CEST imaging is an important molecular MRI technique that allows detection of endogenous, low-concentration biomolecules in tissue. However, CEST quantifications depend on the choice of CEST metric approaches and reference images; thus, care should be taken when doing the quantification of CEST imaging and comparing the measurements in different labs. Herein, we evaluated the reliability of four CEST imaging metrics and potential confounds in tumors at different experimental settings.

A five-pool proton exchange model (free water, semi-solid, amide, amine, and NOE-related protons) combined with the super-Lorentzian lineshape for semi-solid protons was used for the simulation. Four CEST metrics (CEST ratio or CESTR [1-5], CESTR normalized with the reference value or CESTR^nr [6-9], inverse Z-spectrum-based or MTR_Rex [10], and apparent exchange-related relaxation or AREX [11-13]) were compared. All reference signals (Z_ref) were taken from the simulated semi-solid MT signal. For the in vivo study, eight glioma-bearing rats were scanned at 4.7 T.

$$(1) CESTR=\frac{S_{ref}-S_{lab}}{S_{0}}=Z_{ref}-Z_{lab}$$

$$(2) CESTR^{nr}=\frac{S_{ref}-S_{lab}}{S_{ref}}=\frac{Z_{ref}-Z_{lab}}{Z_{ref}}$$

$$(3) MTR_{Rex}=\frac{(S_{ref}-S_{lab})S_{0}}{S_{ref}S_{lab}}=\frac{1}{Z_{lab}}-\frac{1} {Z_{ref}}=\frac{Z_{ref}-Z_{lab}}{Z_{ref}Z_{lab}}$$

$$(4) AREX=\frac{MTR_{Rex}}{T_{1w}}=\frac{(S_{ref}-S_{lab})S_{0}}{S_{ref}S_{lab}T_{1w}}=\frac{Z_{ref}-Z_{lab}}{Z_{ref}Z_{lab}}\cdot\frac{1}{T_{1w}}$$

The five-pool Bloch equation-based simulation results with six RF saturation power levels of 0.5, 1, 1.5, 2, 2.5, and 3 μT are shown in Fig. 1 (for 4.7 T) and 2 (for 9.4 T). (i) Similar CEST signal features at 3.5 ppm and 2 ppm can be seen with all four CEST metrics when a relatively low RF saturation power (< 1 μT) is applied. (ii) The CEST signals using two inverse metrics (MTR_Rex and AREX) are dramatically increased (e.g. MTR_Rex ≈ 43 % and AREX ≈ 30 % at 3.5 ppm, and MTR_Rex ≈ 132 % and AREX ≈ 95 % at 2 ppm with B₁ of 3 μT), while CESTR and CESTR^nr metrics stay the same or similar as the RF power increase up to 3 μT. Such a tremendous increase of the inverse metrics occurs due to the inherent error from small denominators, likely resulting in their applications at high RF saturation powers and low clinical B₀ field strengths (3 T and 4.7 T) problematic. (iii) The inverse metric signals around the water frequency are not reliable due to denominators approaching zero. These peaks definitely lead to an erroneous result, in particular, for the detection of OH groups (hydroxyl) at 1 ppm and NH₂ groups (amine) at 2 ppm, close to the water resonance. (iv) At a high field of 9.4 T, the APT peaks can be relatively well resolved in AREX metrics, as shown in Fig. 2, because of the high spectral resolution compared to 4.7 T, while the AmineCEST peaks is still unreliable at high RF saturation power (> 2 μT).

For the in vivo tumor rat study, quantitative APT signals obtained from these four CEST metrics were assessed at varied saturation power levels (0.5, 0.9, 1.3, 2.1, 3.2, and 4.4 μT) as shown in Fig. 3. The high MTR_Rex and the AREX peaks can be seen clearly around 1 ppm due to the intrinsic error of the inverse metric, very similar to the result from the Bloch equation-based simulation. As expected, the APT signals of the tumor from CESTR^nr and CESTR were significantly higher than those of the normal tissue across all power levels (p < 0.05), while the upfield NOE signals of the tumor from CESTR and CESTR^nr were significantly lower than those of the normal tissue across all power levels (p < 0.01) as shown in Fig. 4. Multiple quantitative MRI maps of a tumor-bearing rat are shown in Fig. 5. Remarkably, the tumor was hyperintense on the MTR_asym, CESTR, CESTR^nr, and MTR_Rex maps at 3.5 ppm when an RF saturation power of 1.3 μT was applied. On the CESTR, CESTR^nr, and MTR_Rex maps at 2 ppm, the tumor was slightly hyperintense. As reported previously [11-13], the AREX maps at 3.5 and 2 ppm showed no contrast between the tumor and normal tissue.

In this study, we evaluated the reliability of two inverse CEST metrics and compared them with other CEST metrics, using a five-pool Bloch equation-based simulation and the rat tumor experiment at 4.7 T. The choice of CEST metrics must be carefully considered according to RF saturation power levels, B₀ field strengths, and specific exchangeable solute protons. MTR_Rex and AREX may not be used for amine and hydroxyl CEST measurements, particularly, with a relatively higher saturation power level (with the large direct water saturation and semi-solid MT effects). At the clinical field strength (3 T and 4.7 T), CESTR and CESTR^nr would be more reliable and valid for APT imaging at the optimal saturation power of 2 μT used currently.