Radiomics in mammographic diagnosis of breast cancer

Background: Breast cancer is the most frequently diagnosed type of cancer among women. Triple-negative breast cancer (TNBC) is a specific subtype of breast cancer that does not express estrogen receptor (ER), progesterone receptor (PR) or human epidermal growth factor receptor 2 (HER2), with high invasiveness, high metastatic potential and poor prognosis. TNBC has a close associations with mutations in the BRCA pathway. The current diagnostic workflow is based on imaging and biopsy to confirm the subtype. Objective: This study evaluates the feasibility of radiomic analysis of mammographic images for identifying radiomic features that can differentiate BRCA-mutated from non-mutated TNBC lesions, offering a potential non-invasive diagnostic and prognostic tool. Methods: We performed a retrospective study on 52 patients histologically diagnosed with TNBC and evaluated at the Oncology Department of Modena University Hospital who underwent digital breast mammography between February 2010 and August 2021. Radiomic features were extracted from manually segmented regions of interest (ROIs) of the tumor and the healthy contralateral gland. A total of 195 features was extracted and the more significant ones were selected using the Maximum Relevance Minimum Redundancy (mRMR) algorithm. Three different models were built: one based solely on tumor-derived features, one based on glandular tissue features, and one combining both tumor and gland features. The selected features in each model were used to train four classifiers. Models were evaluated using a stratified 10-fold cross validation. Results: The results of this work confirm the feasibility of a radiomic analysis on mammographic examination. The model with gland features only showed the best performance, suggesting that including both characteristics from the tumor and the healthy contralateral gland improves the predictive power. Conclusion: These results introduce the possibility of using mammography not only as a valid screening test but also, integrated with radiomic analysis techniques, as a tool to stratify patients for their mutational status and risk of disease. Further studies with multicenter datasets are required to confirm these preliminary results.