Radiomic features based on pyradiomics predict CD276 expression associated with breast cancer prognosis

CD276 is a promising immune checkpoint molecule with significant therapeutic potential. Several clinical trials are currently investigating CD276-targeted therapies.

This study aims to assess the prognostic significance of CD276 expression levels and to predict its expression using a radiomic approach in breast cancer (BC).

A cohort of 840 patients diagnosed with BC from The Cancer Genome Atlas was included in this study. The Cancer Imaging Archive provided 98 magnetic resonance imaging (MRI) scans, which were randomly allocated to training and validation datasets in a 7:3 ratio. The association between CD276 expression and patient survival was assessed using Cox regression analysis. Feature selection was performed using the maximum relevance minimum redundancy algorithm and recursive feature elimination. Subsequently, support vector machine (SVM) and logistic regression (LR) models were constructed to predict CD276 expression.

The expression of CD276 was found to be elevated in BC. It was an independent risk factor for overall survival (hazard ratio = 1.579, 95 % CI: 1.054-2.366). There were eight radiomic features selected in total. In both the training and validation subsets, the SVM and LR models demonstrated favorable predictive abilities with AUC values of 0.744 and 0.740 for the SVM model and 0.742 and 0.735 for the LR model. These results indicate that the radiomic models efficiently differentiate the CD276 expression status.

CD276 expression levels can have an impact on cancer prognosis. The MRI-based radiomic signature described in this study can discriminate the CD276 expression status.