Therapeutic Hybrid Intelligence with Neural and Knowledge-based Expert Reasoning for SRS (THINKERS): an AI model for melanoma brain metastases
Neurosurgical Review · 04.11.2026 · 49 (1)
DOI: 10.1007/s10143-026-04557-7 ↗Kurz gefasst
Fachartikel in der Zeitschrift Neurosurgical Review, online erschienen am 04.11.2026. Verfasst von Jheremy S Reyes und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Neurological Surgery, Center for Image-Guided Neurosurgery, University of Pittsburgh Medical Center, Pittsburgh, USA. reyesbarretojs@up…. Hauptthemen laut MEDLINE: Brain Neoplasms, Melanoma, Radiosurgery, Artificial Intelligence. Die Arbeit nennt 28 Literaturangaben.
Abstract (Originalsprache)
Background
Dose selection for melanoma brain metastases treated with stereotactic radiosurgery (SRS) remains largely based on generalized practice patterns rather than tumor-specific modeling of local failure dynamics. We developed a Therapeutic Hybrid Intelligence with Neural and Knowledge-based Expert Reasoning for SRS (THINKERS) in Melanoma Brain Metastases, an artificial intelligence framework for prognostic prediction.
Methods
We performed a retrospective single-center study of melanoma brain metastases treated with SRS at UPMC. Only variables available at or before treatment were included. The final model used a mixture-of-experts deep neural network with discrete-time survival modeling. Margin dose was incorporated as an explicit input variable, allowing repeated evaluation across candidate dose levels for tumor-specific dose evaluation. Internal validation consisted of grouped 5-fold cross-validation and a grouped holdout test split by patient.
Results
The final analytic cohort included 141 patients with 884 treated melanoma brain metastases. Median age at treatment was 65 years (IQR, 55–74), median pre-treatment KPS was 70 (IQR, 60–70), and median tumor volume was 0.133 cm^3 (IQR, 0.047–0.624). In the grouped holdout test set, the model achieved an AUC of 0.807 (patient-cluster bootstrap 95% CI, 0.714–0.892 and an MAE of 1.47 months (95% CI, 1.18–1.75). Probabilistic performance was favorable, with a Brier score of 0.077, calibration intercept of 0.129, and calibration slope of 0.833, indicating good discrimination, low overall prediction error, and acceptable calibration.
Conclusions
THINKERS-Melanoma provides an internally validated framework for patient-specific SRS dose comparison in melanoma brain metastases and supports the feasibility of AI-guided radiosurgical evaluation.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Neurological Surgery, Center for Image-Guided Neurosurgery, University of Pittsburgh Medical Center, Pittsburgh, USA. reyesbarretojs@upmc.edu.
- Department of Neurological Surgery, Computational Neurosurgery Research Group (CIGNS-CRG), Center for Image-Guided Neurosurgery, University of Pittsburgh Medical Center, Pittsburgh, USA. reyesbarretojs@upmc.edu.
- Department of Neurological Surgery, Center for Image-Guided Neurosurgery, University of Pittsburgh Medical Center, Pittsburgh, USA.
- Department of Neurological Surgery, Computational Neurosurgery Research Group (CIGNS-CRG), Center for Image-Guided Neurosurgery, University of Pittsburgh Medical Center, Pittsburgh, USA.
- Department of Neurological Surgery, Center for Image-Guided Neurosurgery, University of Pittsburgh Medical Center, Pittsburgh, USA. hadjipanayiscg2@upmc.edu.
- Department of Neurological Surgery, Computational Neurosurgery Research Group (CIGNS-CRG), Center for Image-Guided Neurosurgery, University of Pittsburgh Medical Center, Pittsburgh, USA. hadjipanayiscg2@upmc.edu.
Zitieren
APA
Reyes, J. S., Bouras, A., Lunsford, L. D., Niranjan, A., & Hadjipanayis, C. G. (2026). Therapeutic Hybrid Intelligence with Neural and Knowledge-based Expert Reasoning for SRS (THINKERS): an AI model for melanoma brain metastases. Neurosurgical Review, 49(1). https://doi.org/10.1007/s10143-026-04557-7
Vancouver
Reyes JS, Bouras A, Lunsford LD, Niranjan A, Hadjipanayis CG. Therapeutic Hybrid Intelligence with Neural and Knowledge-based Expert Reasoning for SRS (THINKERS): an AI model for melanoma brain metastases. Neurosurgical Review. 2026;49(1). doi:10.1007/s10143-026-04557-7
BibTeX
@article{Reyes2026,
author = {Reyes, Jheremy S and Bouras, Alexandros and Lunsford, L Dade and Niranjan, Ajay and Hadjipanayis, Constantinos G},
title = {Therapeutic Hybrid Intelligence with Neural and Knowledge-based Expert Reasoning for SRS (THINKERS): an AI model for melanoma brain metastases},
journal = {Neurosurgical Review},
year = {2026},
volume = {49},
number = {1},
doi = {10.1007/s10143-026-04557-7}
}
EUVIMED