Speaker
Description
Glioblastoma (GBM) is a highly invasive brain tumor, whose cells infiltrate surrounding normal brain tissue beyond the lesion outlines visible in the current medical scans. Predicting GBM infiltration is critical for designing radiotherapy treatment plans because GBM recurrence is largely driven by diffuse tumor infiltration. However, standard radiotherapy, the mainstay for treating this diffuse infiltration, relies on uniform expansions that neglect patient specific biological and anatomical factors. Mathematical models can complement the data by predicting spatial distributions of tumor cells beyond the visible margins. This requires estimating patient specific parameters of the model from clinical data, which is a challenging inverse problem due to limited temporal data and the limited time between imaging and diagnosis. Here, we discuss biophysical growth models and methods for solving the inverse problem, including new scientific machine-learning methods PINN-GBM and BiLO \cite{zhang_personalized_2025, ZHANG2026114679} that use physically-informed neural networks and GLIODIL \cite{balcerak_individualizing_2025}, which integrates traditional numerical methods with data driven paradigms. Using a newly developed open-source platform (PREDICT-GBM) that integrates a curated, longitudinal dataset of 255 patients with a unified evaluation pipeline \cite{noauthor_brainlesion/predictgbm_2026,zimmer2026predictgbmmulticenterplatformadvance}, we find that the biophysical models significantly outperform standard-of-care protocols in predicting future recurrence sites and demonstrate greater robustness compared to purely data-driven recurrence prediction methods.
Bibliography
@article{zhang_personalized_2025,
title = {Personalized predictions of {Glioblastoma} infiltration: {Mathematical} models, {Physics}-{Informed} {Neural} {Networks} and multimodal scans},
volume = {101},
issn = {13618415},
shorttitle = {Personalized predictions of {Glioblastoma} infiltration},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1361841524003487},
doi = {10.1016/j.media.2024.103423},
language = {en},
urldate = {2026-03-28},
journal = {Medical Image Analysis},
author = {Zhang, Ray Zirui and Ezhov, Ivan and Balcerak, Michal and Zhu, Andy and Wiestler, Benedikt and Menze, Bjoern and Lowengrub, John S.},
month = apr,
year = {2025},
pages = {103423},
}
@article{ZHANG2026114679,
title = {BiLO: Bilevel Local Operator Learning for PDE Inverse Problems},
journal = {Journal of Computational Physics},
volume = {551},
pages = {114679},
year = {2026},
issn = {0021-9991},
doi = {https://doi.org/10.1016/j.jcp.2026.114679},
url = {https://www.sciencedirect.com/science/article/pii/S002199912600029X},
author = {Ray Zirui Zhang and Christopher E. Miles and Xiaohui Xie and John S. Lowengrub},
keywords = {Bilevel optimization, PDE inverse problems, Neural operators, Scientific machine learning},
}
@article{balcerak_individualizing_2025,
title = {Individualizing glioma radiotherapy planning by optimization of a data and physics-informed discrete loss},
volume = {16},
issn = {2041-1723},
url = {https://www.nature.com/articles/s41467-025-60366-4},
doi = {10.1038/s41467-025-60366-4},
language = {en},
number = {1},
urldate = {2026-03-28},
journal = {Nature Communications},
author = {Balcerak, Michal and Weidner, Jonas and Karnakov, Petr and Ezhov, Ivan and Litvinov, Sergey and Koumoutsakos, Petros and Amiranashvili, Tamaz and Zhang, Ray Zirui and Lowengrub, John S. and Yakushev, Igor and Wiestler, Benedikt and Menze, Bjoern},
month = jul,
year = {2025},
pages = {5982},
}
@misc{noauthor_brainlesion/predictgbm_2026,
title = {{BrainLesion}/{PredictGBM}},
copyright = {Apache-2.0},
url = {https://github.com/BrainLesion/PredictGBM},
abstract = {Tools for image processing, brain segmentation and evaluation of glioblastoma growth models},
urldate = {2026-03-28},
publisher = {BrainLesion Suite},
month = mar,
year = {2026},
note = {original-date: 2025-07-09T08:35:26Z},
keywords = {mri, brain, medical-image-processing, glioblastoma},
}
@misc{zimmer2026predictgbmmulticenterplatformadvance,
title={PREDICT-GBM: A multi-center platform to advance personalized glioblastoma radiotherapy planning},
author={L. Zimmer and J. Weidner and M. Balcerak and F. Kofler and M. Krupa and I. Ezhov and S. Cepeda and R. Zhang and J. Lowengrub and B. Menze and B. Wiestler},
year={2026},
eprint={2509.13360},
archivePrefix={arXiv},
primaryClass={eess.IV},
url={https://arxiv.org/abs/2509.13360},
}
