Related Experiment Video
Updated: Jul 1, 2026
12:24
Computed Tomography-guided Time-domain Diffuse Fluorescence Tomography in Small Animals for Localization of Cancer Biomarkers
Published on: July 17, 2012
Gadolinium-Free MR Perfusion Imaging Based on Generative Adversarial Network for Primary Intracranial Tumor
Guoqi Lin1,2,3, Wangbin Ding3, Yihua Chen3
1Department of Radiology, the First Affiliated Hospital, Fujian Medical University, Fuzhou, 350005, China.
Neuro-Oncology
|June 26, 2026
Summary
This study introduces a novel gadolinium-free method using a tumor-aware generative adversarial network (TA-GAN) to synthesize cerebral blood volume (CBVsyn) maps from non-contrast MRI. The developed CBVsyn maps show promise in characterizing intracranial tumors, offering a safer alternative to contrast agents.
Area of Science:
- Neuro-oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate evaluation of tumor vascularity is crucial in neuro-oncology.
- Conventional cerebral blood volume (CBV) mapping relies on gadolinium-based contrast agents, posing potential risks.
- There is a need for gadolinium-free methods to assess tumor vascularity.
Purpose of the Study:
- To develop and validate a novel gadolinium-free model for synthesizing CBV maps (CBVsyn) from non-contrast MRI (NC-MRI).
- To assess the diagnostic performance and clinical utility of CBVsyn maps in characterizing primary intracranial tumors.
Main Methods:
- A multicenter retrospective study involving 1227 MRI examinations of patients with primary intracranial tumors.
- Implementation of a tumor-aware generative adversarial network (TA-GAN) to synthesize CBVsyn maps from various NC-MRI sequences.
- Evaluation of image quality using PSNR and SSIM, correlation with pathologic markers, and comparison of diagnostic performance against conventional DSC-PWI derived CBV (CBVreal) maps.
Main Results:
- The TA-GAN model utilizing T2-weighted and diffusion-weighted imaging achieved high performance in CBVsyn synthesis, with 95.34% of maps being diagnostically acceptable.
- CBVsyn parameters demonstrated a positive correlation with CD105 markers, indicating relevance to tumor pathology.
- CBVsyn maps showed comparable diagnostic performance to CBVreal maps for tumor grading and genotype classification, and improved diagnostic accuracy and consistency when assisting NC-MRI alone.
Conclusions:
- The developed TA-GAN model offers a promising gadolinium-free alternative for synthesizing CBV maps from NC-MRI.
- CBVsyn maps correlate with pathologic findings and show potential for characterizing primary intracranial tumors.
- This approach may enhance diagnostic capabilities in neuro-oncology while avoiding gadolinium-based contrast agents.
