Related Experiment Video
Updated: Jul 30, 2025
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
303
Development and acceptability validation of a deep learning-based tool for whole-prostate segmentation on
Lili Xu1,2, Gumuyang Zhang1, Daming Zhang1
1Department of Radiology, State Key Laboratory of Complex Severe and Rare Disease, Peking Union Medical College Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China.
Quantitative Imaging in Medicine and Surgery
|May 14, 2023
Summary
This study developed a deep learning tool for prostate MRI segmentation. The tool shows accurate and robust performance, especially on T2-weighted imaging, aiding in prostate disease management.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Prostate Cancer Research
Background:
- Accurate whole prostate segmentation on MRI is crucial for managing prostate diseases.
- Deep learning tools offer potential for automated segmentation on T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI).
Purpose of the Study:
- To develop and evaluate a clinically applicable deep learning tool for automatic whole prostate segmentation.
- To assess the tool's performance on T2WI and DWI across multiple centers.
Main Methods:
- A 3D U-Net-based deep learning model was trained on 223 patients' prostate MRI data.
- Validation was performed on internal (n=95) and three external cohorts (n=141, n=30, n=29).
- Quantitative metrics (DSC, 95HD, ABD) and qualitative analysis were used for evaluation; DWI model was fine-tuned.
Main Results:
- The tool demonstrated strong performance on T2WI (DSC: 0.897-0.947) and DWI (DSC: 0.815 post-fine-tuning).
- Fine-tuning significantly improved DWI model performance (DSC: 0.275 vs. 0.815, P<0.01).
- High accuracy was observed in the midgland; 98.6% of T2WI and 72.3% of DWI segmentations were clinically acceptable.
Conclusions:
- The 3D U-Net tool provides accurate and robust automatic prostate segmentation on T2WI.
- Prostate segmentation on DWI is feasible, though fine-tuning may be necessary for diverse scanners.
