Related Experiment Video
Updated: Aug 6, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Multiparametric MRI Radiomics-Based Interpretable Machine Learning Model for the Prediction of Lymphovascular Space
Weijing Meng1, Hongxi Dou2, Jinfeng Yin2
1School of Public Health, Shandong Second Medical University, Weifang, China (W.M., F.S.).
Academic Radiology
|July 20, 2026
Summary
A machine learning model combining MRI radiomics and clinical data accurately predicts lymphovascular space invasion (LVSI) in endometrial cancer (EC). This approach enhances non-invasive preoperative prediction accuracy for better patient management.
Area of Science:
- Oncology
- Radiology
- Artificial Intelligence
Background:
- Endometrial cancer (EC) diagnosis requires accurate assessment of lymphovascular space invasion (LVSI).
- Current methods for LVSI detection can be invasive and may not always be accurate.
- Novel approaches are needed for non-invasive prediction of LVSI in EC.
Purpose of the Study:
- To develop and validate a machine learning (ML) model for predicting LVSI in EC.
- To integrate multiparametric Magnetic Resonance Imaging (MRI) radiomics with clinical indicators.
- To compare the performance of six different ML algorithms for this prediction task.
Main Methods:
- Retrospective analysis of 408 EC patients who underwent preoperative MRI across two centers.
- Extraction of clinical risk factors and intratumoral/peritumoral radiomic features from MRI.
- Development and validation of six ML models, with the NeuralNetwork model showing the highest performance (validation AUC=0.803).
Main Results:
- The optimal ML model combined clinical data (CA125, tumor diameter) with radiomic features.
- This integrated model achieved the highest predictive performance (training AUC=0.863, validation AUC=0.803).
- Model interpretability was enhanced using SHapley Additive exPlanations (SHAP), demonstrating good calibration and clinical utility.
Conclusions:
- A NeuralNetwork-based ML model integrating clinical and radiomic data effectively predicts LVSI in EC.
- This non-invasive approach shows potential for improving preoperative risk stratification in EC patients.
- The study highlights the value of radiomics and ML in enhancing diagnostic accuracy for EC.
