Optimization of tumor treating fields using singular value decomposition and minimization of field anisotropy

Tumor treating fields (TTFields) are increasingly used to treat newly diagnosed and recurrent glioblastoma (GBM). Recently, the authors proposed a new and comprehensive method for efficacy estimation based on singular value decomposition of the sequential field distributions. The method accounts for all efficacy parameters known to affect anti-cancer efficacy of TTFields, i.e. intensity, exposure time, and spatial field correlation. In this paper, we describe a further development, which enables individual optimization of the TTFields activation cycle. The method calculates the optimal device settings to obtain a desired average field intensity in the tumor, while minimizing unwanted field correlation. Finite element (FE) methods were used to estimate the electrical field distribution in the head. The computational head model was based on MRI data from a GBM patient. Sequential field vectors were post-processed using singular value decomposition. A linear transformation was applied to the resulting field matrix to reduce fractional anisotropy (FA) of the principal field components in the tumor. Results were computed for four realistic transducer array layouts. The optimization method significantly reduced FA and maintained the average field intensity in the tumor. The algorithm produced linear gain factors to be applied to the transducer array pairs producing the sequential fields. FA minimization was associated with an increase in total current delivered through the head during a activation cycle. Minimized FA can be obtained for an unchanged total current level, albeit with a reduction in average field intensity. We present an algorithm for optimization of the TTFields activation cycle settings. The method can be used to minimize the spatial correlation between sequential TTFields, while adjusting the total current level and mean field intensity to a desired level. Future studies are needed to validate clinical impact and assess sensitivity towards model parameters.