Related Experiment Video
Updated: Jun 13, 2026
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
Multicenter Study Suggests Unsupervised Learning Derived From MRI Identifies Prognostic Subgroups in Prostate Cancer
Guoqing Hu1,2, Xiaofeng Liu3, Zhangzhe Chen1,2
1Department of Radiology, Fudan University Shanghai Cancer Center, Shanghai, China, shca.org.cn.
Radiology Research and Practice
|June 12, 2026
Summary
This study identified distinct patient subgroups for prostate cancer (PCa) after radical prostatectomy (RP) using MRI radiomics and clinical data. The new model significantly improves prediction of 5-year progression-free survival (PFS).
Area of Science:
- Oncology
- Radiology
- Data Science
Background:
- Prostate cancer (PCa) management after radical prostatectomy (RP) requires accurate prognostication.
- Existing risk assessment models have limitations in predicting long-term outcomes.
- Radiomics analysis of magnetic resonance imaging (MRI) shows promise in uncovering hidden prognostic information.
Purpose of the Study:
- To identify patient subgroups with PCa post-RP using clinical and MRI radiomics features.
- To evaluate the prognostic value of these subgroups in predicting 5-year progression-free survival (PFS).
- To develop and validate a novel Radiomic-Clinical model for improved PCa prognostication.
Main Methods:
- Collected preoperative MRI and clinical data from 400 PCa patients across three centers.
- Extracted radiomics features from index lesions and performed feature selection using LASSO-Cox analysis.
- Utilized K-means clustering to define subgroups and developed a Radiomic-Clinical model, validated externally.
Main Results:
- Identified three distinct prognostic subgroups based on 5 clinical and 13 radiomics features.
- The Radiomic-Clinical model achieved superior predictive accuracy (C-indices: 0.82-0.78) compared to EAU, CAPRA, and PIPEN models.
- Demonstrated statistically significant improvements in predicting 5-year PFS across training and validation cohorts.
Conclusions:
- Unsupervised learning integrating radiomics and clinical data effectively stratifies PCa patients post-RP.
- The developed Radiomic-Clinical model offers enhanced prognostic performance for predicting 5-year PFS.
- This approach provides a valuable tool for personalized risk assessment and treatment planning in PCa.
