CEST imaging provides the benefit of enabling detection of low concentration solute molecules through amplifying their signal onto water. However, discriminating between agents with small chemical shift differences using existing methodology which relies on saturation frequency profiles is still quite difficult. CEST contrast is very sensitive to saturation power (B₁), length (t_sat) and frequency offset (∆ω), which provides a potential second route to discriminate between agents based on modulating the saturation conditions. Herein we utilized this dependence to identify CEST agents according to their different response to B1 due to variances in exchange rate (K_ex) and ∆ω, and termed this the Multi-echo Parametric VARiation Saturation (MePaVARS).MePOVARS Sequence: Similar to our MeLOVARS method^1,2, the MePaVARS sequence (Fig.1a) places gradient-echo image readouts in between N saturation (sat.) pulses, which may have a different B₁. As shown, the sat. pulses are the same length but alternate between a higher B₁ (B₁_high) and a lower one (B₁_low). Simulations and phantom experiments: Conventional MTR_asym spectra were acquired for Glutamate (Glu) and Creatine (Cr) solutions with pH = 7.4, with B₀ = 11.7T, B₁ = 3.6uT and T_sat = 3.5s. Their exchange rates (k_ex) were then fitted using a 2-pool Bloch-equation model. The MePaVARS data with 8 module readouts were then simulated. In vivo mice imaging: PC3 human prostate cancer cells were subcutaneously transplanted in nude mice at the lower flank near the right thigh. MRI was performed in an 11.7T Bruker horizontal scanner, with a 72-mm volume coil as the transmitter and a phase-array surface coil as the receiver. Mice were anesthetized by isoflurane with breath rate monitored during MRI. MePaVARS sequence acquired 6 modules, each including a saturation pulse of 0.5 sec. in length followed by a 4-shots EPI readout. Other parameters: B₁^low = 1.8 uT (Module 1, 3, 5) and B₁^high = 2.8 uT (Module 2, 4, 6); TR/TE = 3500ms/5ms, EPI module time = 7.2 ms, Flip Angle =25^o, Matrix size = 96X64, FOV = 19.5mm X 16mm and slice thickness = 10mm.We first performed Bloch-equation simulations and the phantom studies for two endogenous Metabolites, Glu and Cr. Their conventional CEST spectra (Fig.1b) shows a broader spectrum for Glu due to the faster exchange (k_ex was fitted as ~6000/s) while a sharper peak for Cr with a slower exchange (k_ex = ~360/s). MePaVARS sequence enable fast acquisition of 8 modules, each weighted with different sat. parameters (B₁ and T_sat). Although in Fig.1b Glu and Cr shows very similar MTR_asym values ~2.2ppm, MePaVARS produces an additional signal patterns given the 8-module MTR_asym values at this frequency offset. The fast-exchanging Glu produced a more-oscillated pattern than the slow-exchanging Cr, as its saturation efficiency responds very different to B₁^high and B₁^low. Next, when tested MePaVARS in vivo on a mouse model of prostate tumor, we were able to acquire 6 modules of readout, with CEST MTR_asym spectra of Module 2 & 4 shown in Fig.2a. As seen, module 2 spectrum displays significant difference from module 4 spectrum for tumor region, i.e. signal at ~ 3.5 ppm and ~2.5 ppm increases from module 2 to module 4 while ~1 ppm it descrease. In contrast, control muscle region did not display such an obvious change between the two MTR_asym curves. We plotted the MTR_asym (~2.4ppm) as a function of module number as in Fig.2b, where tumor region displaying a similar high-frequency oscillation pattern as Glu, indicating the fast-exchanging amine protons. This could also easily figured out from the two CEST maps, where averaging module 2&4 displays a much higher MTR_asym in tumor than the averaged map of 3&5 although with longer saturation. We further did a voxel-by-voxel FFT for the signal series as module#, with the magnitude map of 3 oscillations highlighting only a portion of tumor (Fig.2c right). Using simulations and phantom studies of endogenous metabolites, e.g. Glu and Cr., we proved that MePaVARS allow separation of slower exchange species from the faster ones. A preliminary study for mice bearing prostate tumor further validate the feasibility of MePaVARS in vivo and also indicate that parts of tumor contains molecules with faster exchange