Abstract Objective: Tumor Treating Fields (TTFields) is an emerging cancer therapy whose efficacy is closely linked to the electric field (EF) intensity delivered to the tumor. However, current computational workflows for simulating the EF and planning treatment rely on time-consuming manual segmentation and proprietary software, hindering efficiency, reproducibility, and accessibility. Approach: We introduce AutoSimTTF, a fully automatic pipeline for personalized EF simulation and optimized treatment planning for TTFields.The end-to-end workflow utilizes advanced deep learning model for automated tumor segmentation, conducts finite element method (FEM)-based EF simulation, and determines a computationally optimized treatment plan via a novel, physics-based parameter optimization method. Main results: The automated segmentation module achieved high precision, yielding a Dice Similarity Coefficient of 0.91 for the whole tumor. In terms of efficiency, the active planning workflow was completed in approximately 12 minutes, significantly outperforming conventional multi-day manual processes. The pipeline’s simulation accuracy was validated against a conventional semi-automated workflow, demonstrating deviations of less than 14.1% for most tissues. Critically, the parameter optimization generated personalized transducer montages that produced a significantly higher EF intensity at the tumor site (up to 111.9% higher) and substantially improved field focality (19.4% improvement) compared to traditional fixed-array configurations. Significance: AutoSimTTF addresses major challenges in efficiency and reproducibility, paving the way for data-driven personalized TTFields therapy and large-scale computational research.
Xie et al. (Thu,) studied this question.