Artificial intelligence for prostate MRI: open datasets, available applications, and grand challenges.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Artificial intelligence (AI) is increasingly utilized in prostate cancer (PCa) detection and characterization via magnetic resonance imaging (MRI).
• AI-assisted reading of prostate MRI shows potential for reducing radiologist workflow times.
• Existing open datasets comprise over 3,300 multi-vendor prostate MRI cases, varying in acquisition parameters and annotation quality.

Purpose of the Study:

• To review the current landscape of AI applications in prostate MRI.
• To assess the available open datasets for AI development and validation.
• To identify future research directions for AI in prostate cancer management.

Main Methods:

• Analysis of publicly available prostate MRI datasets (approx. 253 GB, 3,369 cases).
• Consideration of seven grand challenges and commercial AI applications from eleven vendors.
• Review of studies focusing on AI for prostate cancer detection, classification, and segmentation.

Main Results:

• Prostate MRI datasets vary significantly in quality and annotation, requiring careful usage.
• AI shows feasibility for workflow reduction in prostate cancer detection and classification.
• Limited prospective validation exists, highlighting the need for large-scale, multi-institutional studies.

Conclusions:

• AI holds promise for clinical integration in prostate cancer MRI, particularly for detection and characterization.
• The heterogeneity of current open datasets necessitates rigorous validation strategies.
• Future AI research should expand beyond detection to encompass prognosis, follow-up, and treatment planning in prostate cancer.