"AI-Assisted vs Unassisted Identification of Prostate Cancer in MRI" —in JAMA Network Open

An international multi-reader, multi-case observer study investigating "AI-Assisted vs Unassisted Identification of Prostate Cancer in Magnetic Resonance Images" from the PI-CAI consortium has been published in JAMA Network Open.

👉 Read the full publication here (open-access): https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2835296

𝗦𝘁𝘂𝗱𝘆 𝗢𝘃𝗲𝗿𝘃𝗶𝗲𝘄 🧪 This large-scale international observer study involved 61 radiologists (34 experts and 27 non-experts) from 53 centers across 17 countries. A total of 360 prostate MRI examinations (three Dutch centers and one Norwegian center, disease prevalence of 34%) were reviewed both with and without AI assistance using the AI system developed within PI-CAI.

𝗢𝘂𝘁𝗰𝗼𝗺𝗲𝘀 📈 AI assistance led to statistically significant improvements in diagnostic performance, including AUROC (+3.3%, from 0.883 to 0.916), sensitivity (+2.5%, from 94.3% to 96.8%), and specificity (+3.4%, from 46.7% to 50.1%). Notably, non-expert readers experienced greater performance gains with AI assistance compared to expert readers.

𝗣𝗲𝗿𝘀𝗽𝗲𝗰𝘁𝗶𝘃𝗲 🩺 Findings provide evidence that AI assistance can improve the accuracy of clinically significant prostate cancer detection in MRI interpretation, especially for radiologists with less experience in prostate MRI reading. There remains a need for continued investigation, especially in prospective clinical settings with more heterogeneous populations, to evaluate workflow integration, generalizability, and overall clinical impact.

𝗢𝗽𝗲𝗻 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 𝗮𝗻𝗱 𝗙𝘂𝘁𝘂𝗿𝗲 𝗪𝗼𝗿𝗸 🚀 In upcoming projects, the PI-CAI consortium aims to improve the generalizability of the PI-CAI AI system and explore applications for prostate cancer detection that extend beyond assistive interpretation. They remain committed to sustaining the PI-CAI benchmark, providing a robust foundation for independent evaluation and future research. The study has been funded, in parts, by the European Commission Horizon 2020: ProCAncer-I project and Health~Holland.

We congratulate Jasper Twilt, Anindo Saha, Joeran Bosma, Jurgen Fütterer, Henkjan Huisman, Maarten de Rooij, and the rest of the PI-CAI consortium for their work!