The value of [¹⁸F]PSMA-1007 PET/CT combined with mpMRI in diagnosing prostate cancer with PI-RADS 3–4 lesions

Background: Accurate differentiation between malignant and benign prostate lesions remains challenging, particularly in patients with Prostate Imaging Reporting and Data System (PI-RADS) 3–4 scores. This study aimed to compare the diagnostic performance of fluorine-18 prostate-specific membrane antigen-1007 ([¹⁸F]PSMA-1007) positron emission tomography/computed tomography (PET/CT) and multiparametric magnetic resonance imaging (mpMRI), and to establish a multiparameter model for improving prostate cancer (PCa) detection.

Methods: A retrospective analysis was conducted on 135 patients with PI-RADS 3–4 lesions who underwent [¹⁸F]PSMA-1007 PET/CT between 2 February 2024 and 30 September 2025. Clinical parameters [age, prostate-specific antigen (PSA), free prostate-specific antigen (fPSA)/total prostate-specific antigen (tPSA), prostate volume (PV), prostate-specific antigen density (PSAD)] and imaging variables [maximum standardized uptake value (SUV_max), lesion-to-background ratio (LBR)] were analyzed. Group comparisons were performed using Student’s t-test or Mann-Whitney U test. Diagnostic performance was evaluated by sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV), with paired comparisons using McNemar’s test. Agreement with pathology was assessed using Cohen’s kappa. Logistic regression identified predictors of PCa. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to assess and compare diagnostic performance.

Results: Significant differences between benign prostatic hyperplasia (BPH) and PCa groups were observed in age, PSA, fPSA/tPSA, PV, PSAD, SUV_max, and LBR (all P<0.05), whereas PSA category showed no significant difference (P>0.05). In the overall cohort, [¹⁸F]PSMA-1007 PET/CT demonstrated higher sensitivity (95.5% vs. 73.1%, P<0.001) and accuracy (83.0% vs. 68.1%, P=0.009) than mpMRI, and showed better agreement with pathology (κ=0.66 vs. 0.36). In the PI-RADS 3 subgroup, SUV_max (P=0.003) and LBR (P=0.002) were significant predictors in univariate analysis, whereas SUV_max showed borderline significance in multivariate analysis [odds ratio (OR) =1.26, 95% confidence interval (CI): 0.999–1.590, P=0.051]. The combined model did not significantly improve diagnostic performance (P>0.05). In the PI-RADS 4 subgroup, PV was an independent predictor. The combined model showed better performance than PV alone (P<0.01), but was not superior to SUV_max or LBR alone (P>0.05).

Conclusions: Our findings suggest that [¹⁸F]PSMA-1007 PET/CT may offer better diagnostic performance compared to mpMRI in patients with PI-RADS 3–4. Combining PV with PET/CT parameters may improve diagnostic accuracy in PI-RADS 4 lesions, whereas no additional benefit is observed for combined models in PI-RADS 3 lesions. Further large-scale validation is required.