Radiogenomic analysis of muscle-invasive bladder cancer using CT-based texture analysis.

Area of Science:

• Oncology
• Radiology
• Genomics

Background:

• Urothelial bladder cancer (UBC) is heterogeneous, with molecular profiling revealing prognostic and therapeutic targets.
• Genomic testing for UBC is limited, hindering clinical application.
• Computed tomography (CT) is standard for staging and surveillance, offering potential for noninvasive tumor biology assessment.

Purpose of the Study:

• To investigate if radiomics features extracted from CT images can predict molecular characteristics of muscle-invasive bladder cancer (MIBC).
• To assess the potential of radiomics as a noninvasive adjunct to genomic profiling in MIBC.

Main Methods:

• Integrated The Cancer Genome Atlas genomic data with Cancer Imaging Archive CT images for 89 MIBC patients.
• Extracted 488 radiomic texture metrics from segmented tumors using an in-house pipeline.
• Trained Random Forest, Extreme Gradient Boosting, and Elastic Net classifiers to predict mutations, tumor mutational burden (TMB), and mRNA expression via 10-fold cross-validation.

Main Results:

• Radiomics reliably predicted mutations in EP300, FGFR3, and ARID1A (AUCs 0.77-0.75).
• Models identified high-TMB tumors (AUC=0.61), poor-prognosis transcriptomic signatures (AUCs 0.73, 0.65), and key gene expressions (e.g., CDKN1A AUC=0.78, CASP3 AUC=0.71).
• Radiomics differentiated the luminal infiltrated molecular subtype (AUC=0.69).

Conclusions:

• CT-derived radiomics features capture biologically and clinically relevant information in MIBC.
• Radiomics shows potential as a noninvasive, scalable tool to complement genomic profiling in MIBC.