Clinical and MRI variables in decision support systems for prostate MRI: A systematic review of decision support...

Workforce shortages and rising demand for MRI have increased interest in clinical decision support systems (CDSSs) to standardise imaging workflows. The Prostate Imaging-Reporting and Data System (PI-RADS) committee recommends real-time radiologist input to guide MRI pathways, but workforce constraints may limit consistent implementation. This systematic review examined CDSSs used in prostate MRI, including decision tools, nomograms, and risk calculators, to identify the clinical and MRI-derived variables they incorporate and assess their relevance for future development.

A systematic search of Medline, Cochrane, CINAHL, Web of Science, and ProQuest was conducted in June 2025. Eligible studies were original research published in English since January 2015 describing development, validation, or clinical use of a CDSS using structured clinical and MRI-derived variables for prostate cancer diagnosis, pre-biopsy risk stratification, or staging. Exclusion criteria included radiomics-only studies, non-primary research, studies without MRI variables, and those lacking external validation. Two reviewers independently screened studies, extracted data, and assessed risk of bias using PROBAST. Certainty of evidence was appraised using the GRADE framework.

Twenty-two studies met inclusion criteria: fifteen evaluated nomograms, five described risk calculators, and two reported predictive models. None assessed a fully implemented CDSS. Common predictors included PI-RADS (82 %), prostate-specific antigen density (64 %), age (64 %), prostate-specific antigen (41 %), and prostate volume (23 %). Most tools showed strong discriminative accuracy (AUC >0.80), though calibration and decision curve analysis were inconsistently reported.

Validated clinical and MRI predictors support robust CDSSs, but heterogeneity and lack of implementation limit evidence. Prospective multicentre validation is needed.

Radiographer-facing tools integrating key predictors could guide contrast use, staging, and workflow decisions, improve diagnostic accuracy and reduce unnecessary contrast administration.