Validation of AI-derived whole prostate volumetry versus ellipsoid formula for PSA density-based risk stratification

Le résumé fourni par la source

PURPOSE: State-of-the-art AI tools offer automated prostate MR volumetry, potentially improving accuracy and efficiency compared to approximation using the ellipsoid formula (A x P x L x 0.52). However, data on the clinical validity of AI-derived volumetry (AI-V) for PSA density (PSAD)-based risk stratification remain limited. We therefore investigated the correlation of AI-V- and ellipsoid formula volume (EL-V)-calculated PSAD with ISUP group and their diagnostic performance for the detection of clinically significant prostate cancer (csPCa). MATERIALS AND METHODS: In this IRB-approved retrospective single-centre study, 3-T prostate MRI datasets of 171 patients (median age 68 years) undergoing TRUS-guided transperineal fusion biopsy were analysed. EL-V was calculated by three readers, while AI-V was obtained using a commercially available AI tool. Associations between PSAD and ISUP group were assessed using Spearman's rho and linear regression, and diagnostic performance using receiver operating characteristic (ROC) analysis and predefined PSAD thresholds. RESULTS: Both AI-derived and EL-V-calculated PSAD showed significant positive correlations with ISUP group (all p<0.001). For csPCa detection, AI-V-derived and EL-V-calculated PSAD demonstrated overall comparable diagnostic performance, with AI-V-derived PSAD achieving an AUC of 0.79 and EL-V-calculated PSAD showing AUC values ranging from 0.76 to 0.78 across readers. Only one reader demonstrated a significantly lower AUC compared with AI-V-derived PSAD (p=0.002). At predefined PSAD thresholds (0.10-0.20), sensitivity and specificity were comparable between both methods. CONCLUSION: AI-based prostate volumetry represents a valid alternative to ellipsoid formula-calculated volumetry with non-inferior diagnostic performance for csPCa detection.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.