Inversion Efficiency Model Yields Improved Accuracy in MP2RAGE‐Based <i>T</i> ₁ Mapping in the Human Brain at 7.0T

Estimation of the longitudinal relaxation time T₁ from the MP2RAGE pulse sequence is based on a monoexponential signal evolution model. However, magnetization transfer (MT) caused by the inversion pulse induces a fast relaxation component, which appears as a reduction in the efficiency of the inversion. This may explain the underestimation of T₁ derived from MP2RAGE. To address this systematic bias, an "apparent" inversion efficiency (f_inv) was introduced, which comprises all mechanisms that affect the inversion in the monoexponential MP2RAGE signal model. The model was then extended by calibrating an empirical linear dependence of f_inv on R₁ = 1/T₁, resulting in increased accuracy of the estimated T₁. The apparent inversion efficiency f_inv and the apparent T₁* (yielding T₁ by auxiliary B₁ ⁺ mapping) were mapped at 7T in healthy adults using phase-sensitive inversion recovery (IR) with four consecutive RAGE trains (PS-MP4RAGE) in conjunction with adiabatic inversion using time-resampled (TR)-FOCI and hyperbolic secant pulses. Upon validation by conventional IR-EPI, PS-MP4RAGE was used to calibrate the linear f_inv model for the human brain. These 3D T₁ maps also served as a reference to assess the improvement of the MP2RAGE-based T₁ estimates. The apparent inversion efficiency f_inv was consistently smaller in white matter (WM) than in gray matter (~0.73 vs. ~0.84). The difference in WM T₁ between MP2RAGE and the reference PS-MP4RAGE technique was reduced by more than 200 ms when using the suggested f_inv model. MT effects after spin inversion in MP2RAGE can be accounted for by calibrating the apparent inversion efficiency f_inv without introducing additional parameters. The proposed empirical model retains the B₁ ⁺ compensation inherent to MP2RAGE and facilitates accurate T₁ quantification in brain tissue.