Development of MRI-based radiomics predictive model for classifying endometrial lesions

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Accurate diagnosis of endometrial lesions is crucial for appropriate patient management.
• Radiomics offers a quantitative approach to extract imaging features for improved diagnostic accuracy.

Purpose of the Study:

• To develop and validate a radiomics-based predictive model for classifying benign versus malignant endometrial lesions.
• To assess the clinical applicability and diagnostic performance of the developed model.

Main Methods:

• Utilized data from 139 patients with pathologically confirmed endometrial lesions across two independent centers.
• Extracted and selected radiomics features, then assessed associations with clinical parameters using logistic regression.
• Evaluated model performance using receiver operator characteristic (ROC) curves, DeLong validation, and decision curve analysis (DCA).

Main Results:

• A radiomics prediction model with 15 selected features achieved high discrimination in both training (AUC=0.90) and test (AUC=0.85) cohorts.
• The radiomics nomogram demonstrated superior discrimination compared to clinical parameters alone (AUC=0.91 training, 0.86 test).
• Decision curve analysis confirmed the clinical usefulness and net benefit of the radiomics nomogram.

Conclusions:

• The developed MRI radiomics and clinical parameter-based model shows high diagnostic efficiency for endometrial lesions.
• This predictive model holds potential for precise identification and prediction of endometrial lesions in clinical practice.
• Radiomics nomograms offer a valuable tool for gynecologists in differentiating benign and malignant endometrial pathologies.