Related Experiment Video
Updated: Oct 5, 2025
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
ProstAttention-Net: A deep attention model for prostate cancer segmentation by aggressiveness in MRI scans
Audrey Duran1, Gaspard Dussert1, Olivier Rouvière2
1Univ Lyon, CNRS, Inserm, INSA Lyon, UCBL, CREATIS, UMR5220, U1206, F-69621, Villeurbanne, France.
Abstract:
Multiparametric magnetic resonance imaging (mp-MRI) has shown excellent results in the detection of prostate cancer (PCa). However, characterizing prostate lesions aggressiveness in mp-MRI sequences is impossible in clinical practice, and biopsy remains the reference to determine the Gleason score (GS). In this work, we propose a novel end-to-end multi-class network that jointly segments the prostate gland and cancer lesions with GS group grading. After encoding the information on a latent space, the network is separated in two branches: 1) the first branch performs prostate segmentation 2) the second branch uses this zonal prior as an attention gate for the detection and grading of prostate lesions. The model was trained and validated with a 5-fold cross-validation on a heterogeneous series of 219 MRI exams acquired on three different scanners prior prostatectomy. In the free-response receiver operating characteristics (FROC) analysis for clinically significant lesions (defined as GS >6) detection, our model achieves 69.0%±14.5% sensitivity at 2.9 false positive per patient on the whole prostate and 70.8%±14.4% sensitivity at 1.5 false positive when considering the peripheral zone (PZ) only. Regarding the automatic GS group grading, Cohen's quadratic weighted kappa coefficient (κ) is 0.418±0.138, which is the best reported lesion-wise kappa for GS segmentation to our knowledge. The model has encouraging generalization capacities with κ=0.120±0.092 on the PROSTATEx-2 public dataset and achieves state-of-the-art performance for the segmentation of the whole prostate gland with a Dice of 0.875±0.013. Finally, we show that ProstAttention-Net improves performance in comparison to reference segmentation models, including U-Net, DeepLabv3+ and E-Net. The proposed attention mechanism is also shown to outperform Attention U-Net.
Insights
This study introduces ProstAttention-Net, an AI model for prostate cancer detection and grading using MRI. It accurately segments prostate glands and cancer lesions, improving diagnostic capabilities beyond current clinical practice.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Multiparametric MRI (mp-MRI) shows promise for prostate cancer (PCa) detection.
- Characterizing PCa aggressiveness and Gleason score (GS) from mp-MRI is challenging in clinical practice.
- Biopsy remains the gold standard for determining PCa aggressiveness.
Purpose of the Study:
- To develop a novel end-to-end multi-class network for joint prostate gland segmentation and PCa lesion detection with Gleason score group grading.
- To improve the characterization of prostate cancer aggressiveness using AI-driven mp-MRI analysis.
Main Methods:
- An end-to-end multi-class network (ProstAttention-Net) was designed with two branches: one for prostate segmentation and another for lesion detection/grading using zonal priors as an attention gate.
- The model was trained and validated using 5-fold cross-validation on 219 MRI exams from three scanners.
- Performance was evaluated using Free-Response Receiver Operating Characteristic (FROC) analysis and Cohen's quadratic weighted kappa (κ) for lesion detection and GS grading.
Main Results:
- ProstAttention-Net achieved 69.0%±14.5% sensitivity for clinically significant lesion detection (GS > 6) across the whole prostate and 70.8%±14.4% in the peripheral zone.
- The model obtained a Cohen's kappa of 0.418±0.138 for automatic GS group grading, outperforming previous lesion-wise GS segmentation.
- State-of-the-art prostate segmentation (Dice of 0.875±0.013) and improved performance over U-Net, DeepLabv3+, and E-Net were demonstrated.
Conclusions:
- ProstAttention-Net shows strong potential for improving prostate cancer detection, segmentation, and Gleason score grading from mp-MRI.
- The proposed attention mechanism enhances diagnostic accuracy and generalization capabilities.
- This AI approach offers a promising non-invasive method to characterize prostate cancer aggressiveness, potentially reducing the need for extensive biopsies.
