Logistic LASSO regression for the diagnosis of breast cancer using clinical demographic data and the BI-RADS lexicon for ultrasonography

• DOI: 10.14366/usg.16045
• Publication Date: 2017-04-14T05:14:47Z
• Authors: Sun Mi Kim; Yongdai Kim; Kuhwan Jeong; Heeyeong Jeong; Jiyoung Kim
• Keywords: Lasso; BI-RADS; Breast imaging; Breast MRI

The aim of this study was to compare the performance image analysis for predicting breast cancer using two distinct regression models and evaluate usefulness incorporating clinical demographic data (CDD) into in order improve diagnosis cancer.This included 139 solid masses from patients who underwent a ultrasonography-guided core biopsy had available CDD between June 2009 April 2010. Three radiologists retrospectively reviewed described each lesion Breast Imaging Reporting Data System (BI-RADS) lexicon. We applied compared methods-stepwise logistic (SL) least absolute shrinkage selection operator (LASSO) regression-in which BI-RADS descriptors were used as covariates. investigated performances these methods agreement terms test misclassification error area under curve (AUC) tests.Logistic LASSO superior (P<0.05) SL regression, regardless whether covariates, errors (0.234 vs. 0.253, without CDD; 0.196 0.258, with CDD) AUC (0.785 0.759, 0.873 0.735, CDD). However, it inferior three 0.168, 0.088, 0.844, P<0.001), but comparable (0.873 0.880, P=0.141).Logistic based on showed better than presence cancer. use supplement significantly improved prediction regression.