Related Experiment Video
Updated: Aug 3, 2026
09:18
Image-Guided Resection of Glioblastoma and Intracranial Implantation of Therapeutic Stem Cell-seeded Scaffolds
Published on: July 16, 2018
8.7K
Enhancing decision-making in glioblastoma surgery through an explainable human-AI collaboration: an international
Julius M Kernbach1,2,3, Urte Schroeder4,5, Karlijn Hakvoort4,5
1Department of Neuroradiology, Heidelberg University, Heidelberg, Germany. julius.kernbach@med.uni-heidelberg.de.
NPJ Precision Oncology
|November 27, 2025
Summary
An explainable AI model predicts glioblastoma resection extent, improving surgical planning. Combining AI with human expertise enhances patient outcomes by outperforming traditional methods.
Area of Science:
- Neurosurgery
- Artificial Intelligence
- Oncology
Background:
- Surgical resection is crucial for glioblastoma survival but predicting extent of resection (EOR) is difficult.
- Accurate EOR prediction aids in optimizing surgical strategy and patient outcomes.
Purpose of the Study:
- To develop and validate an explainable AI model for personalized EOR estimation in glioblastoma patients.
- To assess the clinical impact of the AI model compared to human experts and a combined approach.
Main Methods:
- An explainable AI model was developed and externally validated on 811 glioblastoma patients.
- EOR was classified into gross-total (GTR), near-total (NTR), and subtotal resections (STR).
- Model interpretability and sensitivity analyses were performed; human-AI collaboration was evaluated.
Main Results:
- The AI model demonstrated generalizability with an AUC of 0.78 (CI 0.73-0.82).
- Class-specific AUCs were 0.75 for GTR, 0.59 for NTR, and 0.69 for STR.
- Human-AI collaboration significantly outperformed human experts in accuracy, F1 scores, and Cohen's κ.
Conclusions:
- Explainable AI can accurately predict EOR in glioblastoma, supporting preoperative planning.
- Integrating AI with surgical expertise enhances predictive performance and decision-making.
- This approach highlights the value of machine intelligence in neurosurgical oncology.
