Related Experiment Video
Updated: Oct 17, 2025
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
Synthesizing High-b-Value Diffusion-weighted Imaging of the Prostate Using Generative Adversarial Networks
Lei Hu1, Da-Wei Zhou1, Yun-Fei Zha1
1Department of Diagnostic and Interventional Radiology, Shanghai Jiao Tong University Affiliated Sixth People's Hospital, 600 Yi Shan Road, Shanghai 200233, China (L.H., W.H.X., J.G.Z.); State Key Laboratory of Integrated Services Networks, School of Telecommunications Engineering, Xidian University, Xi'an, China (D.W.Z.); Department of Radiology, Renmin Hospital, Wuhan University, Wuhan, China (Y.F.Z., L.L., H. He, L.Q., Y.K.Z.); MR Application Development, Siemens Shenzhen MR, Shenzhen, China (C.X.F.); and Department of Radiology, The Affiliated Renmin Hospital of Jiangsu University, Zhenjiang, China (H. Hu).
A deep learning framework using generative adversarial networks (GANs) can create synthetic high-b-value diffusion-weighted imaging (DWI) from standard-b-value DWI. This method shows promise for improving prostate cancer detection with good image quality and accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Radiology
- Oncology Imaging
Background:
- Diffusion-weighted imaging (DWI) is crucial for prostate cancer detection.
- Acquiring high-b-value DWI (b=1500 s/mm²) can improve diagnostic accuracy but may prolong scan times or be unavailable.
- Generating synthetic high-b-value DWI from lower b-values could enhance image quality and diagnostic utility.
Purpose of the Study:
- To develop and evaluate a deep learning framework utilizing generative adversarial networks (GANs).
- To synthesize high-b-value (b=1500 s/mm²) DWI (SYNb1500) from standard-b-value DWI (ACQb800 and ACQb1000).
- To assess the image quality and diagnostic performance of the synthesized DWI.
Main Methods:
- A retrospective multicenter study included 395 patients undergoing prostate multiparametric MRI.
- A GAN-based deep learning model (M0) was trained on an internal dataset and compared with a cycle GAN model (Mcyc).
- The model was optimized (Opt-M0) using denoising and edge-enhancement; synthetic datasets (Opt-SYNb1500) were generated and compared with acquired (ACQb1500) and calculated (CALb1500) datasets.
Main Results:
- The GAN model (M0) significantly outperformed the cycle GAN model (Mcyc) in image quality metrics.
- Optimized synthetic high-b-value DWI (Opt-SYNb1500) demonstrated significantly superior image quality compared to ACQb1500 and CALb1500.
- Opt-SYNb1500 also showed a higher area under the curve, indicating improved diagnostic utility.
Conclusions:
- A GAN-based deep learning framework is a promising tool for synthesizing realistic high-b-value DWI.
- The synthesized DWI sets exhibit good image quality and accuracy for prostate cancer detection.
- This approach may offer a valuable alternative for enhancing DWI acquisition protocols.
