Related Experiment Video
Updated: Oct 16, 2025
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
461
A Fully Automatic Artificial Intelligence System Able to Detect and Characterize Prostate Cancer Using
Valentina Giannini1,2, Simone Mazzetti1,2, Arianna Defeudis1,2
1Department of Radiology, Candiolo Cancer Institute, FPO-IRCCS, Candiolo, Italy.
Frontiers in Oncology
|October 18, 2021
Summary
A new artificial intelligence system accurately detects and grades prostate cancer (PCa) aggressiveness from MRI scans. This computer-aided diagnosis (CAD) tool aids in personalized treatment decisions, improving patient outcomes and reducing overtreatment of insignificant tumors.
Area of Science:
- Medical Imaging and Artificial Intelligence
- Oncology and Urology
- Radiomics and Machine Learning
Background:
- Prostate-specific antigen (PSA) testing has reduced prostate cancer (PCa) mortality but leads to overtreatment of indolent cancers.
- Accurate risk stratification is crucial for personalized treatment selection in PCa patients.
- Existing diagnostic tools require manual input and are often limited to single-center data.
Purpose of the Study:
- To develop and validate a fully automated computer-aided diagnosis (CAD) system for PCa detection and aggressiveness characterization.
- To create a tool that assists physicians in selecting appropriate treatment options based on individual patient risk.
- To overcome limitations of previous studies, including manual segmentation and single-center data dependency.
Main Methods:
- Development of an AI-based CAD system integrating multiple MRI sequences.
- Automated tumor candidate identification and aggressiveness scoring using a support vector machine classifier fed with radiomics features.
- Validation on a multi-institutional dataset comprising 131 patients (149 tumors) with training, narrow validation, and external validation sets.
Main Results:
- The CAD system achieved an area under the ROC curve of 0.96 (training) and 0.81 (validation) for distinguishing low vs. high aggressiveness.
- When risk was stratified into three classes (indolent, indeterminate, aggressive), no aggressive tumors were misclassified as indolent.
- The system demonstrated superior performance compared to previous studies, offering automated analysis on multivendor data.
Conclusions:
- The developed AI-powered CAD system shows significant promise for accurate PCa aggressiveness assessment.
- This automated tool can support personalized decision-making, potentially reducing overtreatment and improving patient management.
- The system's multi-institutional validation suggests its potential for widespread clinical adoption in PCa diagnosis.
