Development and validation of a malignant risk prediction model for breast cystic-solid lesions based on clinical and ultrasound features.

OBJECTIVE: Accurate differentiation of benign and malignant breast cystic-solid lesions (BCSLs) is crucial due to the significant differences in their treatment strategies. This study aimed to develop and validate a malignant risk prediction model for BCSLs based on clinical and ultrasound (US) features. METHODS: A total of 312 patients with BCSLs were enrolled and divided into a training cohort (218 cases) and a validation cohort (94 cases). The least absolute shrinkage and selection operator (LASSO) was used to screen the most critical clinical and US features, and a prediction model was constructed using logistic regression. Calibration curves, receiver operating characteristic (ROC) curves, and decision curve analysis (DCA) were applied to evaluate the model’s performance and clinical value. RESULTS: Five key features were identified through LASSO regression: age, maximum diameter, edge, calcification in the solid component, and vascularity in the solid component. A prediction model was subsequently established and visualized as a nomogram. Calibration curves demonstrated good consistency of the model. The area under the curve (AUC) values of the training and validation cohorts were 0.821 (95% CI = 0.764–0.879) and 0.837 (95% CI = 0.754–0.919), respectively. DCA indicated that the model had practical value in clinical application. CONCLUSION: The prediction model based on clinical and US features can accurately predict the malignant risk of BCSLs and optimize clinical decision-making.