# Transvaginal ultrasound-detected endometrial echogenicity heterogeneity in diagnosing endometrial carcinoma: risk factors and nomogram-based prediction model

## TL;DR

This study creates a predictive model using ultrasound findings and risk factors to help diagnose endometrial cancer more effectively.

## Contribution

A novel nomogram model integrating transvaginal ultrasound and clinical factors for endometrial cancer prediction is developed and validated.

## Key findings

- The model showed excellent discrimination and stable predictive consistency in validation.
- Echogenicity heterogeneity and age at menopause were the most influential predictors.
- The model's predictions aligned closely with actual outcomes, showing high agreement.

## Abstract

This study aimed to develop a nomogram prediction model based on risk factors associated with endometrial cancer (EC) diagnosed via transvaginal ultrasound (TVS)-detected non-uniform echogenicity.

A retrospective analysis of 564 female patients (control group: normal/benign lesions, n = 475; observation group: EC, n = 89) was conducted. TVS findings were compared with pathological diagnoses, and receiver operating characteristic (ROC) analysis was performed to assess diagnostic performance. Patients were split 7:3 into training and internal validation sets. Multivariate logistic regression identified predictors for nomogram construction, which was validated for performance and utility. SHAP (SHapley Additive exPlanations) analysis was applied for model interpretability, and clinical cases were used for demonstration.

The area under the curve (AUC) of TVS detection of endometrial echogenicity heterogeneity for EC diagnosis was 0.726. Multivariate logistic regression analysis showed that body mass index (BMI), hypertension, diabetes, age at menopause > 50 years, and non-uniform echogenicity were risk factors for EC. The prediction model constructed demonstrated good calibration performance in the training set and excellent discrimination ability and stable predictive consistency in the internal validation set. Decision curve analysis further confirmed its clinical utility. SHAP analysis of the established nomogram revealed that age at menopause and heterogeneous endometrial echogenicity were the most influential predictors in the model, with echogenicity heterogeneity consistently associated with an increased risk of EC. When the nomogram predicted an EC probability of ≥ 0.5, the number of predicted positive cases was 93 (16.49%), showing no statistically significant difference (P > 0.05) from the 89 actually confirmed EC cases (15.78%). This indicates a high agreement between model predictions and actual outcomes.

TVS detection of heterogeneous endometrial echogenicity holds supplementary diagnostic value for EC. The nomogram model constructed in this study integrates key clinical and sonographic features, demonstrating favorable predictive performance and clinical applicability. SHAP analysis confirmed that echogenicity heterogeneity and age at menopause are important predictors, enhancing the model’s interpretability. This tool aids in early identification of high-risk patients and provides a reference for clinical decision-making.