Standardizing PSMA PET/CT reporting in prostate cancer: from structured frameworks to AI-assisted interpretation

PSMA PET/CT is increasingly used for prostate cancer staging, restaging, treatment selection, and therapy response assessment. In parallel, several interpretation frameworks have been developed to standardize reporting. PROMISE version 2 integrates the PRIMARY score for local disease, molecular imaging TNM staging, and a PSMA expression score. PSMA-RADS 2.0 supports lesion-level characterization, whereas response systems such as the PSMA PET progression criteria and RECIP 1.0 focus on treatment response assessment. Despite the availability of these systems and professional society guidance, real-world reporting remains heterogeneous, with limited framework adoption, inconsistent lesion documentation, variable handling of incidental findings, and suboptimal longitudinal comparison, all of which may complicate multidisciplinary communication and clinical decision-making. Emerging automation and artificial intelligence (AI)-based tools may help address some of these gaps by supporting lesion detection, quantification, longitudinal matching, and prepopulation of structured report elements; however, their current role is best regarded as physician-supervised decision support rather than autonomous interpretation. This article provides a pragmatic interpretation workflow for PSMA PET/CT readers. We propose a clinical question-driven approach to framework selection, a compartment-based search pattern, and a minimal synoptic reporting template. Targeted pearls and pitfalls emphasize common abdominopelvic mimics and tracer-dependent issues. We also address frequent interpretive challenges, including limited sensitivity for microscopic nodal disease, partial-volume effects in uptake assessment, the trade-off between comprehensive frameworks and real-world feasibility, and the evolving role of AI-assisted structured reporting. Take-home message: A practical reporting approach is to standardize the impression with PROMISE V2 staging, describe equivocal lesions using PSMA-RADS terminology, and incorporate these elements into a minimal synoptic template. Physician-supervised automation and AI-assisted tools may further support lesion detection, quantification, longitudinal matching, and structured-report prepopulation within this workflow.