Artificial Intelligence in Prostate MRI: Comparison of an AI-Based Software and an Experienced Radiologist for...

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Multiparametric MRI is crucial for detecting clinically significant prostate cancer (csPCa).
• Diagnostic accuracy in prostate MRI interpretation is influenced by reader experience.
• Artificial intelligence (AI) tools can potentially enhance prostate MRI interpretation and minimize inter-reader variability.

Purpose of the Study:

• To compare the detection rate of an AI-based software (PAROS) against an experienced radiologist in identifying csPCa.
• To evaluate the diagnostic performance of AI in prostate MRI interpretation.

Main Methods:

• Retrospective single-center study of 150 patients undergoing prostate MRI and biopsy.
• MRI interpretation by an experienced radiologist (PI-RADS v2.1) and independent analysis by PAROS AI on biparametric MRI.
• Histopathology as the reference standard; evaluation using sensitivity, specificity, PLR, and NLR at PI-RADS thresholds ≥3 and ≥4.

Main Results:

• 63.3% of patients had csPCa.
• PAROS and the radiologist demonstrated comparable sensitivity and specificity at both PI-RADS thresholds.
• Extremely low negative likelihood ratios (NLRs) indicated excellent rule-out capability for both.
• PAROS identified more lesions, particularly in the transition zone.

Conclusions:

• The AI software (PAROS) achieved csPCa detection rates comparable to an experienced radiologist.
• PAROS shows promise as a decision-support tool for prostate MRI interpretation.
• AI can aid in reducing inter-reader variability in prostate cancer diagnosis.