Skip to main content

Development and evaluation of a unified machine-learning model for predicting normal brain toxicity in single isocenter multi-target stereotactic radiosurgery

Journal articles  - Journal Article
Huang, Z; Zhao, J; Lu, K; Ginn, JS; Yang, Z; Xie, Y; Kim, Y; Mullikin, TC; Wang, C
Published in: Journal of Radiosurgery and Sbrt
January 1, 2026

Purpose: To develop and evaluate a machine learning model for normal brain dose outcome in linear-accelerator-based single isocenter multiple target (SIMT) stereotactic radiosurgery (SRS). Methods: SIMT SRS cases collected totaled 231, and 181/50 cases were randomly assigned for model training/testing. The model characterizes SIMT cases using seven geometric and statistical features describing prescription, target burden, and inter-target spatial relationships. Gradient boosted tree (GBT) regression was employed to predict three dosimetric outcomes: overall V50%, V60%, and V66.7%. Five hyperparameters were optimized through grid search and 10-fold cross-validation. Four models were trained: the first model (Ma,s) predicts all three outcomes as a single output, and the second model (Ma,m) uses three GBTs to report each outcome. In the third (Mc,s) and fourth (Mc,m) models, unsupervised k-means cluster partitioned training samples into geometrically distinct groups based on statistical features, and the models were trained using clustered samples for single or multiple outputs, respectively. Model performance was evaluated using normalized mean absolute error (MAE), mean prediction uncertainty (MPU), and R2. Results: Ma,s achieved the best overall performance, with an average MAE of 6.12cm3, MPU of 7.81cm3 and R2 = 0.88. Clustering slightly improved relative error in Mc,m but increased absolute errors and reduced overall performance in Mc,s. Each prediction required under 5 s. An Eclipse Scripting Application Programming Interface script was developed to integrate the model into the Eclipse™ treatment planning system. Conclusions: The machine-learning framework was developed and validated for predicting SIMT SRS normal-brain dose. By providing accurate and reliable dosimetric predictions, the model supports efficient treatment planning and reduces plan quality variability.

Duke Scholars

Altmetric Attention Stats
Dimensions Citation Stats

Published In

Journal of Radiosurgery and Sbrt

EISSN

2156-4647

ISSN

2156-4639

Publication Date

January 1, 2026

Volume

10

Issue

3

Start / End Page

165 / 174
 

Citation

APA
Chicago
ICMJE
MLA
NLM
Huang, Z., Zhao, J., Lu, K., Ginn, J. S., Yang, Z., Xie, Y., … Wang, C. (2026). Development and evaluation of a unified machine-learning model for predicting normal brain toxicity in single isocenter multi-target stereotactic radiosurgery. Journal of Radiosurgery and Sbrt, 10(3), 165–174.
Huang, Z., J. Zhao, K. Lu, J. S. Ginn, Z. Yang, Y. Xie, Y. Kim, T. C. Mullikin, and C. Wang. “Development and evaluation of a unified machine-learning model for predicting normal brain toxicity in single isocenter multi-target stereotactic radiosurgery.” Journal of Radiosurgery and Sbrt 10, no. 3 (January 1, 2026): 165–74.
Huang Z, Zhao J, Lu K, Ginn JS, Yang Z, Xie Y, et al. Development and evaluation of a unified machine-learning model for predicting normal brain toxicity in single isocenter multi-target stereotactic radiosurgery. Journal of Radiosurgery and Sbrt. 2026 Jan 1;10(3):165–74.
Huang Z, Zhao J, Lu K, Ginn JS, Yang Z, Xie Y, Kim Y, Mullikin TC, Wang C. Development and evaluation of a unified machine-learning model for predicting normal brain toxicity in single isocenter multi-target stereotactic radiosurgery. Journal of Radiosurgery and Sbrt. 2026 Jan 1;10(3):165–174.

Published In

Journal of Radiosurgery and Sbrt

EISSN

2156-4647

ISSN

2156-4639

Publication Date

January 1, 2026

Volume

10

Issue

3

Start / End Page

165 / 174