Objectives: The aim of this study was to explore a modified model to simplify the diagnostic process for American College of Radiology Breast Imaging Reporting and Data System category 3 to 5 breast lesions and evaluate the model's diagnostic efficacy.
Methods: A retrospective review was conducted of breast lesions examined by B-mode ultrasound (US) and contrast-enhanced ultrasound (CEUS) and diagnosed by histopathologic examination from October 2016 to August 2019. The US characteristics of a combined model (US + CEUS model) with relatively high diagnostic value were selected by a lasso regression analysis to establish a modified model. Two nomograms were developed. The results were internally validated by bootstrap resampling.
Results: Overall, 206 breast lesions in 198 patients, 136 benign and 70 malignant, were included. Ultrasound characteristics included in the modified model were shape, margin, microcalcification, enhanced intensity, enhanced scope, and enhanced time. For the US + CEUS model and modified model, sensitivities were 94.3% and 93.3%; specificities were 85.9% and 81.4%; accuracies were 89.3% and 88.7% and areas under the curve were 0.957 and 0.944, respectively. No statistically significant differences were seen in the diagnostic efficacies of the models (P = .097). Bootstrap-corrected sensitivities, specificities. and accuracies of the models were consistent with these results. Bootstrap-corrected areas under the curve for the US + CEUS model and the modified model were 0.955 and 0.940, respectively.
Conclusions: The modified model with fewer indicators conveniently and effectively diagnosed the malignancy of Breast Imaging Reporting and Data System category 3 to 5 breast lesions without reducing diagnostic efficacy.
Keywords: Breast Imaging Reporting and Data System; breast; diagnosis; nomogram; ultrasound.
© 2020 American Institute of Ultrasound in Medicine.