Enhancing Access to Cerebrovascular Imaging Through Routine MRI for Preoperative Evaluation of Intracranial Tumor Patients: Development and Multi-Center Validation of an AI Model

By
Chaoyue Chen
Zhouyang Huang
Yanjie Zhao
Haoze Jiang
Yuen Teng
Xiaoping Ran
Yang Zhang
Shuangyi Zhang
Junkai Zheng
Clare Liu
Yu Hua
Fumin Zhao
Yi Zhang
Lei Zhang
Jianguo Xu
February 20, 2026
0 min

Bmc Medicine

Overview

This study presents the development and validation of a deep learning model for automatic vessel segmentation on contrast-enhanced T1-weighted MRI images in patients with intracranial tumors. The model aims to improve access to cerebrovascular imaging, particularly in low- and middle-income countries, where traditional angiographic methods may be limited.

Background

Intracranial tumor resection is a complex surgical procedure with significant risks, including vascular injury that can lead to severe complications. Comprehensive imaging, particularly MRI, is essential for preoperative assessment, yet access to advanced angiographic techniques is often inequitable. This study addresses the need for an alternative imaging method that can provide similar vascular information using routinely acquired MRI scans.

Data Highlights

The study involved a multicenter dataset of 1093 patients diagnosed with intracranial tumors. The model was rigorously evaluated for its performance in vessel segmentation on T1C images compared to traditional methods.

Key Findings

The deep learning model demonstrated improved vessel segmentation accuracy on T1C MRI images compared to traditional methods.
Access to advanced imaging techniques like TOF-MRA is limited in many regions, making this model a valuable alternative.
Integration of AI in imaging can enhance preoperative assessments and potentially reduce surgical risks.
The study highlights the importance of validating AI models across diverse clinical settings to ensure generalizability.
Routine MRI can provide sufficient vascular information, potentially reducing the need for more invasive angiographic procedures.

Clinical Implications

The development of this AI model for vessel segmentation on T1C MRI could significantly enhance preoperative planning for neurosurgeons, particularly in resource-limited settings. By improving access to critical imaging data, it may lead to better surgical outcomes and reduced complications associated with intracranial tumor resections.

Conclusion

This study underscores the potential of AI-driven imaging solutions to bridge gaps in cerebrovascular assessment, enhancing surgical safety and efficacy in intracranial tumor management.