Creating Conductivity Maps at 200 Khz of Brain and Tumor Tissue of Glioblastoma Patients with Water-Content Based Electric Properties Tomography

Abstract

Computational head models are employed to study the electric field distributions in the head of glioblastoma (GBM) patients receiving Tumor Treating Fields (TTFields) therapy with a device operating at 200 kHz. The distribution of underlying electric properties shapes the electric field, which are usually assigned as homogeneous values according to tissue type. However, tissue segmentation is daunting and time-consuming specifically with pathologies present. Even more, the electric conductivity (σ) of tumor tissue shows great heterogeneity and strong variation among patients. For patient-specific treatment planning it would be highly desirable to rapidly produce individual head models that accurately predict the heterogeneous and individual σ distribution within the tumor region. Here, we investigate an adaptation of the water-content based electric properties tomography (wEPT) technique in order to create and compare σ maps of 3 GBM patients. The original wEPT approach maps σ at 128 MHz as monotonic function of tissues’ water-content which is estimated with a transfer function from the image ratio of two MRIs sharing the same sequence parameters except for either having a short or a long repetition time. In a previous animal study we established a model adaption for σ estimation at 200 kHz. One input MRI might be a conventional T1w sequence and the other resembles a proton density (PD) image, also often acquired when scanning GBM patients. From the EF14 trial database 3 patients were selected according to quality of MRIs at baseline and tumor composition. Maps of water-content and EPs in the brain and tumor were created for the 3 patients and values were compared between different tissue types in the healthy brain (gray and white matter) and the tumor area (enhancing and non-enhancing part and necrosis). The wEPT-estimated values of water-content and σ in the healthy brain agree well with reported literature values. They are reasonably homogeneous and consistent among patients. Contrary, wEPT-estimates of water-content and σ in tumor tissues are very heterogeneous (greater standard deviation compared to healthy brain) with strong local differences within tumors and large variability between patients. The predicted results emphasize the need for patient-specific individual head model creation. It can be concluded that binary segmentation masks with assigned pre-defined σ values are not recommended for the heterogeneous and variable tumor tissues. However, this method needs to be validated with actual σ measurements of tumor tissues at 200 kHz and more patients need to be studied. Yet, the presented approach holds great promise for rapid creation of patient-specific computational head models, since only conventional MRIs are used without the need of complex tissue segmentation.