Amide proton transfer (APT) MRI probes amide protons from endogenous proteins/peptides, which has shown promising results in defining acidosis, a surrogate metabolic biomarker during acute stroke^1,2. Indeed, pH complements the widely-used perfusion and diffusion images for enhanced stratification of heterogeneous ischemic tissue injury^3,4. However, the endogenous APT effect depends not only on pH but also on tissue water content, relaxation rates, and experimental conditions. Particularly, there are concomitant RF irradiation effects including direct RF saturation, magnetization transfer and nuclear overhauser effects (NOE) ^5-7. Hence, conventional in vivo APT analysis provides pH-weighted contrast, limiting its specificity to tissue acidification. Our study evaluated Magnetization transfer and water Relaxation-normalized APT (MRAPT) MRI in an animal model of acute stroke for semiautomatic lesion segmentation.Ten normal and ten stroke rats were imaged (intraluminal MCAO model) at 4.7 Tesla. MRI (5 slices, slice thickness/gap=1.8/0.2 mm, FOV=20x20 mm², image matrix=48x48) was acquired with EPI. Briefly, diffusion MRI was obtained using isotropic DWI (b= 250 and 1000 s/mm², TR/TE=3250/54ms). For pH-weighted APT MRI, we set the recovery time to 5000 ms (TS1/TS2=4500/500 ms) for an RF irradiation amplitude of 0.75 μT applied at ±3.5ppm and perfusion was quantified with ASL MRI (TR/TE=6500/15 ms, NSA=32, TS=3250 ms, B1=4.7 μT). T1-weighted images were acquired using inversion recovery EPI, with inversion delays ranging from 250 to 3,000 ms (TR/TE=6500/15 ms, NSA=4); T2-weigthed images were obtained with TE of 30 and 100 ms (TR=3250 ms, NSA=16). Fig. 1 shows the association between R_1w, R_2w and mean MTR (MMTR) with pH-weighted MTR_asym, per pixel. There was significant correlation between R_1w*MTR_asym and R_1w (Fig. 1a), R2w (Fig. 1b) and MMTR (Fig. 1c). We also evaluated multivariate regression to enhance the prediction of MTR_asym. Notably, R_2w was no longer a significant predicator. We had R_1w*MTR_asym=-12.7% +25.5%*R1w +51.6% *MMTR -112.1%*MMTR*R1w (Fig. 1d, R2=0.83, P<0.001). When analyzed for all normal animals, R2 was found to be 0.80 ± 0.07. Fig. 2 demonstrates minimization of intrinsic heterogeneity in APT MRI with MRAPT algorithm. Fig. 2a shows conventional R_1w*MTR_asym while Fig. 2b shows R_1w*MTR_asym estimated from MMTR and R_1w. Fig. 2c shows the difference between experimentally measured and predicted R_1w*MTR_asym map (ΔMRAPTR), which displays little heterogeneity between brain WM and GM. Indeed, we calculated the CNR between striatum and cortex and found CNR for the proposed ΔMRAPTR MRI was 0.43±0.51, substantially smaller than that of commonly used R_1w*MTR_asym map, being 3.07±0.71 (P<0.01, paired t-test). Fig. 3 evaluates the proposed MRAPT MRI in acute ischemic stroke rats. Figs. 3a and 3b show T_1w and T_2w maps, respectively. CBF map (Fig. 3c) and diffusion MRI (Fig. 3d) shows noticeable PWI/DWI mismatch. Ischemic lesions were determined with a threshold-based algorithm. Whereas the threshold-based algorithm could not faithfully detect the ischemic lesion in MTR_asym map due to the confounding intrinsic heterogeneity (Fig. 3e), the ΔMRAPTR map showed improved lesion segmentation (Fig. 3f). Importantly, the lesion size decreased in an order from PWI (297 mm3), pH or ΔMRAPTR (205 mm3) to ADC lesions (157 mm³). Moreover, MRAPT shows significant different acidosis among PWI/pH mismatch, pH/DWI mismatch and ADC lesions, indicating heterogeneous acidification within the ischemic lesion.Multivariate regression analysis enables substantial reduction of intrinsic non-pH heterogeneity in pH-weighted APT MRI. The MRAPT approach enables semiautomatic lesion segmentation, demonstrating graded tissue acidification in the acute stroke setting for refined tissue classification.