Objective: To explore the value of constructing a malignant risk prediction model for non mass breast lesions (NML) based on clinical data and ultrasound image features. Methods: 78 lesions from 78 patients with NML who were treated in our hospital from January 2020 to February 2023 were selected to analyze the clinical data and ultrasound characteristics differences between benign and malignant NML patients, and to construct a risk model for predicting malignant NML. Results: Among the 78 patients, 36 were malignant and 42 were benign. Malignant lesions are mainly infiltrating carcinoma, while benign lesions are mainly fibrocystic breast disease. The proportion of malignant disease patients aged ≥ 50 years old was 63.89%, which was higher than that of benign disease patients (P<0.05). The proportion of patients with malignant lesions who have a history of breastfeeding was 19.44%, which was lower than that of patients with benign lesions (P<0.05). The family history of breast cancer in patients with malignant lesions was 44.44%, which was higher than that in patients with benign lesions (P<0.05). The proportion of CEA>2.1 ng/ml in patients with malignant lesions was 58.33%, which was higher than that in patients with benign lesions (P<0.05).The proportion of CA15-3>18.7 U/ml in patients with malignant lesions was 63.89%, which was higher than that in patients with benign lesions (P<0.05). The proportion of malignant disease patients with miR-194>1.60 was 72.22%, which was higher than that of benign disease patients (P<0.05).The proportion of malignant lesions with structural distortion, blood flow signals II-III, posterior echo attenuation, and microcalcification was 44.44%, 41.67%, 52.78%, and 66.67%, respectively, which was higher than that of benign lesions (P<0.05). Logistic regression analysis showed that family history of breast cancer, structural distortion, blood flow type and posterior echo attenuation,CEA,CA15-3 and miR-194 were the influencing factors of NML malignancy (P<0.05), the area under the ROC curve predicted by the equation for malignant NML was 0.846 (95% CI: 0.762-0.930), P<0.05, sensitivity and specificity were 73.50% and 77.30%, respectively. Conclusion: Building a predictive model for malignant NML based on clinical data and ultrasound image features has certain application value and is worthy of further clinical research.