Related Experiment Video
Updated: Jul 24, 2025
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
287
Improving accelerated 3D imaging in MRI-guided radiotherapy for prostate cancer using a deep learning method.
Ji Zhu1, Xinyuan Chen1, Yuxiang Liu1,2
1National Cancer Center, National Clinical Research Center for Cancer, Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, 100021, China.
Radiation Oncology (London, England)
|July 1, 2023
Summary
Deep learning enhances prostate cancer radiotherapy by improving high-speed MRI quality. This method significantly reduces scan time while maintaining accurate image registration for adaptive treatment.
Area of Science:
- Medical Imaging
- Radiotherapy
- Artificial Intelligence
Background:
- Online adaptive radiotherapy requires high-quality MR images for accurate treatment planning.
- High-speed MRI acquisition often compromises image quality, posing challenges for real-time adjustments.
Purpose of the Study:
- To improve image quality for high-speed MRI in prostate cancer radiotherapy using a deep learning method.
- To evaluate the impact of enhanced image quality on image registration accuracy.
Main Methods:
- A CycleGAN deep learning model was developed to generate synthetic high-quality MR images (synLSHQ) from high-speed, low-quality (HSLQ) images.
- The model was trained and validated using 1.5 T MR images from an MR-linac, comparing synLSHQ with low-speed, high-quality (LSHQ) images.
- Image quality was assessed using nMAE, PSNR, SSIM, and EKI; deformable image registration was evaluated using JDV, DSC, and MDA.
Main Results:
- The synLSHQ images achieved comparable quality to LSHQ images, reducing imaging time by approximately 66%.
- Compared to HSLQ images, synLSHQ showed significant improvements in image quality metrics (nMAE, SSIM, PSNR, EKI).
- Registration accuracy was enhanced with synLSHQ, demonstrating superior mean JDV and preferable DSC and MDA values over HSLQ.
Conclusions:
- The proposed deep learning method effectively generates high-quality MR images from high-speed sequences.
- This approach has the potential to shorten MRI scan times in radiotherapy without compromising treatment accuracy.
