CT imaging analysis differentiating papillary renal neoplasm with reverse polarity from papillary renal cell carcinoma: combined with a radiomics model

Abstract Purpose To assess the computed tomography (CT) findings of papillary renal neoplasm with reverse polarity (PRNRP) and develop a radiomics-based model to distinguish PRNRPs from papillary renal cell carcinomas (PRCCs). Materials and methods We analyzed 31 PRNRPs and 68 PRCCs using preoperative kidney CT. We evaluated CT features that could discriminate PRNRPs from PRCCs. A radiomics signature was constructed using features selected through a least absolute shrinkage and selection operator algorithm. A radiomics-based model incorporating a radiomics signature and subjective CT parameters using multivariate logistic regression was developed. The diagnostic performance of the CT parameters, radiomics model, and their combination was evaluated using the area under the curve (AUC). Results Most of PRNRPs had a round shape (93.5%), well-defined margin (100%), and persistent enhancement (77.4%). Compared with PRCC, PRNRPs exhibited distinct CT features including small size (16.7 vs. 37.7 mm, P < 0.001), heterogeneity (64.5 vs. 32.4%, P = 0.004), enhancing dot sign (16.1 vs. 1.5%, P = 0.001), and high attenuation in pre-contrast CT (44.2 vs. 35.5 HU, P = 0.003). Multivariate analysis revealed smaller mass size (odds ratio OR: 0.9; 95% confidence interval CI 0.9–1.0, P = 0.013), heterogeneity (OR: 8.8; 95% CI 1.9–41.4, P = 0.006), and higher attenuation in pre-contrast CT (OR: 1.1; 95% CI 1.0–1.2, P = 0.011) as significant independent factors for identifying PRNRPs. The diagnostic performance of the combination model was excellent (AUC: 0.923). Conclusion Smaller tumor size, heterogeneity, and higher attenuation in pre-contrast CT were more closely associated with PRNRPs than with PRCCs. Though the retrospective design, small sample size, and single-center data of this study may affect the generalizability of the findings, combining subjective CT features with a radiomics model is beneficial for distinguishing PRNRPs from PRCCs.