Quantitative T2 mapping has been successfully applied in white matter to quantify myelin changes in multiple sclerosis [1], in cartilage to early track osteoarthritic changes [2] and in skeletal muscle to quantify edema in neuromuscular diseases [3]. T2 has been traditionally measured using multi-echo spin-echo (MESE) sequences with known problems related to the refocusing pulses in the presence of B1 and B0 inhomogeneities. T2-prepared sequences have been emerging to address the limitations of MESE, using either composite hard pulses with MLEV weighting [4] or adiabatic pulses [5] to reduce the sensitivity to B1 and B0. Specifically, the adiabatic BIR-4 RF pulse, previously used for generating T2-weighted contrast [6], has been recently adjusted with gaps (modified BIR-4) for T2 quantification [7]. Given that the adiabatic condition is fulfilled, BIR-4 is insensitive to B1 offsets [8]. However, with B0 offsets (ΔB0), the magnetization can experience both T1 and T2 relaxation leading to T2 quantification errors. The present work proposes a methodology to make T2 quantification based on the modified BIR-4 robust to the presence of large ΔB0.

Theory: The employed sequence contains a T2 preparation module (based on the modified BIR-4, where gaps are introduced for different T2-weighting), a spoiler gradient and a 3D TSE readout (Fig. 1a). In the presence of ΔB0, the magnetization vector is not completely rotated into the transverse plane after the 1st segment of the BIR-4. Due to BIR-4’s nature, this magnetization is mostly refocused on the longitudinal axis at the end of the entire preparation [8]. However, part of the magnetization experiences T1 instead of T2 relaxation during the gaps (Fig.1), which leads to an overestimation of T2 (T1 >> T2). To correct this error, we propose to use an additional saturation preparation scan executing only the 1st segment of the BIR-4 and a spoiler gradient before the readout (Fig. 1b). This measured magnetization is then used as input to the fitting model that was derived by following the magnetization pathways and including an effective T2 decay during the pulse [9].

$$ S(t) = M_0 \ast e^{-\frac{t}{T_2}} \ast e^{-g \frac{T_{\text{BIR4}}}{T_2}} \ast \left( 1 - \left( \frac{M_{\text{Sat}} \ast e^{g \frac{1}{4} \frac{T_{\text{BIR4}}}{T_2}}}{M_0}\right)^2 \right) + \frac{M^2_{\text{Sat}} \ast e^{-g \frac{1}{2} \frac{T_{\text{BIR4}}}{T_2}}}{M_0} $$

where M₀: denotes initial magnetization, T_BIR4: length of BIR-4, g: effective T2 weighting constant, M_Sat: measured magnetization after 1st segment of BIR-4 and t: gaps duration.

Simulations: The dependence on B1 and B0 error offsets was investigated with Bloch simulations, using: BIR-4 duration: 10 ms, B1: 13.5 µT, frequency sweep: 3.7 kHz, T2Prep module duration: 20/40/60/80 ms, T1/T2: 1400/70ms. T2 maps were extracted using the standard 2-parameter-fit and Eq. (1) incoorporating the SAT Prep measurement.

Phantom and in vivo measurements: 4 phantoms with different agar concentration were investigated in a 3 T system (Ingenia, Philips Healthcare) for a range of manually set resonance frequency offsets, with sequence paramters: voxel size: 1.75×1.75×4 mm³, FOV: 12×12×6 cm³, TSE-factor: 50, TR/TE = 1.5 s/20 ms, 6 dynamic scans (duration: 15/30/45/60/75/90 ms) and the SAT Prep scan. The shoulder of one healthy subject was also scanned with different T2Prep duration (20/30/40/50/60ms) and the SAT Prep scan, with FOV = 16×16×8 cm³, TR/TE = 1.5 s/20 ms, TSE factor = 50, acquisition voxel = 2×2×6 mm³. Additionally, a B0 map was acquired. All measurements were processed with standard fitting and the proposed method.

Figure 2 shows the great insensitivity regarding B1 errors when using the standard fitting due to the adiabatic nature of the BIR-4. However, the standard approach fails when larger ΔB0s are present, whereas the proposed method minimizes the error in T2 quantification (Fig. 2). The phantom results also show that the standard fitting tends to overestimate T2 with larger ΔB0 and the proposed method improves the T2 quantification for a range of ΔB0 up to 150 Hz. In vivo, a large B0 offset was observed in the subscapularis muscle (Fig. 4) due to its proximity to the lungs. In this area (red circle in Fig. 4) the standard fitting overestimates the T2 of muscle quite notably, whereas the proposed fitting with saturation preparation gives homogeneous T2 maps.A novel method was proposed for minimizing the sensitivity of T2 preparation using the modified BIR-4 to large ΔB0. It was shown that with only one additional scan the stability of the fitted T2 value can be greatly improved. The proposed method would be useful for T2 mapping applications in areas with large inhomogeneity (either B1 or B0) as for instance the shoulder/neck area or near metal implants and at high field strengths.