which comprises a coaxial line featuring a tapered outer conductor and a dipole antenna, aims to produce a nearly spherical ablation zone without the need for any additional matching network. Capable of operating at both 2.45 GHz and 5.8 GHz with minor structural modifications, it offers flexibility in tumor ablation systems. The research further incorporates and compares the sigmoidal model, a well-established computational method, and a recently developed parametric model for evaluating temperature-dependent properties in modeling the 3-D liver tissue, identifying differences in the ablation zone during MWA. Additionally, since both under and over ablation are major concerns during the MWA procedure, resulting in damage to healthy tissue and tumor recurrence, respectively, this study introduces a Taguchi Artificial Neural Networks (TNN) framework for the prediction of ablation zone in advance, thereby, significantly reducing the number of required training datasets without compromising performance metrics.