Differentiating triple-negative breast cancer and atypical fibroadenomas using an ultrasound-based radiomics nomogram

To investigate the application value of ultrasound radiomics technology in differentiating TNBC from atypical fibroadenoma (AFA) and to develop a nomogram model.

In this study, 75 patients with TNBC and 90 patients with AFA who underwent surgery were enrolled and randomly divided into training (n = 114) and validation cohorts (n = 51). Radiomic features were extracted from the images. LASSO regression analysis and stepwise regression were used for features selection. A prediction model was developed by combining multivariate logistic regression with the selected imaging biomarkers, resulting in the generation of a nomogram. A confusion matrix was used to visualize the distribution of correct and misclassified classifications. Finally, the validity of the model was assessed by using the receiver operator characteristic curve and calibration curve.

Multivariate analysis based on ultrasound features identified elasticity score, fibrous strands within the mass and lateral acoustic shadow as independent factors for differentiating TNBC from AFA. The radiomics signature, composed of 4 selected features, achieved good diagnostic performance. The nomogram incorporating the radiomics signature demonstrated favorable diagnostic efficacy, and the AUC of the training set and the validation set were 0.977 and 0.982, which outperformed the ultrasound model (AUC = 0.939). The calibration curve demonstrated the good clinical utility of the radiomics nomogram.

This study systematically analysed the ability of ultrasound radiomics to differentiate TNBC from AFA and the ultrasound-based radiomics nomogram effectively improved the diagnostic accuracy of TNBC.