Using radiomics model for predicting extraprostatic extension with PSMA PET/CT studies: a comparative study with the Mehralivand grading system

PMCID: PMC12177976

PMID: 40533824

DOI: 10.1186/s40644-025-00894-w

Journal: Cancer imaging : the official publication of the International Cancer Imaging Society

Publication Date: 2025-6-18

Authors: Bian L, Liu F, Peng Y, Liu X, Li P, et al.

Key Points

• Radiomics model using PSMA PET/CT demonstrates superior EPE prediction compared to traditional MRI-based methods
• AUC of 76.8% with 81.5% specificity and 81.6% negative predictive value
• Offers promising potential for more accurate preoperative prostate cancer staging and surgical planning

Summary

This study investigated the potential of a radiomics model using PSMA PET/CT to predict extraprostatic extension (EPE) in prostate cancer, comparing its performance against the Mehralivand Grading System based on multiparametric MRI. The research involved 206 patients who underwent radical prostatectomy, with a subset of 63 patients receiving both PSMA PET/CT and mpMRI imaging. The radiomics model, developed using Support Vector Machine and Random Forest algorithms, demonstrated superior diagnostic performance in identifying EPE compared to traditional reader-based assessments.

The radiomics model achieved a notable area under the curve (AUC) of 76.8% (95% CI: 64.4–86.5%), with a sensitivity of 72.0%, specificity of 81.5%, and high negative predictive value of 81.6%. Critically, statistical analysis using DeLong's test revealed that the radiomics model significantly outperformed all three independent readers (p-values ranging from 0.001 to 0.013), suggesting its potential as a more objective and accurate method for preoperative EPE assessment in prostate cancer.