Accepted for/Published in: Journal of Medical Internet Research
Date Submitted: Jul 20, 2025
Date Accepted: Oct 11, 2025
Artificial intelligence-enabled imaging for predicting preoperative extraprostatic extension in prostate cancer: a systematic review and meta-analysis
ABSTRACT
Background:
Artificial intelligence (AI) techniques, particularly those employing machine learning (ML) and deep learning (DL) to analyze multimodal imaging data, have shown considerable promise in enhancing preoperative prediction of extraprostatic extension (EPE).
Objective:
This meta-analysis explores the diagnostic performance of artificial intelligence-enabled imaging techniques versus radiologists for predicting preoperative EPE in prostate cancer (PCa).
Methods:
We conducted a systematic review of literature from PubMed, Embase, and Web of Science, following PRISMA-DTA guidelines. The included studies applied AI techniques to predict EPE using mpMRI and PSMA PET imaging. Sensitivity, specificity, and area under the curve (AUC) for both internal and external validation sets were extracted and combined using a bi-variate random-effects model. The quality of the included studies was assessed using the modified QUADAS-2 tool.
Results:
A total of 21 studies were analyzed. The mpMRI-based AI demonstrated a pooled sensitivity of 0.78, specificity of 0.76, and an AUC of 0.84, significantly outperforming traditional radiologists, whose pooled sensitivity for detecting EPE was 0.69, specificity was 0.72, and AUC was 0.76. Conversely, the PSMA PET-based AI showed a pooled sensitivity of 0.73, specificity of 0.61, and an AUC of 0.74, indicating moderate performance but no significant advantage over either mpMRI-based AI or radiologists.
Conclusions:
This study indicates that the mpMRI-based AI model has higher sensitivity and AUC values compared to radiologists. However, the PSMA PET-based AI shows no additional advantage in diagnostic performance for predicting preoperative EPE in prostate cancer, whether compared to mpMRI or human-interpreted PET. Limitations include the retrospective design and high heterogeneity which may introduce bias and affect generalizability. Larger, diverse cohorts are essential for confirming these findings and optimizing the integration of AI in clinical practice.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.