Radiomics analysis of multiparametric MRI for preoperative prediction of microsatellite instability status in...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Endometrial cancer (EC) treatment can be guided by microsatellite instability (MSI) status.
• Accurate pre-treatment prediction of MSI status is crucial for personalized therapy.

Purpose of the Study:

• To develop and validate a multiparametric MRI-based radiomics model for predicting MSI status in EC patients.
• To assess the model's performance and clinical utility.

Main Methods:

• Radiomics features were extracted from T2WI, CE-T1WI, and ADC maps of 225 EC patients (training/internal validation) and 132 (external validation).
• Machine learning models (SVM, LR, KNN, NB, RF) were trained to predict MSI status.
• Model performance was evaluated using ROC and DCA.

Main Results:

• The Support Vector Machine (SVM) model achieved an AUC of 0.905 in the training cohort.
• The model demonstrated robust performance in internal (AUC 0.875) and external (AUC 0.862) validation cohorts.
• Decision curve analysis indicated favorable clinical utility.

Conclusions:

• A multiparametric MRI radiomics model effectively predicts MSI status in EC.
• This model has the potential to assist in clinical treatment decision-making for EC patients.