Current Value of Biparametric Prostate MRI with Machine-Learning or Deep-Learning in the Detection, Grading, and...

Area of Science:

• Radiology and Oncology
• Artificial Intelligence in Medicine

Background:

• Prostate cancer detection relies on PI-RADS guidelines using MRI, including morphologic imaging, diffusion-weighted imaging, and perfusion.
• Biparametric MRI protocols omit contrast-enhanced perfusion for efficiency, impacting diagnostic capabilities.

Purpose of the Study:

• To review the utility of biparametric MRI combined with machine learning (ML) and deep learning (DL) for prostate cancer detection, grading, and characterization.
• To compare ML/DL performance against human radiologists where data is available.

Main Methods:

• Systematic PubMed search identifying 29 relevant studies.
• Analysis of studies evaluating biparametric MRI with ML/DL for prostate cancer assessment.

Main Results:

• ML and DL demonstrate feasibility in detecting clinically significant prostate cancer.
• These AI techniques show promise in differentiating cancerous from non-cancerous tissue.
• Some ML/DL approaches achieve performance comparable to radiologists in PI-RADS lesion classification.

Conclusions:

• Biparametric MRI integrated with ML/DL offers a viable approach for prostate cancer diagnosis.
• AI-powered analysis shows potential to match or exceed human radiologist performance in specific tasks.
• Further research is warranted to fully integrate these advanced techniques into clinical practice.