Prostate: simplified MRI anatomy recognition and training—the P-SMART study: a novel approach to defining prostatic apex, mid-gland and base

Objectives Precise lesion localisation on prostate MRI—specifically, identification of whether at the apex, mid-gland, or base—is essential for effective communication and clinical management. This study aimed to assess interobserver agreement of current practice using the Prostate Imaging-Reporting and Data System (PI-RADS) against a novel anatomy-based approach, P-SMART (Prostate Simplified MRI Anatomy Recognition and Training). Materials and methods This multi-reader, two-phase agreement study, used axial and sagittal T2-weighted MRI from a publicly available dataset. Forty MRI cases were annotated with four regions of interest (ROIs). Readers were asked to document ROI location—apex, mid-gland, or base—first using the PI-RADS method, then following a 3-week washout period and short educational intervention, using the P-SMART method. Interobserver agreement was evaluated using Fleiss’ kappa (κ) statistics and compared following bootstrap resampling using a paired t -test. Subgroup analyses were performed by reader expertise and prostate size. Results Following quality assurance, 156 ROIs were evaluated by 43 readers in each phase of the study. Overall interobserver agreement improved with P-SMART vs PI-RADS (κ = 0.571 vs 0.399; p < 0.001). In subgroup analyses, improvements were consistent for experts (0.663 vs 0.435; p < 0.001), non-experts (0.530 vs 0.393; p < 0.001), large prostates (0.518 vs 0.325; p < 0.001), and small prostates (0.620 vs 0.461; p < 0.001). Conclusion The P-SMART approach, designed to align with the PI-RADS sector map, yielded significantly higher interobserver agreement for prostate lesion-level localisation than the PI-RADS method, across reader experience and gland size, though overall agreement was only moderate. Improved reproducibility may support biopsy targeting, active surveillance monitoring, and treatment planning. Key Points Question Can a novel method of dividing the prostate into apex, mid-gland, and base improve reader agreement compared to current practice? Findings The P-SMART method to divide the prostate outperformed PI-RADS (Fleiss’ kappa 0.571 vs 0.399), improving consistency across experience levels and prostate sizes. Relevance statement More reproducible prostate level assignment supports accurate lesion localisation for biopsy planning, active surveillance, and treatment planning. If validated, P-SMART may improve communication and could be integrated into reporting workflows and software tools.