Semantic Computed Tomography Features for Predicting BRCA1-associated Protein 1 and/or Tumor Protein p53 Gene...

The purpose of this study was to explore the semantic computed tomography (CT) features associated with BRCA1-associated protein 1 (BAP1) and/or tumor protein p53 (TP53) mutation in clear cell renal cell carcinoma (ccRCC).

Clinical characteristics and gene mutation information of 336 ccRCC patients were retrieved from The Cancer Genome Atlas-Kidney Renal Clear Cell Carcinoma database (TCGA-KIRC). Kaplan-Meier analysis was performed to examine prognosis by gene mutation. The CT imaging data and gene mutation information of 156 ccRCC patients treated between January 2019 and January 2021 (the training cohort) were retrospectively analyzed. The CT imaging information and gene mutation data of 123 patients with ccRCC were downloaded from The Cancer Imaging Archive and The Cancer Genome Atlas database (the external validation cohort). Univariate Chi-square test and multivariate binary logistic regression analysis were performed to determine predictors of gene mutation; a nomogram was developed using these predictors. Receiver operating characteristic curve analysis and the Hosmer-Lemeshow test were performed to evaluate the performance of the nomogram.

Kaplan-Meier analysis showed that BAP1 and/or TP53 mutation was significantly correlated with worse survival outcome. Multivariate binary logistic regression analysis indicated ill-defined margin (P = .001), spiculated margin (P = .018), renal vein invasion (P = .002), and renal pelvis invasion (P = .001) were independent predictors of BAP1 and/or TP53 mutation. A nomogram containing these 4 semantic CT features was constructed; the area under the receiver operating characteristic curves was 0.872 (95% CI, 0.809-0.920). The Hosmer-Lemeshow test showed acceptable goodness-of-fit for the nomogram (X² = 1.194, P = .742). The nomogram was validated in the validation cohort; it showed good accuracy (area under the receiving operating characteristic curve = 0.819, 95% CI, 0.740-0.883) and was well calibrated (X² = 3.934, P = .559).

Semantic CT features are a potential and promising method for predicting BAP1 and/or TP53 mutation status in ccRCC patients.