Value of the diagnostic algorithm in differentiating glioblastoma from solitary brain metastasis and primary central nervous system lymphomas using Dynamic susceptibility contrast-enhanced MRI (DSCE-MRI) and Proton magnetic resonance spectroscopy (1H-MRS).

Purpose: Our study aims to differentiate and develop the diagnostic algorithm for glioblastoma (GBM), solitary brain metastasis (SM), and primary central nervous system lymphomas (PCNSLs) using Dynamic susceptibility contrast-enhanced MRI (DSCE-MRI) and Proton magnetic resonance spectroscopy (1H-MRS).

Material and methods: This retrospective study included 91 patients (51 GBM, 18 SM, and 22 PCNSLs) who underwent preoperative imaging with a standard 3T MRI brain tumor protocol, including conventional Magnetic Resonance Imaging (cMRI), DSCE-MRI, and 1H-MRS. All patients underwent surgery or stereotactic biopsy with histopathological confirmation. On DSCE-MRI, the ratios of tumor regions (t) and peritumoral regions (e) to normal white matter (n) in CBV and CBF maps were analyzed, including rCBVt, rCBFt, rCBVt/n, rCBFt/n; rCBVe, rCBFe, rCBVe/n, and rCBFe/n. On 1H-MRS, metabolite ratios of the tumor and peritumoral regions were evaluated, comprising tCho/NAA, tCho/Cr, pCho/NAA, and pCho/Cr. Conducted the statistical analysis using the Fisher test or Chi-square test, One-way ANOVA tests, and decision tree analysis.

Results: The differences in indices between the two imaging modalities were statistically significant in differentiating tumor types: (1) GBM vs. SM: the values of rCBVe, rCBVe/n, rCBFt, rCBFe, rCBFt/n, tCho/Cr, tCho/NAA, eCho/Cr, and eCho/NAA were significantly higher in GBM compared to SM (p < 0.05). (2) GBM vs. PCNSLs: the values of rCBVt, rCBVe, rCBVt/n, rCBVe/n, rCBFt, rCBFe, rCBFt/n, rCBFe/n, eCho/Cr, and eCho/NAA were significantly higher in GBM compared to PCNSLs (p < 0.001). (3) SM vs. PCNSLs: The values of rCBVt, rCBVt/n, rCBFt, and rCBFe/n of SM were significantly higher, while tCho/NAA of SM were lower compared to PCNSLs (p < 0.001). The diagnostic algorithm using rCBVe/n, rCBFt/n, rCBVt, rCBVe, and tCho/NAA achieved 100% accuracy in diagnosing GBM and PCNSLs, and 94.7% for SM, with a misclassification risk estimate of 2.1%, the sensitivity of 100%, specificity of 98.9%, and an AUC of 0.993.

Conclusion: The values obtained from DSCE, 1H-MRS, and the diagnostic model play a crucial role in differentiating GBM, SM, and solitary PCNSLs. Our DSCE and 1H-MRS-based algorithm accurately differentiates GBM, SM, and PCNSLs, achieving 100% sensitivity and 98.9% specificity. This non-invasive method enhances pre-treatment diagnosis and prognosis, improving diagnostic quality for patients.