Artificial Intelligence-Assisted Biparametric MRI for Detecting Prostate Cancer-A Comparative Multireader Multicase...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Oncology Diagnostics

Background:

• Multiparametric MRI (mpMRI) is standard for prostate cancer (PCa) detection.
• Biparametric MRI (bpMRI) offers a streamlined approach but may have limitations.
• AI integration aims to enhance diagnostic performance and accessibility.

Purpose of the Study:

• To assess the diagnostic accuracy of AI-assisted bpMRI (AI-bpMRI) for PCa detection.
• To compare AI-bpMRI with traditional mpMRI and standard bpMRI.
• To evaluate the impact of reader experience on AI-bpMRI performance.

Main Methods:

• Retrospective analysis of 105 men with suspected PCa.
• Independent scoring of bpMRI, mpMRI, and AI-bpMRI by radiologists of varying experience.
• PI-RADS V2.1 classification and histopathological correlation.
• Area Under the Curve (AUC) analysis for diagnostic accuracy, with non-inferiority testing.

Main Results:

• AI-bpMRI demonstrated non-inferiority to mpMRI in detecting clinically significant PCa (Gleason score ≥ 3+4).
• AI-bpMRI was superior to bpMRI in detecting any PCa (Gleason score ≥ 3+3).
• The diagnostic benefits of AI-bpMRI were most pronounced in radiologists with lower to medium experience.

Conclusions:

• AI-bpMRI achieves diagnostic accuracy comparable to mpMRI for prostate cancer detection.
• AI-bpMRI represents a promising alternative to mpMRI, potentially improving consistency across different experience levels.
• Further validation may support AI-bpMRI as a cost-effective and efficient tool in PCa diagnostics.