Purpose: To determine optimal conditions for precise measurement of arterial input function (AIFs) in dynamic susceptibility contrast (DSC) perfusion MRI.
Materials and methods: Magnitude-based (DeltaR(2)*) and phase-based (Deltaphi) AIFs were numerically simulated for several doses and baseline MRI noise levels [SNR(I(0))]. Random noise (1000 realizations) was added to real/imaginary MRI signals (derived from an internal carotid AIF), and AIF signal, noise, and signal-to-noise ratio (SNR) were determined. The optimal dose was defined as the dose that maximizes mean AIF SNR over the first-pass (SNR(mean)), rather than SNR at the AIF peak (SNR(peak)) because, compared to SNR(peak), doses predicted by SNR(mean) reduced the AIF-induced variability in cerebral blood flow (CBF) by 24% to 40%.
Results: The AIF SNR is most influenced by choice of AIF signal, then optimal dosing, each with little penalty. Compared to DeltaR(2)*, Deltaphi signal has 4 to 80 times the SNR over all doses and time points, and approximately 10-fold SNR(mean) at respective optimal doses. Optimal doses induce 85% to 90% signal drop for the DeltaR(2)* method, and 70% to 75% for Deltaphi, with two-fold dose errors causing approximately 1.7-fold loss in SNR(mean). Increases in SNR(I(0)) proportionally increase AIF SNR, but at a cost.
Conclusion: AIF SNR is affected most by signal type, then dosing, and lastly, SNR(I(0)).