Dual contrast CEST MRI for pH-weighted imaging in stroke

Abstract

Purpose: pH enhanced (pH_enh ) CEST imaging combines the pH sensitivity from amide and guanidino signals, but the saturation parameters have not been optimized. We propose pH_dual as a variant of pH_enh that suppresses background signal variations, while enhancing pH sensitivity and potential for imaging ischemic brain injury of stroke.

Methods: Simulation and in vivo rodent stroke experiments of pH_enh MRI were performed with varied RF saturation powers for both amide and guanidino protons to optimize the contrast between lesion/normal tissues, while simultaneously minimizing signal variations across different types of normal tissues. In acute stroke, contrast and volume ratio measured by pH_dual imaging were compared with an amide-CEST approach, and perfusion and diffusion MRI.

Results: Simulation experiments indicated that amide and guanidino CEST signals exhibit unique sensitivities across different pH ranges, with pH_enh producing greater sensitivity over a broader pH regime. The pH_enh data of rodent stroke brain demonstrated that the lesion/normal tissue contrast was maximized for an RF saturation power pair of 0.5 μT at 2.0 ppm and 1.0 μT at 3.6 ppm, whereas an optimal contrast-to-variation ratio (CVR) was obtained with a 0.7 μT saturation at 2.0 ppm and 0.8 μT at 3.6 ppm. In acute stroke, CVR optimized pH_enh (i.e., pH_dual ) achieved a higher sensitivity than the three-point amide-CEST approach, and distinct patterns of lesion tissue compared to diffusion and perfusion MRI.

Conclusion: pH_dual MRI improves the sensitivity of pH-weighted imaging and will be a valuable tool for assessing tissue viability in stroke.

Keywords: CEST; amide; guanidino; pH; stroke.

Fig. 1.. Simulation of Amide CEST and Guanidino CEST signals exhibit differing pH dependence across varying saturation powers.

(A) Amide CEST signal increases monotonically with pH across powers ranging from 0.3 μT to 2.0 μT. (B) Guanidino CEST signal exhibits a maximum at particular pH values that increase as power increases from 0.3 μT to 2.0 μT. (C) Subtracting the Amide CEST signal at pH 7 across all pH shows stronger changes from pH 7 to 7.4 compared to below 7. (D) Guanidino CEST signal contrast with pH 7.0 can be either positive or negative, depending on both the pH and the saturation power.

Fig. 2.. Simulation of pH_enh contrast in four types of voxels in either lesion or normal tissue and gray or white matter across different saturation power parameter combinations.

(A) Lesion/Normal Contrast may be optimal close to 0.5 μT at 2.0 ppm and 1.0 μT at 3.6 ppm, however, normal gray matter exhibits a higher signal than white matter lesion tissue. (B) Using a power pair of 0.7 μT at 2.0 ppm and 0.8 μT at 3.6 ppm shows lower gray/white matter contrast improving differentiation of lesion and normal tissue. (C) Lesion/normal tissue contrast across different power combinations shows maximal contrast with 0.56 μT at 2.0 ppm and 0.95 μT at 3.6 ppm with contrast only slightly decreasing as powers diverge from this maximum. (D) The absolute value of gray/white matter contrast reaches a minimum at around 0.79 μT at 2.0 ppm and 0.86 μT at 3.6 ppm with this contrast staying relatively low around this ratio of powers.

Fig. 3.. Simulated Lesion/Normal Tissue Contrast, Variation with 0.1% Noise, and Contrast to Variation with either 0.1% or 0.5% noise in tissue with either 90% or 70% gray matter.

Simulation of (A) lesion/normal tissue contrast, (B) variation with 0.1% noise, (C) contrast to variation with 0.1% noise, and (D) contrast to variation with 0.5% noise assuming a tissue distribution of 90% gray matter and 10% white matter. Under an assumption of a tissue distribution of 70% gray matter and 30% white matter, simulation was performed of (E) lesion/normal tissue contrast, (F) variation with 0.1% noise, (G) contrast to variation with 0.1% noise, and (H) contrast to variation with 0.5% noise.

Fig. 4.. Lesion/Normal tissue contrast, variation, and contrast to variation in MCAO rodents as a function of RF saturation powers.

(A) Lesion/Normal tissue contrast exhibited a maximal contrast of 2.72 ± 0.33% with 0.5 μT at 2.0 ppm and 1.0 μT at 3.6 ppm. (B) Contralateral variation was minimized along a diagonal with the minimum variation of 0.32 ± 0.09% at 0.5 μT at 2.0 ppm and 0.6 μT at 3.6 ppm. (C) Total variation was minimized along a diagonal with the minimum variation of 0.70 ± 0.26% at 0.5 μT at 2.0 ppm and 0.6 μT at 3.6 ppm. (D) Contrast to variation was maximized along the same diagonal with the maximum of 3.80 ± 1.90 being observed with 0.7 μT at 2.0 ppm and 0.8 μT at 3.6 ppm.

Fig. 5.. pH_enh maps of two exemplary MCAO rodents using CEST power pairs that optimize contrast (0.5 μT/1.0 μT) or optimize CVR (0.7 μT/0.8 μT) for 2.0 ppm and 3.6 ppm.

ROIs depicting infarcted lesions (red) and the contralesional (blue) were shown on ADC maps. Broad contrast differences between the lesion and the contralesional tissue on contrast-optimized maps but lesion bounds were difficult to delineate. CVR-optimized maps showed clear margins of the lesion regardless of being gray or white matter.

Fig. 6.. pH_dual and APT* maps of two exemplary MCAO rodents and box and whisker plots comparing signals averaged over the lesion and contralesional ROIs.

Lesions were clearly identifiable on both pH_dual and APT* maps with greater contrast using pH_dual (2.53%, SD = 0.49%) over APT* (1.04%, SD = 0.10%) and higher CVR.

Fig. 7.. Comparison of ADC, CBF, and pH_dual lesions in MCAO rodents.

(A, B) In two examples, a clear difference in lesion size was detected among ADC, CBF, and pH_dual maps where pH_dual depicted an ischemic region larger than the ADC core yet smaller than the region of CBF deficit. The orange arrows indicated the mismatch between the ADC and pH_dual lesions. (C) both the mismatch between ADC and pH_dual and the ADC-CBF mismatch were strongly correlated with the volume fraction of the ADC core. (D). Group comparison showed that both CBF and pH_dual lesion volume was significantly larger than that of the ADC core (p<0.01). Moreover, the pH deficit volume detected by pH_dual was significantly larger than that detected by APT* (p<0.01).