This study aimed to identify the risk factors and sonographic variables that could be integrated into a predictive model for endometrial cancer (EC) and atypical endometrial hyperplasia (AEH) in women with abnormal uterine bleeding (AUB). This retrospective study included 1837 patients who presented with AUB and underwent endometrial sampling. Multivariable logistic regression was developed based on clinical and sonographic covariates [endometrial thickness (ET), resistance index (RI) of the endometrial vasculature] assessed for their association with EC/AEH in the development group (n=1369), and a predictive nomogram was proposed. The model was validated in 468 patients. Histological examination revealed 167 patients (12.2%) with EC or AEH in the development group. Using multivariable logistic regression, the following variables were incorporated in the prediction of endometrial malignancy: metabolic diseases [odds ratio (OR)=7.764, 95% confidence intervals (CI) 5.042-11.955], family history (OR=3.555, 95% CI 1.055-11.971), age ≥40 years (OR=3.195, 95% CI 1.878-5.435), RI ≤0.5 (OR=8.733, 95% CI 4.311-17.692), and ET ≥10 mm (OR=8.479, 95% CI 5.440-13.216). A nomogram was created using these five variables with an area under the curve of 0.837 (95% CI 0.800-0.874). The calibration curve showed good agreement between the observed and predicted occurrences. For the validation group, the model provided acceptable discrimination and calibration. The proposed nomogram model showed moderate prediction accuracy in the differentiation between benign and malignant endometrial lesions among women with AUB.