Diagnostic Accuracy of Semiautomatic T2 Subtraction plus Quantitative Susceptibility Mapping in the Detection of New Multiple Sclerosis Lesions

The ability to identify new MRI lesions in patients with multiple sclerosis (MS) on follow-up imaging is of great importance in monitoring disease activity and informing therapeutic decision-making. Gadolinium (Gd)-enhancing lesions tend to be isointense or slightly hyperintense on QSM images whereas non-enhancing tend to be hyperintense. However, characterization of QSM signal of MS lesions in isolation can be difficult without coregistered T2-weighted imaging. For this reason, we developed an algorithm of T2-subtraction based on two time points of FLAIR images, as well as an automatic lesion mask to help detect new MS lesions with the overall goal of combining this technique with QSM to predict the enhancement status of MS lesions. We found that T2 subtraction+QSM has a sensitivity of 90.9% to predict new enhancing lesions that had been previously identified by experienced neuroradiologists on T1w+Gd imaging. In discriminating between new enhancing versus new but nonenhancing lesions, our T2 subtraction+QSM protocol had a sensitivity of 87.5%, and specificity of 89.7%. Receiver operating characteristic (ROC) curve analysis using region-of-interest of susceptibility values on QSM showed an optimal cutoff susceptibility value of -4.92 ppb (referenced to CSF) in distinguishing new enhancing lesions from new but nonenhancing lesions (sensitivity 88.9%, specificity 80.0%). Our results suggest that T2 subtraction plus QSM no Gd protocol can be a useful tool in detecting the new enhancing MS lesions in clinical practice without Gd injection.