Magnetic resonance imaging texture analysis for discrimination between malignant and benign lesions in the floor of the mouth.

Abstract

Objective: Owing to their overlapping radiological characteristics, distinguishing between malignant and benign lesions of the floor of the mouth (FOM) remains challenging. The aim of this study was to assess the effectiveness of magnetic resonance imaging (MRI) texture analysis in discriminating between malignant and benign lesions.

Methods: MRI texture analysis was performed in 30 patients with FOM lesions (eight inflammatory diseases, ten cystic lesions, three benign tumors, and nine malignant tumors). Based on the histopathological examinations, the lesions were classified as malignant or benign. The lesion type was considered the predictor variable and MRI texture features were the outcome variables. Short-tau inversion recovery images were used for texture feature extraction, with 10 features selected using Fisher's coefficient. One-way analysis of variance and Tukey's post-hoc test were used to compare the texture features. Receiver operating characteristic (ROC) curves were used to assess the ability of MRI texture features to distinguish between lesions. The significance level was set at P < .05.

Results: ROC curve analysis showed that cystic lesions and malignant lesions could be distinguished with very high accuracy; malignant and benign lesions were differentiated with high accuracy. The accuracy in distinguishing between inflammatory diseases and malignant lesions was relatively low.

Conclusions: MRI texture features were useful in distinguishing malignant from some benign lesions within the FOM, although were less effective in distinguishing inflammatory from malignant lesions.