Creating individually computed head models to simulate TTFields distribution
PO-1357
Abstract
Creating individually computed head models to simulate TTFields distribution
Authors: BOMZON|, Zeev(1)*[zbomzon@novocure.com];Kinzel|, Adrian(2);Urman|, Noa(1);Levi|, Shay(1);Naveh|, Ariel(1);Manzur|, Doron(1);Hershkovich|, Hadas Sara(1);
(1)Novocure Ltd., Research and Development, Haifa, Israel;(2)Novocure GmbH, Medical, Munich, Germany;
Show Affiliations
Hide Affiliations
Purpose or Objective
Tumor Treating Fields (TTFields) are locally applied alternating electric fields of intermediate frequency used to treat glioblastoma multiforme (200 kHz). Currently, TTFields are also investigated in clinical phase III trials in other solid cancers. TTFields distribution in the tissue is affected by the position of the transducer arrays delivering the therapy, but also by the patient’s anatomy and the electric properties of tissue and tumor. For investigating the influence of TTFields distribution on patient outcome, we aimed to design realistic, patient-specific computational head models in a rapid manner that is robust even when MRI image quality is restricted.
Material and Methods
We created our patient models using a detailed head model of a healthy person as a deformable template. After pre-processing (denoising and reducing background noise, if needed: super-resolution algorithms), we manually segmented and masked the tumor to leave only healthy tissue in the MRI. This is then registered to the template space transforming the patient space to template space. In the next step, the template is deformed into the patient space by inverse transformation before placing back the tumor to create the full patient model. Landmarks on the patient’s head are automatically identified for positioning of the transducer arrays to include them into the model. We then simulated field distribution with the Finite Elements Method (Sim4Life V3.0, ZMT-Zurich).
Results
We simulated distribution in 340 TTFields-treated patients of the EF-14 trial that led to the therapy approval in GBM. Our method enables to accurately contour tissues known to have a great impact on the electric field distribution such as the skull, scalp, CSF, or ventricles, a fundamental prerequisite for the subsequent study investigating correlation of TTFields spatial distribution and patient outcome.
Conclusion
Our method for fast patient-specific model creation enabled us to rapidly create patient-specific head models even when the quality of the available MRI images was low. This ultimately prepared the ground for investigations of correlations between spatial electric field distribution and disease progression.