Ultrasound Radiomics‐Based Logistic Regression Model to Differentiate Between Benign and Malignant Breast Nodules.

Objectives: To explore the potential value of ultrasound radiomics in differentiating between benign and malignant breast nodules by extracting the radiomic features of two‐dimensional (2D) grayscale ultrasound images and establishing a logistic regression model. Methods: The clinical and ultrasound data of 1000 female patients (500 pathologically benign patients, 500 pathologically malignant patients) who underwent breast ultrasound examinations at our hospital were retrospectively analyzed. The cases were randomly divided into training and validation sets at a ratio of 7:3. Once the region of interest (ROI) of the lesion was manually contoured, Spearman's rank correlation, least absolute shrinkage and selection operator (LASSO) regression, and the Boruta algorithm were adopted to determine optimal features and establish a logistic regression classification model. The performance of the model was assessed using the area under the receiver operating characteristic curve (AUC), and calibration and decision curves (DCA). Results: Eight ultrasound radiomic features were selected to establish the model. The AUC values of the model were 0.979 and 0.977 in the training and validation sets, respectively (P =.0029), indicating good discriminative ability in both datasets. Additionally, the calibration and DCA suggested that the model's calibration efficiency and clinical application value were both superior. Conclusions: The proposed logistic regression model based on 2D grayscale ultrasound images could facilitate differential diagnosis of benign and malignant breast nodules. The model, which was constructed using ultrasound radiomic features identified in this study, demonstrated good diagnostic performance and could be useful in helping clinicians formulate individualized treatment plans for patients. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Ultrasound in Medicine is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)