Objective: To investigate the diagnostic efficacy of a nomogram prediction model constructed based on ultrasound parameters for malignant lesions of the endometrium in perimenopausal women. Methods: A total of 637 patients who underwent hysteroscopic examination from August 2021 to December 2023 at our hospital were selected as the study subjects. All patients underwent hysteroscopic examination and pathological diagnosis, and were divided into two groups based on the pathological results ,benign endometrial lesion group (n=567) and malignant endometrial lesion group (n=70). The general data, pelvic ultrasound, and laboratory indicators of the two groups were compared. Univariate and multivariate analyses were used to select indicators with statistical differences, and a nomogram model was constructed and evaluated for diagnostic efficacy. Results: There were statistical differences in age, BMI, functional uterine bleeding, postmenopausal vaginal bleeding, oligomenorrhea, menstrual cycle, menopause, hypertension, breast hyperplasia, cervical polyps, pelvic inflammatory disease, CA125, fasting blood sugar, triglycerides, lesion volume, echogenicity, blood flow signal, endometrial thickness, and endometrial PI, etc., between the two groups (P<0.05). After selecting factors with statistical differences, LASSO regression and multivariate Logistic regression analysis were performed in the validation set, and it was ultimately shown that age, BMI, triglycerides, lesion blood flow signal, endometrial PI, and endometrial thickness, etc., had statistical significance (P<0.05). A nomogram model was constructed in the training set, and the predictive model"s indicators were the model"s calibration and discrimination. After 1000 times of Bootstrap resampling internal validation in the training set, the calibration was good (Brier=0.077), and the discrimination was good (C-statistic=0.794); when the predicted risk value >0.45, the calibration was high, when the risk value >0.2, the risk value was underestimated, and when the risk value was between (0.05-0.2), the risk value was overestimated. After 1000 times of Bootstrap resampling internal validation in the validation set, the calibration was good (Brier=0.077), and the discrimination was good (C-statistic=0.767); when the predicted risk value >0.45, the calibration was high, when the risk value was between (0.25-0.80), the risk value was underestimated, the negative predictive function was good, and when the risk value was between (0.05-0.25), the risk value was overestimated. The clinical application value of the model at different decision threshold probabilities was evaluated through DCA from the net benefit indicator. After drawing the DCA curves of different models (ultrasound combined with clinical comprehensive prediction model vs. ultrasound-only prediction model), it was found that within the decision threshold probability interval of (0-0.75), the clinical application value of the comprehensive prediction model was higher. Then, the risk stratification was carried out using the nomogram, and the clinical impact curve was drawn, showing that when the decision threshold probability >0.2, the prediction model was roughly close to the true probability. Conclusion: Univariate and multivariate analyses showed that age, BMI, triglycerides, lesion blood flow signal, endometrial PI, and endometrial thickness are high-risk factors for malignant changes of endometrial lesions in perimenopausal women. Based on this, a comprehensive nomogram for predicting malignant changes of endometrial lesions in perimenopausal women based on ultrasound parameters was constructed, which has high accuracy, good calibration, and discrimination, and can provide guidance for formulating the best benefit-risk individualized follow-up treatment plans for patients with endometrial lesions.