Related Experiment Video
Updated: Oct 3, 2025
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
419
Low-Dose 68 Ga-PSMA Prostate PET/MRI Imaging Using Deep Learning Based on MRI Priors
Fuquan Deng1,2,3, Xiaoyuan Li4, Fengjiao Yang4
1Lauterbur Research Center for Biomedical Imaging, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China.
Frontiers in Oncology
|February 14, 2022
Summary
Deep learning restores low-dose 68 Ga-PSMA PET/MRI scans to full-dose quality, enabling up to 50% radiation dose reduction for prostate cancer patients without impacting diagnosis.
Area of Science:
- Medical Imaging
- Nuclear Medicine
- Artificial Intelligence
Background:
- 68 Ga-prostate-specific membrane antigen (PSMA) PET/MRI is crucial for prostate cancer diagnosis.
- Reducing radiation exposure in medical imaging is a significant clinical goal.
Purpose of the Study:
- To apply deep learning for low-dose 68 Ga-PSMA PET/MRI image restoration.
- To evaluate the impact of synthesized full-dose PET (FDPET) images on diagnostic accuracy.
Main Methods:
- Utilized deep learning with MRI priors to synthesize FDPET images from low-dose PET (LDPET) data of 41 patients.
- Assessed image quality using quantitative scores from nuclear medicine doctors and imaging metrics (PSNR, SSIM, NMSE, RCNR).
Main Results:
- Synthesized FDPET images from 50% dose PET data achieved high quantitative scores (4.03±0.17) and excellent image quality metrics (PSNR: 39.88±3.83, SSIM: 0.896±0.092).
- Image quality indicators demonstrated the effectiveness of the deep learning restoration method.
Conclusions:
- Deep learning-based synthesis of FDPET images from 50% dose PET data does not compromise prostate cancer diagnosis.
- This method allows for significant radiation dose reduction (up to 50%) in 68 Ga-PSMA PET/MRI scans for prostate cancer patients.
