Diagnostic Accuracy of Arterial Spin Labeling in Comparison With Dynamic Susceptibility Contrast-Enhanced Perfusion for Brain Tumor Surveillance at 3T MRI.

Purpose: We aimed to compare arterial spin labeling (ASL) with dynamic susceptibility contrast (DSC) enhanced perfusion MRI for the surveillance of primary and metastatic brain tumors at 3T, both in terms of lesion perfusion metrics and diagnostic accuracy. Methods: In this retrospective study, we included 115 patients, who underwent both ASL and DSC perfusion in the same 3T MRI scanning session between 1 January and 31 December 2019. ASL-derived cerebral blood flow (CBF) maps and DSC-derived relative cerebral blood volume (rCBV) maps, both uncorrected and corrected for leakage, were created with commercially available software. Lesions were identified as T2-/T2-FLAIR hyperintensity with or without contrast enhancement. Measurements were done by placing a region of interest in the visually determined area of highest perfusion, copying to the contralateral normal appearing white matter (NAWM), and then propagating to the other perfusion maps. Pearson's correlation coefficients were calculated between the CBF and rCBV ratios of tumor versus NAWM. Accuracy for diagnosing tumor progression was calculated as the area under the receiver operating characteristics (ROC) curve (AUC) for the ASL-CBF and leakage corrected DSC-rCBV ratios. Results: We identified 178 lesions, 119 with and 59 without contrast enhancement. Correlation coefficients between ASL-derived CBF versus DSC-derived rCBV ratios were 0.60-0.67 without and 0.72-0.78 with leakage correction in all lesions (n = 178); these were 0.65-0.80 in enhancing glioma (n = 80), 0.58-0.73 in non-enhancing glioma, and 0.14-0.40 in enhancing metastasis (n = 31). No significant correlation was found in enhancing (n = 8) or non-enhancing (n = 7) lymphomas. The areas under the ROC curves (AUCs) for all patients were similar for ASL and DSC (0.73-0.78), and were higher for enhancing glioma (AUC = 0.78-0.80) than for non-enhancing glioma (AUC = 0.56-0.62). In brain metastasis, the AUC was lower for ASL-derived CBF (AUC = 0.72) than for DSC-derived rCBV ratios (AUC = 0.87-0.93). Conclusion: We found that ASL and DSC have more or less the same diagnostic accuracy. Our findings suggest that ASL can be used as an alternative to DSC to measure perfusion in enhancing and non-enhancing gliomas and brain metastasis at 3T. For lymphoma, this should be further investigated in a larger population.