Related Experiment Video
Updated: Apr 29, 2026
Hybrid PET/MRI Imaging of Alzheimer's Disease Based on 18F-AV-1451
Published on: April 18, 2025
Prediction of FLAIR MRI from 18F-FDG PET/CT for the Evaluation of White Matter Hyperintensity Using Generative
Kyeong Taek Oh1, Sangwon Lee2, Dongwoo Kim2,3
1Department of Biomedical Engineering, Yonsei University College of Medicine, Seoul, Republic of Korea.
Abstract:
White matter hyperintensities (WMH) may decrease cortical glucose metabolism and appear hypodense on 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT), respectively. Currently, T2-weighted fluid-attenuated inversion recovery (FLAIR) images on magnetic resonance imaging (MRI) are considered as a sequence of choice to evaluate WMH. This study aimed to use a generative adversarial network (GAN) to predict FLAIR MRI images from 18F-FDG PET/CT. From 2017 to 2019, we selected 167 patients who had MRI and FDG PET/CT scans within 6 months. We categorized WMH into three groups using the Fazekas scale and trained a GAN model to predict MR FLAIR images from PET and CT data (pix2pix-PT), or only CT data (pix2pix-CT). We compared these predicted images with actual MR FLAIR images, then performed WMH segmentation and volume estimation, assessing their agreement. To predict ground-truth FLAIR images, the pix2pix-PT method demonstrated superior performance compared with pix2pix-CT, as evidenced by the lower NMAE and higher PSNR in all groups. Integrating these findings with the segmentation results, the performance of the pix2pix-PT model in WMH segmentation was differential across groups. Notably, the pix2pix-PT model accurately segmented WMH lesions over 0.3 cm2 without false positives or negatives and maintained a DSC above 0.7 for lesions over 2 cm2. For WMH volume estimation, the pix2pix-PT method showed excellent correlations in Group 2 (r = 0.903) and Group 3 (r = 0.984), and moderate in Group 1 (r = 0.780). In this study, the prediction of T2-weighted FLAIR MR images using the GAN model was better achieved when both FDG PET and CT data were provided to the model, compared to CT data alone. Predicted T2-FLAIR images derived from our model could aid in selecting patients who need MRI to assess WMH and related vascular pathology.
Insights
This study shows that generative adversarial networks (GANs) can predict T2-weighted FLAIR MRI images from 18F-FDG PET/CT scans. Combining PET and CT data improves white matter hyperintensities (WMH) assessment, aiding patient selection for MRI.
Area of Science:
- Neuroimaging
- Medical Physics
- Artificial Intelligence in Medicine
Background:
- White matter hyperintensities (WMH) can impact cortical glucose metabolism and are visualized on 18F-FDG PET/CT and MRI.
- T2-weighted FLAIR MRI is the standard for WMH evaluation.
- There is a need for non-MRI methods to assess WMH, especially in patients where MRI is contraindicated or unavailable.
Purpose of the Study:
- To develop and evaluate a generative adversarial network (GAN) model for predicting T2-weighted FLAIR MRI images from 18F-FDG PET/CT data.
- To compare the performance of a GAN model using both PET and CT data versus CT data alone.
- To assess the utility of predicted FLAIR images for WMH segmentation and volume estimation.
Main Methods:
- A GAN model (pix2pix-PT) was trained to predict FLAIR MRI from combined 18F-FDG PET/CT data, and another model (pix2pix-CT) used only CT data.
- 167 patients with both MRI and FDG PET/CT scans were included.
- Predicted FLAIR images were compared to actual FLAIR images, followed by WMH segmentation and volume analysis using the Fazekas scale for categorization.
Main Results:
- The pix2pix-PT model (using PET and CT) outperformed pix2pix-CT (using CT alone) in predicting FLAIR images, showing lower NMAE and higher PSNR.
- WMH segmentation performance was differential across groups, with accurate segmentation of lesions >0.3 cm² and DSC >0.7 for lesions >2 cm².
- Excellent correlations were found for WMH volume estimation in Fazekas Groups 2 and 3 (r=0.903, r=0.984) and moderate in Group 1 (r=0.780).
Conclusions:
- Generative adversarial networks can effectively predict T2-FLAIR MRI images from 18F-FDG PET/CT data.
- Integrating both PET and CT data significantly improves the accuracy of predicted FLAIR images and subsequent WMH analysis compared to CT alone.
- This GAN-based approach shows potential for aiding in patient selection for MRI-based WMH assessment and evaluating related vascular pathology.
