Assessing the Performance of Artificial Intelligence Assistance for Prostate MRI: A Two-Center Study Involving...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) holds promise for improving radiologist performance in detecting clinically significant prostate cancer (csPCa) on MRI.
• Further research is needed to validate AI's impact across varying levels of radiologist experience.

Purpose of the Study:

• To evaluate the effectiveness of AI assistance in csPCa detection among radiologists with different experience levels.
• To compare radiologist performance with and without AI support.

Main Methods:

• Retrospective analysis of 900 prostate MRI scans from patients who underwent biopsy.
• Ten less-experienced and six experienced radiologists reviewed cases twice, with and without AI assistance, using PI-RADS v2.1.
• Performance metrics including sensitivity, specificity, AUC, reading time, and diagnostic confidence were assessed.

Main Results:

• AI significantly improved lesion-level sensitivity and patient-level AUC for less-experienced radiologists.
• Experienced radiologists showed improved sextant-level AUC with AI assistance.
• AI reduced reading time and increased diagnostic confidence for all radiologists, enhancing consistency among experienced readers.

Conclusions:

• AI-assisted MRI reading enhances the detection of csPCa, especially for radiologists with less experience.
• AI tools can augment radiologist capabilities, leading to more efficient and confident diagnoses.