Related Experiment Video
Updated: May 14, 2026
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
Dual-conditioned diffusion model with anatomical guidance for geometric distortion correction in prostate MRI
Inye Na1, Qi Miao2, Jonghun Kim1
1Department of Electrical and Computer Engineering, Sungkyunkwan University, Suwon, Republic of Korea.
European Radiology Experimental
|May 13, 2026
Summary
Geometric distortion in prostate diffusion-weighted imaging (DWI) complicates MRI interpretation. DeDistortNet corrects these distortions using T2-weighted images, improving anatomical accuracy without extra scans.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Prostate Cancer Diagnosis
Background:
- Susceptibility artifacts in diffusion-weighted imaging (DWI) cause geometric distortion in prostate MRI.
- This distortion degrades anatomical fidelity, hindering accurate clinical interpretation.
Purpose of the Study:
- To develop and evaluate DeDistortNet, a generative dual-conditioned diffusion model.
- To correct geometric distortions in prostate DWI without needing paired distorted-undistorted data.
Main Methods:
- Utilized the PROSTATEx dataset with simulated distortions for training.
- Employed a dual-conditioned diffusion model (DeDistortNet) integrating distorted DWI and T2-weighted images.
- Evaluated performance via quantitative analysis on simulated data and indirect validation on clinical data.
Main Results:
- DeDistortNet significantly improved image quality metrics (SNR, SSIM) in simulated data.
- Achieved substantial improvements in anatomical concordance (Dice similarity) with T2 references in clinically distorted data.
- Radiologist assessments confirmed enhanced geometric fidelity and prostate boundary delineation.
Conclusions:
- DeDistortNet effectively corrects geometric distortions in prostate DWI, restoring anatomical fidelity.
- The model performs particularly well in the peripheral zone, enhancing diagnostic reliability.
- This method eliminates the need for additional acquisitions or specialized protocols.
