Various strategies have been employed to combat T1 or T2* leakage artifacts resulting from contrast agent extravasation in brain tumors, including adjusting the flip angle, use of a preload, and diverse post-processing leakage correction algorithms. The current study uses DSC simulations and experimental testing to optimize the image acquisition parameters and leakage correction algorithms to obtain rCBV measurements with the highest accuracy.
METHOD AND MATERIALSDSC data was simulated using developments from Quarles et al.(1). Flip angle, TE, TR, and preload dosages were varied to simulate various MRI protocols. Monte Carlo simulation of 100 tumors, with varying Ktrans, ve, and T10, was used to test each protocol. Uncorrected rCBV, unidirectional leakage correction, and bidirectional leakage correction(2) estimates were compared to a gold standard (tumor without contrast extravasation).
RESULTSLow flip angles, high TE, high TR, and preload dosage increase T2*-weighting. The smaller the deviation of the uncorrected deltaR2* curve from the gold standard, as measured by uncorrected rCBV error, the smaller the residual error was after leakage correction. Furthermore, rCBV estimates with bidirectional correction had lower percentage error than those with unidirectional correction both with and without preload by a factor of 2.0±0.2. Optimal protocols had flip angle 60 degrees with TE = 35 ms, TR ranging from 1.5-2 s, and preload dosage varying from 0 to ½, with the DSC-MRI injection using the remaining portion of a full dose.
CONCLUSIONResults suggest that the best performance occurs when the acquisition protocol yields approximately equal T1 and T2* contributions from contrast agent extravasation, rather than a DSC-MRI signal with dominant T1- or T2*-weighting. The bidirectional rCBV estimates consistently outperform the unidirectional and uncorrected rCBV estimates.
CLINICAL RELEVANCE/APPLICATIONClinical trials using DSC-MRI perfusion should focus on using protocols that minimize errors for the uncorrected rCBV to best recover “true” rCBV.1) Quarles et al., Phys Med Biol, 2009; 2) Leu et al., JMRI, 2016