Research Results Show 20% Improved Prostate Cancer Detection on Biparametric MRI Using AI

New  research highlights the promise of artificial intelligence (AI) in improving prostate cancer detection, particularly in helping radiologists identify smaller and harder-to-spot lesions on MRI. According to the study, radiologists' sensitivity improved by nearly 20% when aided by AI detection tools.

The findings suggest AI could help reduce variability in prostate MRI interpretation, a challenge that persists despite the imaging modality's high diagnostic accuracy.

Jiule Ding, MD, with the department of radiology at Third Affiliated Hospital of Soochow University in China noted, “Prostate MRI has proven highly effective for PCa detection, leading to its adoption in international guidelines as a primary diagnostic tool for suspected cases. Despite its accuracy, prostate MRI interpretation varies significantly across medical institutions and depends heavily on reader expertise.” 

While many studies have shown that AI can improve prostate cancer detection on MRI, most have been limited in scope. This study aimed to fill that gap using a rigorous fully-crossed, multi-reader, multi-case (MRMC) trial design.

“The MRMC design reduces bias by ensuring every reader interprets every case, increasing the validity of our findings,” the authors explained. “To date, only three studies have used this design to evaluate AI’s role in prostate cancer detection on MRI.”

In this study, 10 non-expert radiologists from three institutions interpreted 407 biparametric prostate MRI cases (T2-weighted and diffusion-weighted imaging), both with and without AI support. Their performance was compared to assess the impact of AI and to benchmark against AI as a standalone reader.

With AI support, lesion-level sensitivity rose significantly—from 67.3% without AI to 85.5% with it. Case-level sensitivity also improved by about 4%, and AFROC-AUC scores increased from 76.1% to 86.9%. Although the AI system performed well independently, it did not outperform radiologists using AI assistance.

The benefit was especially notable for small lesions (≤1 cm), where radiologist sensitivity increased from 38.3% to 94.8% with AI. For larger lesions (>3 cm), the sensitivity increase was more modest, at just 7.2%.

“These results suggest that AI systems are particularly effective in improving detection and localization of smaller prostate cancer lesions,” the researchers noted.

While the study had limitations, the authors believe their trial design offers strong evidence supporting the integration of AI tools into clinical practice for prostate cancer imaging.