Related Experiment Video
Updated: Oct 10, 2025
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
451
Deep learning model for automatic prostate segmentation on bicentric T2w images with and without endorectal coil
Summary
This study presents an automatic prostate segmentation algorithm using U-Net transfer learning. The method achieves over 85% accuracy on internal and external datasets, aiding computer-aided diagnosis for prostate cancer detection.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Prostate MRI segmentation is crucial for computer-aided diagnosis (CAD) of cancer.
- Variations in MRI data across different centers and scanners pose challenges for automated segmentation.
Purpose of the Study:
- To develop an automatic prostate segmentation algorithm using transfer learning with a U-Net architecture.
- To validate the algorithm's performance in a bi-center setting, addressing data variability.
Main Methods:
- A U-Net model with transfer learning was employed for prostate segmentation.
- Training and internal validation used T2w MRI from 80 patients at Center A (with/without endorectal coil).
- External validation utilized T2w MRI from 20 patients at Center B (without endorectal coil).
Main Results:
- The algorithm achieved a Dice similarity coefficient exceeding 85% on both internal and external validation datasets.
- This demonstrates robust performance despite variations in MRI acquisition parameters.
Conclusions:
- The developed automatic prostate segmentation algorithm shows high accuracy and generalizability.
- Integration into CAD systems can optimize computational resources for prostate cancer detection.
