Related Experiment Video
Updated: Mar 23, 2026
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
1.9K
In vivo MRI based prostate cancer localization with random forests and auto-context model
Chunjun Qian1, Li Wang2, Yaozong Gao2
1School of Science, Nanjing University of Science and Technology, Jiangsu, China; Department of Radiology and BRIC, University of North Carolina at Chapel Hill, NC, United States.
Summary
This study introduces a new AI framework for precise prostate cancer localization in MRI scans. The method accurately identifies cancerous regions across the entire prostate, aiding in diagnosis and treatment planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Prostate cancer is a leading cause of male cancer deaths.
- Magnetic resonance (MR) imaging is crucial for localizing prostate cancer.
- Existing automated methods often focus only on specific prostate zones or regions of interest.
Purpose of the Study:
- To develop a fully automatic method for prostate cancer localization in the entire prostate region using in vivo MRI.
- To integrate multi-source imaging data for improved detection accuracy.
- To address limitations of previous methods that focused on limited areas or regions.
Main Methods:
- A novel learning-based multi-source integration framework was proposed.
- Random forests were used to integrate features from multi-parametric MRIs (T2, DWI, dADC).
- Iteratively-estimated and refined tissue probability maps were incorporated.
Main Results:
- The method accurately localized cancerous sections in 26 patient datasets.
- Section-based evaluation (SBE) and ROC analysis demonstrated high performance.
- Area Under the Curve (AUC) values of 0.832 (common ROC) and 0.883 (ROI-based ROC) were achieved.
Conclusions:
- The proposed framework shows promise for in vivo MRI-based prostate cancer localization.
- Accurate localization can guide prostate biopsy and focal therapy planning.
- The method supports active surveillance and treatment decision-making.
