Amide proton transfer (APT) imaging has been developed as one of the endogenous chemical exchange saturation transfer (CEST) imaging techniques introduced by the group of Zhou et al.¹. APT imaging exploits the exchange of protons between the amide protons (-NH) in endogenous mobile proteins and bulk-water protons. Recent studies have shown that APT imaging is useful in grading gliomas², assessment of therapeutic response in GBMs³, and in differentiation of radiation necrosis and active/recurrent tumors⁴. APT imaging is highly sensitive to the presence of B0 inhomogeneity and thus accurate correction methods are necessary. The separate B0 mapping with dual-echo gradient sequence has been shown to work for this purpose. It was reported that 2D separate B0 mapping method was accurate and reproducible in APT imaging of brain tumors⁵. However, the separate B0 mapping is time consuming, and susceptible to alteration in B0 distribution during the scans due to patient’s head movement. Recently, the FSE Dixon APT acquisition protocol with intrinsic B0 correction was developed and implemented on 3T clinical MRI scanners^6-7. This technique allows simultaneous acquisition of APT imaging and intrinsic B0 mapping without increasing scan time. The purpose of the present study was to assess the quantitative performance of the 3D FSE Dixon APT imaging of brain tumors in comparison with the separate B0 mapping.

[Patients] Twenty-two patients with brain tumors (54.2±18.7 year-old, 12 males and 10 females) who underwent subsequent surgical resection were included in the study. Histological types of brain tumors were as follows: 4 low-grade gliomas, 8 high-grade gliomas, 5 metastases, 2 meningiomas, and 2 malignant lymphomas and one central neurocytoma.

[MRI] MR imaging was performed on a 3T clinical scanner (Achieva 3.0TX, Philips Healthcare). APT imaging with intrinsic B0 correction by Dixon method was performed using 2-channel body coil transmission, 8-channel head coil reception and the following parameters in a 3D FSE sequence: RF saturation with Tsat=2s, B1,rms=2.0μT, 40 sinc-Gaussian pulses (50ms), frequency offsets=±3.1ppm, ±3.5ppm, ±3.9ppm and -1560ppm (S0). Two additional scans at +3.5ppm (averaged for signal intensity), 9 slices, FOV=212×184×40mm, voxel size=1.8×1.8×4.4mm, TR/TE=7200ms/6.2ms, total acquisition time 4min30s. Acquisition windows and readout gradients were shifted (echo-shift ES) by -0.4ms(ES1), 0ms(ES2), and +0.4ms(ES3) at 3.5ppm for Dixon B0 mapping. A separate 3D B0 mapping was performed with 3D GRE sequence (δTE=1ms) for comparison. A single-slice 2D APT imaging was performed in the following parameters: RF saturation with Tsat=2s, B1,rms=2.0μT, 40 sinc-Gaussian pulses (50ms), 25 frequency offsets (-5.0...+6ppm, step 0.5ppm) and -1560ppm (S0), FOV=230×230mm, voxel size=1.8×1.8×5mm, TR/TE=5000ms/6ms, total acquisition time 2min20s. A separate 2D B0 mapping was performed with 2D GRE sequence (δTE=1ms).

[Image Analysis] The MTR asymmetry (MTR_asym, APT signal)=(S_[−3.5ppm]−S_[+3.5ppm])/S₀ was calculated with a point-by-point B0 correction. Three types of APT images were generated; (1) 3D APT reconstructed via Dixon-B0 mapping (3D-Dixon), (2) 3D APT reconstructed via separate B0 mapping (3D-Sep), (3) 2D APT reconstructed via separate B0 mapping (2D-Sep, as a reference standard). Mean and 25-, 50-, 75- and 90-percentile of APT signal (MTRasym at 3.5 ppm) were measured in the whole tumor ROI in the slice with maximum area of each tumor.

[Statistical Analysis] Intraclass correlation coefficient (ICC), Pearson correlation, and Bland-Altman plot were performed to compare the methods.

Table 1 and Figure 1 show ICCs and correlations between each 3D method and 2D-Sep for mean and percentile values of tumors, respectively. The mean and 90-percentile APT signals obtained by 3D-Dixon showed excellent agreements (ICC=0.964 and 0.972, respectively) and correlations (r=0.93 and 0.95, respectively) with those obtained by 2D-Sep. These agreements and correlations were better than those obtained using the 3D-Sep (mean: ICC=0.811, r=0.70, 90-percentile: ICC=0.865, r=0.77). The Bland-Altman plot analyses (Figure 2) show the 3D-Dixon showed narrower 95% limits of agreement than 3D-Sep for both APT mean and 90-percentile. Figure 3 demonstrates a case with malignant lymphoma where 3D-Dixon APT imaging shows equivalent image quality and measurements to 2D-Sep APT imaging.

These results indicated that the intrinsic B0 map using Dixon method could be more accurate than separately acquired B0 map in 3D imaging. 3D Dixon method might be less sensitive to motion and alteration of B0 shimming during a scan. 3D separately acquired B0 maps might have larger frequency shifts due to difficulty in B0 shimming over imaging volume and patient's motion.

The 3D FSE Dixon APT imaging of brain tumors showed the high quantitative performance equivalent to single-slice 2D separate B0 mapping method. The intrinsic B0 mapping and correction would lead to easy work flow in clinical use, and volume coverage with 3D acquisition should be useful in therapy follow-up studies.