Abstract
Introduction: TTFields are an anti-mitotic therapy designed to target solid tumors. TTFields are FDA approved for the treatment of glioblastoma multiforme (GBM) and multi-pleural mesothelioma (MPM). Delivery of TTFields in GBM patients is done by placing two pairs of transducer array on the patient's scalp. Previous studies has shown that the location of the arrays may have significant influence on the field distribution within the patient's head and tumor. Planning the placement of the arrays can hence provide better prognosis for patients. As of today, treatment of GBM patients is performed using the NovoTAL system. A new generation of patient specific planning software is currently under development. In order to plan the treatment, a computerized model is created from the patient's T1 MRI, and various options for arrays layout are simulated. The system can then asses which layout would provide the higher TTFields dose. The accuracy of these simulations may, however, depend on the accuracy of the model. Therefore there is a need to quantify how errors affect TTFields intensity distribution within the model.
Methods: To test the effect of model accuracy on TTFields distribution we used a patient specific model (prepared in house) which consisted of the following tissues: scalp, skull, cerebrospinal fluid (CSF), grey matter (GM) and white matter (WM). These tissues were automatically segmented using an in-house algorithm utilizing the MARS and SPM Matlab packages. In addition, the tumor, necrotic core and scar tissue were manually segmented into the model. To test the effect of segmentation errors on the field distribution in the tumor, we purposely inserted large defects into the model in the form of conductive (2 S/m) spheres placed at various locations within the model (i.e. WM, GM and CSF). Simulations were performed using Sim4Life v4.4 (ZMT Zurich), and the effect of the defects on the field distribution within the brain and tumor was assessed.
Results: Comparison of field distributions in models into which a defect had been introduced with field distributions in the model without the defects revealed that the defects into the model perturbed the field in close vicinity to the defect s with little to no effect on the field distribution in regions that were ata distant from the defect. Typically, introducing a metallic sphere with a diameter of 2cm into altered field intensities up to a distance of abut 4cm from the center of the sphere, with little to no effect on field distributions beyond this distance.
Discussion: The introduction of a metallic ball into the model represents the introduction of a very large inaccuracy in segmentation. Nonetheless, the perturbations caused by such inaccuracies seems to influence field distribution only in the immediate vicinity of the defect (up to 2cm away from it's edge). This suggests that when creating patient-specific models for planning TTFields therapy utilizing simulations, it is important to accurately segment the tissue in the vicinity of the tumor, whereas inaccuracies in regions distant from the tumor are unlikely to affect the field distribution in and around the tumor, and hence unlikely to affect the results of a planning process aimed at maximizing the field intensity within that target region.
Citation Format: Ze'ev Bomzon, Ariel Naveh, Oshrit Ze'evi, Hadas S. Hershkovich. The influence of computational model inaccuracies on calculated electric field distributions when simulating TTFields therapy [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5498.
