Related Experiment Video
Updated: Aug 7, 2025
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
Mutually communicated model based on multi-parametric MRI for automated segmentation and classification of prostate
Kewen Liu1,2, Piqiang Li1,2, Martins Otikovs3
1State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan, P.R. China.
A novel deep learning network improves prostate cancer diagnosis by integrating segmentation and classification. This mutually communicated deep learning segmentation and classification network (MC-DSCN) enhances accuracy in detecting prostate cancer using MRI scans.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Multiparametric magnetic resonance imaging (mp-MRI) is a key noninvasive tool for prostate cancer (PCa) detection and characterization.
- Accurate segmentation and classification of PCa from mp-MRI are crucial for effective diagnosis and treatment planning.
Purpose of the Study:
- To develop and evaluate a mutually communicated deep learning segmentation and classification network (MC-DSCN).
- To leverage mp-MRI for enhanced prostate segmentation and PCa diagnosis using the MC-DSCN.
Main Methods:
- The MC-DSCN facilitates mutual information transfer between segmentation and classification components for a bootstrapping effect.
- It utilizes segmentation masks to refine classification by excluding irrelevant regions and employs classification localization for improved segmentation.
- The network was trained and validated on mp-MRI data (T2-weighted, ADC) from two medical centers, with statistical analysis including DeLong and paired t-tests.
Main Results:
- The MC-DSCN demonstrated superior performance compared to single-task networks in both segmentation and classification.
- Segmentation accuracy (IOU) improved from 84.5% to 87.8% (Center A) and 83.8% to 87.1% (Center B).
- Prostate cancer classification AUC improved from 0.946 to 0.991 (Center A) and 0.926 to 0.955 (Center B).
Conclusions:
- The proposed MC-DSCN architecture effectively transfers mutual information between segmentation and classification tasks.
- This synergistic approach significantly enhances diagnostic performance for prostate cancer detection and segmentation from mp-MRI.
- The MC-DSCN outperforms networks designed for single tasks, highlighting the benefit of integrated segmentation and classification.
