Objective To develop and validate a machine learning model for predicting disease-free survival (DFS) in invasive breast cancer by integrating clinicopathologic and ultrasound-radiomics features. Methods A total of 978 eligible patients with invasive breast cancer were retrospectively enrolled, and 4 414 ultrasound images were collected and allocated to training, validation, and test sets at a patient-level ratio of 6∶2∶2. Pyradiomics software was employed to extract tumor ultrasound radiomic features (from both 2-dimensional and color Doppler images). Feature dimensionality reduction was performed using t-distributed stochastic neighbor embedding (t-SNE), least absolute shrinkage and selection operator (LASSO), and correlation coefficients. The selected features were subsequently incorporated into a nomogram prediction model. Model performance was evaluated using the concordance index (C-index), area under curve (AUC) of the receiver operating characteristic curve (ROC), precision, recall, specificity, and F1-score. Additionally, the contribution of individual features to model predictions was quantified with Shapley additive explanations (SHAP). Results Twelve radiomic features and 6 clinicopathological features closely associated with DFS events in invasive breast cancer patients were ultimately selected. The nomogram model constructed from these features achieved a C-index of 0.787, AUC of 0.827, precision of 0.623, recall of 0.701, specificity of 0.812, and F1-score of 0.584. SHAP analysis revealed that the radiomics score (Rad-Score) exerted a significantly positive influence on model outputs. Conclusion The model based on ultrasound radiomics and machine learning algorithms can effectively predict DFS events in invasive breast cancer patients, with radiomic features serving as significant independent prognostic factors for DFS events.