Differential Diagnostic Value of Texture Feature Analysis of Magnetic Resonance T2 Weighted Imaging between Glioblastoma and Primary Central Neural System Lymphoma

Abstract

Objective: To investigate the difference in tumor conventional imaging findings and texture features on T2 weighted images between glioblastoma and primary central neural system (CNS) lymphoma. Methods: The pre-operative MRI data of 81 patients with glioblastoma and 28 patients with primary CNS lymphoma admitted to the Chinese PLA General Hospital and Hainan Hospital of Chinese PLA General Hospital were retrospectively collected. All patients underwent plain MR imaging and enhanced T1 weighted imaging to visualize imaging features of lesions. Texture analysis of T2 weighted imaging (T2WI) was performed by use of GLCM texture plugin of Image J software, and the texture parameters including Angular Second Moment (ASM), Contrast, Correlation, Inverse Difference Moment (IDM), and Entropy were measured. Independent sample t-test and Mann-Whitney U test were performed for the between-group comparisons, regression model was established by Binary Logistic regression analysis, and receiver operating characteristic (ROC) curve was plotted to compare the diagnostic efficacy. Results: The conventional imaging features including cystic and necrosis changes (P = 0.000), ‘Rosette’ changes (P = 0.000) and ‘incision sign’ (P = 0.000), except ‘flame-like edema’ (P = 0.635), presented significantly statistical difference between glioblastoma and primary CNS lymphoma. The texture features, ASM, Contrast, Correlation, IDM and Entropy, showed significant differences between glioblastoma and primary CNS lympoma (P = 0.006, 0.000, 0.002, 0.000, and 0.015 respectively). The area under the ROC curve was 0.671, 0.752, 0.695, 0.720 and 0.646 respectively, and the area under the ROC curve was 0.917 for the combined texture variables (Contrast, cystic and necrosis, ‘Rosette’ changes, and ‘incision sign’) in the model of Logistic regression. Binary Logistic regression analysis demonstrated that cystic and necrosis changes, ‘Rosette’ changes and ‘incision sign’ and texture Contrast could be considered as the specific texture variables for the differential diagnosis of glioblastoma and primary CNS lymphoma. Conclusion: The texture features of T2WI and conventional imaging findings may be used to distinguish glioblastoma from primary CNS lymphoma.