OBJECTIVE: To refine the existing CT algorithm to enhance inter-reader agreement and improve the diagnostic performance for clear-cell renal cell carcinoma (ccRCC) in solid renal masses less than 4 cm. METHODS: A retrospective collection of 331 patients with pathologically confirmed renal masses were enrolled in this study. Two radiologists independently assessed the CT images: in addition to heterogeneity score (HS) and mass-to-cortex corticomedullary attenuation ratio (MCAR), measured parameters included ratio of major diameter to minor diameter at the maximum axial section (Major axis / Minor axis), tumor-renal interface, standardized heterogeneity ratio (SHR), and standardized nephrographic reduction rate (SNRR). Spearman's correlation analysis was performed to evaluate the relationship between SHR and HS. Univariate and multivariate logistic regression analyses were employed to identify independent risk factors and then CT-score was adjusted by those indicators. The diagnostic efficacy of the modified CT-scores was evaluated using ROC curve analysis. RESULTS: The SHR and heterogeneity grade (HG) of mass were correlated positively with the HS (R = 0.749, 0.730, all P <
0.002). Logistic regression analysis determined that the Major axis / Minor axis (>
1.16), the tumor-renal interface (>
22.3 mm), and the SNRR (>
0.16) as additional independent risk factors to combine with HS and MCAR. Compared to the original CT-score, the two CT algorithms combined tumor-renal interface and SNRR showed significantly improved diagnostic efficacy for ccRCC (AUC: 0.770 vs. 0.861 and 0.862, all P <
0.002). The inter-observer agreement for HG was higher than that for HS (weighted Kappa coefficient: 0.797 vs. 0.722). The consistency of modified CT-score was also superior to original CT-score (weighted Kappa coefficient: 0.935 vs. 0.878). CONCLUSION: The modified CT algorithms not only enhanced inter-reader consistency but also improved the diagnostic capability for ccRCC in small renal masses.