Radiomics analysis of parotid gland tumors using DWI and DCE-MRI: comparison of multiple machine learning models for differential diagnosis

Purpose: To develop and validate the value of different radiomics models based on diffusion-weighted imaging (DWI) and dynamic contrast-enhanced MRI (DCE-MRI) images for preoperative discrimination of parotid gland tumors (PGTs). Material and methods: = 48) cohorts. A total of 963 radiomics features were extracted from DWI and DCE-MRI images. After dimensionality reduction and feature selection, three radiomics models based on DWI, DCE-MRI, and DWI + DCE-MRI were constructed by several machine learning algorithms. The optimal radiomics model was selected using receiver operating characteristic (ROC) curve analysis. The performance of the models was evaluated using ROC curve and area under the curve (AUC) analysis, and decision curve analysis (DCA) was conducted to estimate the clinical values. Results: Logistic regression (LR) was selected as the optimal classifier for distinguishing benign and malignant PGTs. The radiomics model based on DWI + DCE-MRI achieved the highest AUC values in the training and test cohorts (AUC = 0.915 and 0.861), outperforming the DCE-only model (AUC = 0.903 and 0.847) and the DWI-only model (AUC = 0.839 and 0.769). The DCA based on the DWI + DCE-MRI radiomics model demonstrated high clinical usefulness. Conclusions: Radiomics is a useful tool for distinguishing malignant from benign PGTs. The radiomics predictive model that combines DWI and DCE-MRI yielded outstanding performance for improving clinical decision-making regarding the treatment of PGTs.