Computed Tomography Radiomic Features Reflect ATP-Binding Cassette Subfamily G Member 2 Expression in Ovarian Serous Cystadenocarcinoma: A Radiogenomic Study in the Context of Chemoresistance

Ovarian cancer is frequently associated with recurrence and chemoresistance, representing a major clinical challenge. Among the mechanisms underlying treatment resistance, ATP-binding cassette subfamily G Member 2 (ABCG2) plays an important role in drug efflux and tumor adaptation. At the same time, radiomics has emerged as a non-invasive approach for identifying imaging biomarkers that may reflect tumor molecular characteristics. This study investigated the relationship between computed tomography (CT)-derived radiomic features and ABCG2 expression in ovarian cancer. This retrospective radiogenomic study included 70 patients from the Cancer Genome Atlas with available contrast-enhanced CT images and matched gene expression data. A single axial CT slice showing the largest tumor area was selected for analysis. Tumors were manually segmented using Horos, and the statistical analysis included univariate testing with false discovery rate correction and exploratory multivariable modeling using Least absolute shrinkage and selection operator (LASSO) and ridge logistic regression. Five radiomic features showed nominally significant differences between ABCG2-positive and ABCG2-negative tumors, mainly involving tumor shape and attenuation characteristics. However, none remained significant after correction for multiple comparisons. Multivariable models showed limited predictive performance, with area under the curve values of 0.508 for LASSO and 0.598 for ridge regression. Although predictive performance was modest, CT-derived radiomic features appeared to capture subtle imaging differences associated with ABCG2 expression. These findings support the potential role of radiomics as a non-invasive tool for exploring molecular tumor phenotypes in ovarian cancer.