Evaluating an AI-driven Triaging Workflow for MRI-based Clinically Significant Prostate Cancer Diagnosis: A...

Area of Science:

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

Background:

• Prostate cancer diagnosis relies heavily on MRI interpretation.
• Current workflows face challenges in efficiency and accuracy.
• AI offers potential for optimizing diagnostic pathways.

Purpose of the Study:

• To simulate and evaluate an AI-driven triaging workflow for prostate MRI.
• To compare AI assessment against radiologists for clinically significant prostate cancer (csPCa).
• To estimate potential workload reduction through AI triaging.

Main Methods:

• Retrospective analysis of 500 prostate MRI examinations from four European centers.
• AI triaging thresholds calibrated on 100 cases, simulated on 400 cases with 62 radiologists.
• Comparison of AI-driven vs. conventional workflow using multireader, multicase analysis of variance.

Main Results:

• AI pathway maintained sensitivity (89.0%) comparable to radiologists (89.4%) but significantly improved specificity (69.2% vs. 57.7%).
• The AI system triaged and diagnosed 49% of examinations with high sensitivity (94.7%) and specificity (94.7%).
• Simulated workflow demonstrated improved efficiency without compromising diagnostic accuracy for csPCa.

Conclusions:

• AI-driven triaging enhances the efficiency of prostate MRI interpretation for csPCa.
• The AI system shows potential for reducing radiologist workload while maintaining diagnostic performance.
• This AI approach offers a promising tool for optimizing oncology diagnostic workflows.