Prediction of Lymphovascular Space Invision in Endometrial Cancer based on Multi-parameter MRI Radiomics Model

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Endometrial carcinoma (EC) staging requires accurate assessment of lymphatic vascular space invasion (LVSI).
• Current methods for predicting LVSI preoperatively have limitations.
• Multi-parametric MRI radiomics offers potential for non-invasive prediction.

Purpose of the Study:

• To evaluate the application value of a combined model integrating multi-parameter MRI radiomics and clinical features.
• To predict preoperative lymphatic vascular space invasion (LVSI) in endometrial carcinoma (EC).

Main Methods:

• Retrospective analysis of 218 EC patients' clinicopathological and imaging data.
• Feature extraction from multi-parametric MRI sequences (ADC, CE-sag, CE-tra, DWI, T2WI-sag-fs, T2WI-tra).
• Radiomics and clinical feature selection using MRMR, LASSO, and logistic regression; construction of a predictive nomogram.

Main Results:

• A combined radiomics model achieved an AUC of 0.884.
• The nomogram integrating radiomics, age, and maximum tumor diameter (MTD) showed high predictive accuracy (AUC 0.914 training, 0.912 validation).
• The nomogram demonstrated good calibration and clinical utility for predicting LVSI.

Conclusions:

• The combined model effectively predicts LVSI in EC patients.
• Multi-parametric MRI radiomics integrated with clinical data provides valuable preoperative risk stratification for EC.
• This approach enhances diagnostic efficiency and clinical decision-making for EC management.