Impact of Model Inaccuracy on Dose Estimation in TTFields Therapy

Abstract

<h3>Purpose/Objective(s)</h3> Tumor Treating Fields (TTFields) is a modality for treating Glioblastoma (GBM). A recent study combining clinical data and simulations demonstrated that the simulation-based dose estimation at the tumor level is directly correlated with patient survival (Ballo M, et al. <b>Int J Radiat Oncol Biol Phys</b>. 2019;104(5):1106-1113). This study highlights the importance of computer-modelling based TTFields treatment planning in the clinic. Performing such planning in a meaningful manner requires an understanding about how inaccuracies in the patient model influence calculated TTFields distributions. In this study we show the effect of local perturbations in the model, errors in tumor segmentation, and inaccuracy in the patient head model on TTFields dose estimation. <h3>Materials/Methods</h3> Computational studies were performed with software. 1) Local model uncertainty outside of tumor region: To create defects in the models, conductive spheres with varying conductivities and radii were placed into the model's brains at different distances from the tumor. Virtual transducer arrays were placed on the models, and delivery of TTFields numerically simulated. The error in the electric field induced by the defects as a function of defect conductivity, radius, and distance to tumor was investigated. 2) Tumor segmentation uncertainty: For illustrating segmentation errors in tumors and necrotic cores, layered spheres with various conductivities and radii were placed in different head locations. Tumor conductivity uncertainty was defined as (σ-σ<b>_ref</b>)/σ<b>_ref</b> where σ<b>_ref</b> = 0.24 S/m is the standard tumor conductivity and σ is the varied conductivity over the tumor. The normalized tumor radius was calculated when each of the varied radii was normalized to the results obtained by all the other combinations. <h3>Results</h3> The results show that that when a defect of radius R is placed at a distance, d, from the tumor that is larger than seven times R, the error is below 1% regardless of the defect conductivity. <h3>Conclusion</h3> Our models and results show the impact of uncertainty in segmentation of the tumor or other tissue distant from it on the TTFields dose estimation. These results could serve as a guideline for model creation and tissue segmentation that would lead to optimal dose estimation in TTFields treatment planning.