Therapeutic Hybrid Intelligence with Neural and Knowledge-based Expert Reasoning for SRS (THINKERS): an AI model for...

Prescription dose selection for recurrent glioblastoma treated with Gamma Knife radiosurgery remains highly individualized and is largely guided by tumor volume, anatomical constraints, prior radiation exposure, and physician judgment rather than patient-specific estimates of second progression risk. We developed THINKERS-GBM, a mixture-of-experts artificial intelligence framework for personalized dose evaluation after Gamma Knife radiosurgery.

We performed a retrospective single-center patient-level study of recurrent glioblastoma treated with Gamma Knife radiosurgery. Variables available before or at treatment were used to train a mixture-of-experts neural network with discrete-time survival modeling. Prescription dose was incorporated as a queryable input to enable repeated candidate dose evaluation. Internal validation used grouped 5-fold cross-validation and a grouped holdout test split. Performance was assessed using area under the receiver operating characteristic curve (AUC) for 12-month second progression, mean absolute error (MAE) for expected time to progression, Brier score, and calibration metrics.

The cohort included 200 patients with recurrent glioblastoma. In grouped cross-validation, THINKERS-GBM achieved a raw mean AUC of 0.828 and calibrated mean AUC of 0.839 for 12-month second progression. In the grouped holdout set, the calibrated AUC was 0.870 (95% CI, 0.80-0.93); the calibrated Brier score was 0.009 (95% CI, 0.006-0.019), and MAE for expected time to progression was 1.03 months (95% CI, 0.83-1.95). Dose-sweeping generated individualized dose-response profiles.

THINKERS-GBM provides an internally validated framework for second progression prediction and dose-policy evaluation after Gamma Knife radiosurgery for recurrent glioblastoma. External validation is required before clinical deployment.

Not applicable.