Improved Differentiation of Low-Grade and High-Grade Gliomas and Detection of Tumor Proliferation Using APT Contrast...

The purpose of the study is to demonstrate the value of quantitative amide proton transfer (APT) imaging for differentiating glioma grades and detecting tumor proliferation.

This study included 32 subjects with 16 low-grade gliomas (LGG) and 16 high-grade gliomas (HGG) confirmed by histopathology. Chemical exchange saturation transfer (CEST) magnetic resonance imaging with APT weighting was performed on a 3 T scanner. After B₀ correction, Z-spectra were fitted with Lorentzian functions corresponding to the upfield semi-solid magnetization transfer and nuclear overhauser enhancement (MT&NOE) effect, the direct saturation (DS) effect, and the downfield APT effect centered at around - 1.5, 0, and + 3.5 ppm, respectively. To compute the Z-spectral fitted APT (fitted_APT) in solid tumor tissue, double-peak histogram fitting of pixel MT&NOE effect from the whole tumor was used to remove necrosis regions. The fitted APT was then compared with the conventional APT based on magnetization transfer ratio asymmetry. Receiver operating characteristic (ROC) analysis was used to compare the performance between Z-spectral fitted contrasts and the con_APT for LGG versus HGG differentiation. Additionally, the correlations between the imaging contrasts (fitted_APT, con_APT, and fitted_MT&NOE) and Ki-67 labeling index for tumor proliferation were also evaluated.

Z-spectral fitted_APT shows improved statistical power for differentiating HGG and LGG (7.58 ± 0.99 vs. 6.79 ± 1.05 %, p < 0.05) than con_APT (4.34 ± 0.95 vs. 4.05 ± 2.02 %, p > 0.05) in solid tumor tissues. Analyses of whole tumor, on the other hand, have less differentiating power for both fitted_APT (p from 0.032 to 0.08) and con_APT (p from 0.696 to 0.809). Similarly, based on ROC analyses, fitted_APT shows larger area under the curve (AUC = 0.723) than con_APT (AUC = 0.543). The combination of fitted APT, DS, and MT&NOE further improved the specificity (75 %), diagnostic accuracy (78.2 %), and area under the curve (0.758) in differentiating LGG and HGG. Consistently, fitted_APT (r = 0.451, p = 0.018) is better correlated with Ki-67 than con_APT (r = 0.331, p = 0.092).

Fitted APT from Z-spectrum improves differentiation of low- and high-grade gliomas and better correlated with tumor proliferation than conventional APT.