Prostate MRI interpretation remains challenging and highly dependent on radiologist experience. This study investigated whether QP-Prostate®, a commercially available AI decision-support system, can improve the detection of clinically significant prostate cancer (csPCa) when used alongside radiologists with different levels of experience.

In this retrospective, prostatectomy-validated multi-reader study, four radiologists with 1, 3, 16 and 25 years of prostate MRI experience interpreted biparametric MRI both independently and with AI assistance. Performance was evaluated at the lesion level against whole-mount prostatectomy histopathology, providing spatial correspondence between imaging findings and confirmed cancer.

AI assistance significantly improved lesion-level discrimination for all four radiologists, with AUC increases ranging from +0.013 to +0.037. Sensitivity also increased significantly for three of four readers, while no statistically significant change in specificity was observed. The largest gains were seen among less-experienced radiologists. In the index-lesion analysis, sensitivity for the least-experienced reader increased from 73.1% to 86.5% with AI assistance.

Importantly, the findings support AI as an assistive rather than autonomous tool: standalone AI showed high specificity but relatively low lesion-level sensitivity, while combining AI outputs with radiologist interpretation improved performance. The results suggest that AI-based decision support may help radiologists identify subtle lesions that might otherwise be overlooked, with particularly relevant potential for reducing experience-related differences in prostate MRI interpretation.