Objective The aim was to investigate the relationship between imaging phenotypic features and TP53 mutation patterns in bladder cancer(BLCA)based on radiomics,deep learning,and pathomic through multi-omics analysis.Methods Multi-omics data of 57 BLCA patients were downloaded from public databases.Radiomics and deep learning features were extracted from pre-operative arterial phase CT scans,while pathomic features were extracted from post-operative hematoxylin and eosin(H&E)stained pathological images.After dimensionality reduction using principal component analysis and Relief,a nomogram for TP53 mutation prediction was developed based on the random forest algorithm.The performance of the nomogram was evaluated using the area under the receiver operating characteristic curve(AUC).Results After dimensionality reduction of 21 radiomics features,9 deep learning features,and 9 pathomic features,24 features were identified for developing the nomogram.The AUCs for the training and validation cohorts were 0.95 and 0.87,respectively,with accuracy of 0.88 and 0.88,sensitivity of 0.87 and 0.90,and specificity of 0.88 and 0.86.Conclusion The complementary role of multi-omics imaging phenotypic information was utilized to elucidate the relationship between imaging phenotypic features and BLCA-TP53 mutation patterns.This approach could serve as a non-invasive surrogate marker for TP53 mutations,providing a basis for precision medicine management.
NETL
NSTL
万方数据
