Jacquelyn Fields (Medicine) secured first place at the 26th Annual Graduate & Professional Student Research Forum at UNLV with her poster: CEM radiomics for distinguishing lesion from background parenchymal enhancement in patients with invasive breast cancer.
Collaborating with B.A. Varghese, M. Perkins, S. Cen, X. Lei, J. Jamie, B. Desai, M. Thomas, D.H. Hwang, S. Lee, L.H. Larsen, and M. Yamashita from USC and UCLA, Fields explored contrast-enhanced mammography (CEM) to differentiate between invasive breast cancer lesions and background parenchymal enhancement in 41 women. An experienced radiologist manually outlined regions of interest on the CEM images for analysis. The team used LifEx software to extract 109 radiomics metrics from these regions, spanning six texture families. They then developed predictive models with Random Forest, Real Adaboost, and ElasticNet classifiers, achieving AUCs of 0.83, 0.82, and 0.74, respectively, in distinguishing malignant lesions from background enhancements. The study observed a decline in discrimination ability as background enhancement increased. Further analysis based on hormone receptor status identified significant metric differences across cancer subtypes, demonstrating the models’ effectiveness in discriminating malignant lesions with AUCs greater than 0.8. This research highlights CEM radiomics' value in refining breast cancer diagnosis and tailoring treatment strategies.
