Reid Master

Background: Glioblastoma (GBM) is the most aggressive adult primary brain tumor, with a median survival of 14–16 months despite multimodal treatment involving surgery, radiotherapy (RT), and chemotherapy. Poor prognosis stems from diffuse infiltration, resistance to standard treatments, and difficulty in defining tumor margins due to interobserver variability and microscopic spread beyond visible lesions¹. Traditional RT relies on anatomical imaging and expert delineation of gross tumor volume (GTV), clinical target volume (CTV), and planning target volume (PTV). Despite modern RT precision on the scale of millimeters, CTV uncertainty remains on the centimeter scale. Artificial Intelligence (AI) shows promise in improving tumor delineation, automating segmentation, and optimizing dose delivery with deep learning using imaging modalities such as MRI and PET. However, most models remain in early phases with small training datasets, lacking clinical deployment and survival data². This review addresses the gap between current RT practices and emerging AI tools aiming to enhance precision while reducing toxicity in GBM management.

Methods: A PubMed search was conducted using the key terms “radiotherapy planning,” “artificial intelligence,” and “glioblastoma.” Highly relevant papers were added from review articles to provide comprehensive coverage.

Results: Multiple studies demonstrated that convolutional neural networks (CNNs), particularly U-Net architectures, achieved high Dice similarity coefficients (DSC), outperforming manual segmentation in consistency and efficiency^3,4. A CNN using diffusion tensor imaging enhanced CTV estimation following EORTC and RTOG guidelines⁵. Machine learning (ML) models trained on clinical datasets generated RT plans under 30 minutes, with comparable coverage and reduced mean brain dose⁶. One Bayesian ML framework incorporated MRI and FET-PET to infer tumor cell densities, enabling dose escalation to radioresistant regions while sparing healthy tissue⁷. Predictive models showed AUCs from 0.74–0.91, with up to 75% accuracy in forecasting RT response⁸. However, all referenced studies used single-center data, typically with fewer than 40 patients, raising overfitting concerns and limiting generalizability².

Conclusions: AI integration into RT planning for GBM has shown early promise in segmentation, dose prediction, and treatment personalization. These tools may mitigate interobserver variability, shorten planning time, and enable biologically informed dose escalation. Despite technological advances, no AI models are yet clinically approved⁴. Future work should prioritize multi-institutional datasets, prospective validation, and incorporation of the personalized precision care model to integrate data from epigenetics, wearables, and molecular data⁹. RCT must be conducted to validate whether models have any notable impact on patient outcomes.

Works Cited:

1. Kanderi, T., Munakomi, S. & Gupta, V. Glioblastoma Multiforme. in StatPearls (StatPearls Publishing, Treasure Island (FL), 2025).
2. d’Este, S. H., Nielsen, M. B. & Hansen, A. E. Visualizing Glioma Infiltration by the Combination of Multimodality Imaging and Artificial Intelligence, a Systematic Review of the Literature. Diagnostics 11, 592 (2021).
3. Lu, S.-L. et al. Randomized multi-reader evaluation of automated detection and segmentation of brain tumors in stereotactic radiosurgery with deep neural networks. Neuro-Oncol. 23, 1560–1568 (2021).
4. Di Nunno, V. et al. Machine learning in neuro-oncology: toward novel development fields. J. Neurooncol. 159, 333–346 (2022).
5. Peeken, J. C. et al. Deep learning derived tumor infiltration maps for personalized target definition in Glioblastoma radiotherapy. Radiother. Oncol. 138, 166–172 (2019).
6. Tsang, D. S. et al. A pilot study of machine-learning based automated planning for primary brain tumours. Radiat. Oncol. 17, 3 (2022).
7. Lipkova, J. et al. Personalized Radiotherapy Design for Glioblastoma: Integrating Mathematical Tumor Models, Multimodal Scans, and Bayesian Inference. IEEE Trans. Med. Imaging 38, 1875–1884 (2019).
8. Gutsche, R. et al. Radiomics outperforms semantic features for prediction of response to stereotactic radiosurgery in brain metastases. Radiother. Oncol. 166, 37–43 (2022).
9. Velu, U. et al. Precision Population Cancer Medicine in Brain Tumors: A Potential Roadmap to Improve Outcomes and Strategize the Steps to Bring Interdisciplinary Interventions. Cureus (2024) doi:10.7759/cureus.71305.