In Brief | Radiology & Imaging | Oncology | Prostate Cancer
AI Assistance Enhances Accuracy in Prostate Cancer MRI Diagnosis, Especially for Non-expert Readers
Time to read: 03:08Time to listen: 06:43
Originally Published: 5June 2025
Source: JAMA Network Open
Type of article: In Brief
MedED Catalogue Reference: MRDIb003
Category: Radiology & Imaging
Cross Reference: Oncology, Men's Health
Keywords:AI, prostate cancer, oncology, MRI, Diagnosistic
This article is a review of recent studies originally published in the JAMA Network Open, 5 June 2025. Reproduced here under CC-BY license. This article does not represent the original research, nor is it intended to replace the original research. Access the full Disclaimer Information.
Listen to this Review
Study Context and Purpose
Prostate cancer is the most frequently diagnosed malignancy in men globally, accounting for over 14% of all cancer cases. According to the 2022 National Cancer Registry in South Africa, the lifetime risk of developing prostate cancer is 1 in 15.¹
MRI-targeted biopsy has become central to diagnosis, improving detection of clinically significant prostate cancer (csPCa) while reducing unnecessary procedures.
The Prostate Imaging Reporting and Data System (PI-RADS) was introduced to standardise MRI interpretation, but diagnostic accuracy remains limited by inter-reader variability and the level of expertise required.²
With the increasing demand for prostate MRI, artificial intelligence (AI) has emerged as a promising solution. While early studies suggest AI can enhance diagnostic accuracy and consistency, many have lacked robust validation.
The Prostate Imaging–Cancer AI (PI-CAI) challenge was a global, multi-reader, multi-centre initiative designed to rigorously benchmark AI performance in detecting csPCa on MRI.³
It involved over 10,000 MRI exams, of which 2,440 had histologically confirmed Gleason grade group ≥2 prostate cancer. In a subset of 400 test cases, the AI system achieved a superior AUROC of 0.91, outperforming 62 radiologists. This large-scale evaluation confirmed the AI system’s capability to outperform expert readers in CS-PCa detection.
About this Study
Building on the findings of the PI-CAI challenge, researchers conducted a follow-up international observer study to evaluate the clinical impact of AI-assisted MRI interpretation.
The primary objective was to assess whether the high-performing AI system could improve the detection of clinically significant prostate cancer (csPCa) compared to unaided reader assessments.
Secondary objectives included comparing the diagnostic benefit of AI support for non-expert versus expert readers and examining performance across varying diagnostic thresholds to understand how AI influences reader decision-making.
Study Methodology
Between March and July 2024, 61 readers (34 experts and 27 non-experts) from 53 centres in 17 countries participated in a diagnostic observer study using the AI system developed through the international PI-CAI Consortium.
Each reader assessed prostate MRIs with and without AI support, assigning PI-RADS scores (3–5) and patient-level suspicion scores (0–100).
The dataset included 780 biparametric prostate MRI examinations from men suspected of prostate cancer, with histopathology or long-term follow-up confirming disease presence or absence.
The AI model was recalibrated on 420 Dutch cases to generate lesion-detection maps (score 1–10), and the remaining 360 cases, from Dutch and Norwegian centres, were used in the study.
The primary outcome was the detection of csPCa, measured by AUC, sensitivity, and specificity at a PI-RADS threshold of≥3. Secondary outcomes examined diagnostic performance in relation to reader expertise and threshold variation.
Study Findings
Among 360 men included in the observer study (median age 65 years, IQR 62–70), 122 cases (34%) were confirmed to harbour clinically significant prostate cancer (csPCa).
AI assistance was associated with a statistically significant improvement in diagnostic performance.
- The area under the receiver operating characteristic curve (AUC) increased by 3.3%, from 0.882 (95% CI, 0.854–0.910) without AI to 0.916 (95% CI, 0.893–0.938) with AI support (P < .001). ?
- Sensitivity improved from 94.3% to 96.8% (a 2.5% gain; 95% CI, 1.1%–3.9%; P < .001), while specificity increased from 46.7% to 50.1% (a 3.4% gain; 95% CI, 0.8%–6.0%; P = .01) at a PI-RADS threshold of ≥3.
- Secondary analyses showed consistent improvements across varying diagnostic thresholds, with non-expert readers experiencing greater performance gains with AI support compared to their expert counterparts.
AI assistance was associated with a statistically significant improvement in diagnostic performance.
Discussion
This study had several limitations, including its retrospective nature, the use of mixed cohort types, and a controlled reading environment that differed from real-world clinical settings. Generalisability remains to be confirmed in external cohorts with varying disease prevalence, imaging quality, and clinical contexts. Workflow efficiency and clinical applicability were not assessed.
Despite these limitations, the findings demonstrate that AI assistance significantly improves the diagnostic accuracy of biparametric MRI for csPCa, with gains in AUROC, sensitivity, and specificity at a PI-RADS threshold of 3 or higher. Non-expert readers, in particular, benefited most, highlighting the potential of AI to support diagnostic consistency.
Original Study
Twilt JJ, Saha A, Bosma JS, Padhani AR, ...; PI-CAI Consortium et al. AI-Assisted vs Unassisted Identification of Prostate Cancer in Magnetic Resonance Images. JAMA Netw Open. 2025 Jun 2;8(6):e2515672. doi: 10.1001/jamanetworkopen.2025.15672. PMID: 40512493; PMCID: PMC12166490.
Additional References
1.Cancer.org. Prostate Cancer. Retrieved 1 July 2025. https://cansa.org.za/prostate-cancer/
2. Scott, R., Misser, S., Cioni, D., & Neri, E. (2021). PI-RADS v2.1: What has changed and how to report. South African Journal of Radiology, 25(1), 13 pages. doi:https://doi.org/10.4102/sajr.v25i1.2062
3.Saha A, Bosma JS, Twilt JJ, et al; PI-CAI consortium. Artificial intelligence and radiologists in prostate cancer detection on MRI (PI-CAI): an international, paired, non-inferiority, confirmatory study. Lancet Oncol. 2024;25(7):879-887. doi:10.1016/S1470-2045(24)00220-1
Disclaimer
This article is compiled from several resources researched and compiled by the contributor. It is in no way presented as an original work. Every effort has been made to attribute quotes and content correctly. Where possible all information has been independently verified. The Medical Education Network bears no responsibility for any inaccuracies which may occur from the use of third-party sources. If you have any queries regarding this article contact us
Fact-checking Policy
The Medical Education Network makes every effort to review and fact-check the articles used as source material in our summaries and original material. We have strict guidelines in relation to the publications we use as our source data, favouring peer-reviewed research wherever possible. Every effort is made to ensure that the information contained here is an accurate reflection of the original material. Should you find inaccuracies, out of date content or have any additional issues with our articles, please make use of the contact us form to notify us.
