Saturation Power Dependence of Amide Proton Transfer (APT) Image Contrasts in Human Brain Tumors and Strokes at 3T

Abstract

Amide proton transfer (APT) imaging is capable of detecting mobile cellular proteins and peptides in tumor and monitoring pH effects in stroke, through the saturation transfer between irradiated amide protons and water protons. In this work, four healthy subjects, eight brain tumor patients (four with high-grade glioma; one with lung cancer metastasis; three with meningioma), and four stroke patients (average 4.3 ± 2.5 days after the onset of the stroke) were scanned at 3T, using different radiofrequency saturation powers. The APT effect was quantified using the magnetization-transfer-ratio (MTR) asymmetry at 3.5 ppm with respect to the water resonance. At a saturation power of 2 μT, the measured APT-MRI signal of the normal brain tissue was almost zero, due to the contamination of the negative conventional MTR asymmetry. This irradiation power caused an optimal hyperintense APT-MRI signal in the tumor and an optimal hypointense signal in the stroke, compared to the normal brain tissue. The results suggest that the saturation power of 2 μT is ideal for APT imaging of these two pathologies at 3T with the existing clinical hardware.

Many cellular activities are performed by proteins. Thus, spatial information at the protein level, such as the concentration and the proton exchange properties of these protein molecules, is very useful for assessing various diseases, such as tumors and strokes (1,2). Recently, Balaban et al. proposed a chemical exchange saturation transfer (CEST) enhancement scheme (3–5), which can be used to detect low-concentration metabolites in solution indirectly via the water resonance (6–13). Based on the same principle, endogenous mobile proteins and peptides in tissue can be imaged by selectively labeling their water-exchangeable amide protons at 8.3 ppm (3.5 ppm downfield from the water resonance) through radiofrequency (RF) irradiation. This approach is called amide proton transfer (APT) imaging (14,15). APT imaging is a potentially important molecular MRI technique. It has been demonstrated that the endogenous protein and peptide-based MRI contrast can not only provide valuable information about the presence and grade of malignant brain tumors, based on the increasing mobile protein and peptide content (16–18), but also offer extra information about the infarct extent in ischemic strokes, by the virtue of pH changes (14,19–21). Currently, APT studies on brain tumors have been performed successfully on both pre-clinical models (15,22,23) and human patients (16–18). However, studies on strokes are still limited to animals (14,19–21).

The APT approach (14,15) makes it possible to image cellular mobile proteins and peptides (μM to mM concentration range) in vivo (24,25). However, one should note that the APT image contrast depends strongly on the used experimental parameters, such as the RF irradiation power and time. Currently, the APT imaging experimental parameters are to a certain extent arbitrary for human studies on clinical scanners (16–18,26–29). To better evaluate the clinical applications of APT imaging, it is very helpful to use standard and optimized experimental parameters. In APT imaging, the RF saturation time used is usually a few seconds in animal studies (14,15,19–23); however, it is typically limited to 0.5–1 second on human scanners (16–18,26–29). Presently, the dependence of APT contrast on the RF saturation power (or the mean RF saturation power, if the pulse train is used) is still not fully understood in clinical settings. The purposes of this study are to demonstrate the feasibility of APT imaging of stroke patients on a clinical 3T scanner, to experimentally evaluate the RF radiation power dependence of the APT signals in the healthy human brain and in brain tumors and strokes, as well as to propose a standard RF radiation scheme that may be used for APT imaging of the human brain. By using this standard experimental scheme, the clinical APT results acquired from different research centers may be compared to one another.

MATERIALS AND METHODS

Subjects

Total 16 subjects comprising of eight brain tumor patients (high-grade glioma in four, lung cancer metastasis in one, meningioma in three; all at initial presentation prior to treatment), four stroke patients (average 4.3 ± 2.5 days after the onset of the stroke; range 1–7 days), and four healthy volunteers were recruited to this study. Written and informed consents were obtained from all subjects. The study protocol was approved by the local institutional review board.

MRI Acquisition

All subjects were scanned on a Philips 3T MRI scanner (Achieva 3.0T; Philips Healthcare Systems, Best, the Netherlands) equipped with pencil beam second-order shim, using a body coil for RF transmission and an eight-channel sensitivity-encoding (SENSE) coil for reception. Several standard MR images, which are T₂-weighted (T₂w), T₁-weighted (T₁w), fluid attenuated inversion recovery (FLAIR), and gadolinium-enhanced T₁w (Gd-T₁w), were acquired for reference. APT imaging was acquired before the administration of the Gd contrast agent.

An off-resonance continuous-wave RF saturation pulse with a duration of 500 ms (which currently is the longest allowed for our body coil) was used for APT imaging in this study. For patient studies, three different saturation powers of 1, 2, and 3 μT were applied. For healthy volunteers, eight power levels from 0.5 to 4 μT with a step of 0.5 μT were used. Some other imaging parameters were as follows: a single-slice turbo spin echo (TSE) imaging readout with a SENSE factor of 2; repetition time = 3 sec; echo time = 11 msec; matrix = 128×64; field of view = 240×240 mm²; and slice thickness = 6 mm. Pencil beam second-order shimming was employed. We used a multi-offset, multi-acquisition APT imaging protocol (18), in which the APT image scan and z-spectrum scan were combined. Using this APT acquisition, the procedure to turn off the pre-scan between the APT image scan and z-spectrum scan, which can affect the shim and frequency offset settings, was not needed. This protocol (31 offsets = 0, ±0.25, ±0.5, ±0.75, ±1, ±1.5, ±2, ±2.5, ±3 (2), ±3.25 (4), ±3.5 (8), ±3.75 (4), ±4 (2), ±4.5, ±5, ±6 ppm; the values in parentheses were the number of acquisitions, which was 1, if not specified; an unsaturated image was acquired for the signal normalization) can provide B₀ inhomogeneity-corrected APT images with sufficient signal-to-noise ratios within a clinically relevant time frame, which was about three minutes per power level.

Theoretical Considerations

To better quantify APT effects, as described previously (17), it is necessary to reduce the interference of other saturation effects that are concurrent with APT measurements, such as direct water saturation and conventional magnetization transfer (MT) (30). This is usually achieved by the so-called magnetization-transfer-ratio (MTR) asymmetry analysis with respect to the water resonance, namely, by subtracting the MTR (MTR = 1 - S_sat/S₀, where S_sat and S₀ are the signal intensities with and without RF irradiation, respectively) obtained at the negative offset from that obtained at the corresponding positive offset:

$$\begin{array}{cc}\hfill {\text{MTR}}_{\text{asym}}\left(\text{offset}\right)& =\text{MTR}(+\text{offset})-\text{MTR}(-\text{offset})\hfill \\ \hfill & ={\mathrm{S}}_{\text{sat}}(-\text{offset})∕{\mathrm{S}}_0-{\mathrm{S}}_{\text{sat}}(+\text{offset})∕{\mathrm{S}}_0.\hfill \end{array}$$ [1]

Specifically for APT imaging, we have:

$$\begin{array}{cc}\hfill {\text{MTR}}_{\text{asym}}\left(3.5\text{ppm}\right)& ={\mathrm{S}}_{\text{sat}}(-3.5\text{ppm})∕{\mathrm{S}}_0-{\mathrm{S}}_{\text{sat}}(+3.5\text{ppm})∕{\mathrm{S}}_0\hfill \\ \hfill & \approx {\text{MTR}}_{\text{asym}}^{\prime }\left(3.5\text{ppm}\right)+\text{APTR}.\hfill \end{array}$$ [2]

Namely, the measured APT-MRI signal in tissue, MTR_asym(3.5ppm), consists of not only the APTR, based on the exchangeable protons of mobile cellular proteins and peptides, but also the residual non-APT component or conventional MTR asymmetry (13,31), ${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$. The presence of${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$ would cause a negative background for obtained MTR_asym(3.5ppm) images, which greatly complicates APT imaging quantification in vivo. To reflect that the APT-MRI signals measured are contaminated by the conventional MTR asymmetry, the MTR_asym(3.5ppm) images were called the APT-weighted (APT_w) images.

Data Analysis

All data were processed using programs written in interactive data language (IDL; Research Systems, Inc., Boulder, CO, USA). Briefly, the raw image data were first organized into the z-spectrum (the normalized signal intensities, S_sat/S₀, as a function of 31 offsets, where S_sat and S₀ are the signal intensities with and without radiofrequency irradiation, respectively). Then, the z-spectrum was fitted through all offsets using a 12th-order polynomial (the maximum order available with IDL) on a voxel-by-voxel basis (17,18). The fitted curve was interpolated using an offset resolution of 1 Hz. After this step, the corresponding B₀ field inhomogeneity was calculated according to the deviation of the minimum of the fitted curve from 0 ppm. To correct for the field inhomogeneity effect, the original z-spectrum was interpolated and centered along the direction of the offset axis to shift its lowest intensity to 0 ppm. The realigned z-spectra were interpolated back to 31 points. Finally, the APT_w image (namely, MTR_asym(3.5ppm) image) was calculated using the B₀-corrected data at the offset of ±3.5 ppm. The image was thresholded based on the signal intensity of S₀ image to remove voxels outside the brain.

Regions of interest (ROIs) were carefully chosen by experienced radiologists. For healthy normal brain tissues, the ROIs were drawn on relatively homogenous white matter. For brain tumors, the ROIs were placed on the Gd-enhancing tumor cores and the contralateral normal-appearing white matter (CNAWM). For stroke cases, the lesions were selected according to the signal abnormality on T₂w images and FLAIR images. All results were expressed as mean ± standard deviation. A one-way analysis of variance (ANOVA) test, followed by the Tukey multiple comparisons, was applied to analyze the statistical differences. The level of significance was set at P < 0.05.

RESULTS

Figure 1 shows z-spectra and MTR_asym spectra for several RF saturation power levels from healthy human brains (n = 4). The imaging signal intensities in the z-spectra were substantially attenuated (Fig. 1a). This is mainly due to the effects of direct water saturation and conventional MT (30). The MTR asymmetry analysis was used to isolate the APT effects from these concomitant saturation effects (14,15). The resulting curves of MTR asymmetry depended strongly on the applied RF radiation power (Fig. 1b).

Fig. 1.

Measured z-spectra and MTR_asym spectra of white matter from healthy subjects (n = 4). a: z-spectra. b: Corresponding MTR_asym spectra. c: Plot of MTR_asym(3.5ppm) as a function of RF saturation power levels. The error bars are too small to see clearly. MTR_asym(3.5ppm) is roughly zero for the RF saturation power level of 2 μT.

More interesting results can be observed from the plots of MTR_asym(3.5ppm) as a function of RF saturation power applied (Fig. 1c). Within the range of radiation power levels from 0.5 to 4 μT with a step of 0.5 μT, the measured MTR_asym(3.5 ppm) changed from approximately −2.2% to 2.2%. Particularly, it should be pointed out that the MTR_asym(3.5ppm) curve reached a minimum at the RF radiation power of ~1 μT and crossed zero at ~2 μT (MTR_asym(3.5ppm) = 0.21% ± 0.17% at 2μT). Then, MTR_asym(3.5 ppm) increased gradually with increasing RF power levels and reached a plateau at about 3.5–4 μT. Unlike animal research systems, clinical scanners have specific absorption rate (SAR) limitations. Thus, no power levels >4 μT were used in this study. However, it is reasonable to predict that this MTR_asym(3.5ppm) curve would go down for higher RF power levels (>4 μT), due to the increasing effects of direct water saturation and conventional MT. According to Eq. [2], the resulting MTR_asym(3.5 ppm) comprises the inherent conventional MTR asymmetry (13,31), ${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$ (which is negative), and the APTR associated with the cellular mobile protein and peptides. Our experimental result shows that at 2 μT, ${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$ and APTR nearly canceled each other out, so MTR_asym(3.5 ppm) was almost zero.

Figure 2 shows z-spectra, MTR_asym spectra, and ΔMTR_asym spectra (MTR_asym for the lesion - MTR_asym for the CNAWM) acquired using three RF radiation power levels for brain tumors (n = 8). The measured image signal intensities for both the tumor and the CNAWM (Fig. 2a, d, and g) clearly decreased as the RF radiation power increased. In addition, the z-spectra for the tumor shifted upward, compared to the CNAWM. This may be attributed to several factors, such as increasing water content that may affect the relaxation times and increasing macromolecular mobility in these regions (32,33).

Fig. 2.

z-spectra, MTR_asym spectra, and ΔMTR_asym spectra for three RF saturation power levels (a–c: 1 μT; d–f: 2 μT; g–i: 3 μT) measured from brain tumor patients (n = 8). The MTR_asym(3.5ppm) value for the CNAWM is approximately zero for 2 μT. The ΔMTR_asym spectra are maximized at the offsets of about 3–4 ppm with respect to the water resonance. The ΔMTR_asym(3.5ppm) difference is small across the three RF power levels.

When the MTR asymmetry analysis was used, the resulting MTR_asym spectra for both tumor and CNAWM, over the resonance frequency range (2–4 ppm downfield of the water resonance) of mobile protein and peptide amide protons, increased with increasing RF radiation power (Fig. 2b, e, and h). The MTR_asym spectra for all RF power levels were higher in the tumor than in the CNAWM (except at <1 ppm). The MTR_asym difference between the tumor and the CNAWM at each RF power level was most dominated by the higher mobile protein and peptide content in tumor (2,34), as discussed below. It is important to note that the MTR asymmetry analysis did not remove all interferences because the conventional MT effect was somewhat asymmetric with respect to the water resonance (13,31). Thus, the measured APT-MRI signals had a contribution from this conventional MTR asymmetry. Because of this contamination, at the lower RF power of 1 μT, the tumor MTR_asym spectra varied around zero, although there was an APT effect. At 2 μT, the magnitude of MTR_asym(3.5ppm) in the CNAWM was around zero (0.18% ± 0.23%), which is in agreement with the above measurements of the normal brain at 2 μT (Fig. 1b and c).

To better understand the APT contrast characteristics between the tumor and the CNAWM, it is important to inspect the ΔMTR_asym spectra (Fig. 2c, f, and i). We can see that the ΔMTR_asym spectra for all RF powers showed a rise from 1–3 ppm and then a decreasing trend after 4.5 ppm, a phenomenon that would be more convincing if more higher offsets (>5 ppm) are acquired, as observed in our previous studies (14,17). The ΔMTR_asym spectra were maximized at 3–4 ppm, with respect to the water resonance (4.75 ppm), which corresponds to the resonance frequency range of mobile protein and peptide backbone amide protons in the proton MR spectrum (35).

Figure 3 shows z-spectra, MTR_asym spectra, and ΔMTR_asym spectra of stroke lesions (n = 4) for three RF radiation power levels. Compared to the CNAWM, the z-spectra for the stroke show visible upward shifts (Fig. 3a, d, and g), which may be primarily due to increases in T₂ in these acute or subacute stroke lesions. As seen from Fig. 3b, e, and h, the MTR_asym spectra of stroke and CNAWM both increased with the RF saturation power levels from 1 to 3 μT. For each RF power level, the MTR_asym spectra were lower in the stroke than in the CNAWM (except at <1 ppm), suggesting the presence of the possible APT effect in the lesion that may be associated with lower pH (36) and some other factors (see the Discussion section below). Therefore, there was a negative APT-MRI contrast between the stroke and the CNAWM (Fig. 3c, f, and i), as observed previously in pre-clinical models (14,19–21).

Fig. 3.

z-spectra, MTR_asym spectra, and ΔMTR_asym spectra for three RF saturation power levels (a–c: 1 μT; d–f: 2 μT; g–i: 3 μT) measured from stroke patients (n = 4). The magnitude of ΔMTR_asym(3.5ppm) increases with applied RF saturation power levels.

Figure 4 compares the RF power dependence characteristics of APT imaging of human brain tumors and strokes at 3T. It can be seen that the apparent APT-MRI signals (namely, MTR_asym(3.5ppm), which is equal to ${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$ + APTR) measured for both pathologies increased significantly, when the RF saturation powers increased from 1 to 2, then to 3 μT (all P < 0.001, except for the stroke from 2 to 3 μT, P < 0.01). On the other hand, the corresponding APT-MRI contrasts (exactly speaking, ΔMTR_asym(3.5ppm) contrasts) for these two pathologies demonstrated a different RF power dependence. In the case of tumors, it seems that the measured magnitude of ΔMTR_asym(3.5ppm) decreased as the RF saturation power was increased, but the changes in ΔMTR_asym(3.5ppm) were statistically insignificant (both P > 0.2). However, for the stroke case, the magnitude of ΔMTR_asym(3.5ppm) was lower at 1 μT, but increased when 2 and 3 μT were used. The changes in ΔMTR_asym(3.5ppm) were significant from 1 to 2 μT (P < 0.05), but insignificant from 2 to 3 μT (P > 0.2).

Fig. 4.

MTR_asym(3.5ppm) (a) and ΔMTR_asym(3.5ppm) (b) for tumors and strokes as a function of RF irradiation power levels. MTR_asym(3.5ppm) increases significantly with RF saturation power levels for the tumors (P < 0.001, both from 1 to 2 μT and from 2 to 3 μT) and the strokes (P < 0.001, from 1 to 2 μT; P < 0.01, from 2 to 3 μT). The changes in ΔMTR_asym(3.5ppm) are significant for the strokes as the RF irradiation power increases from 1 to 2 μT (P < 0.05), but insignificant for the others (all P > 0.2). * P < 0.05; ** P < 0.001; *** P < 0.001; n.s.: nonsignificant.

Examples of the APT_w MR images of brain tumors and strokes using the RF saturation power of 2 μT are shown in Fig. 5. In the case of brain tumors (Fig. 5a–d), there was a clear increase in APT signal intensity in tumor mass, compared to the tumor periphery and the surrounding normal-appearing white matter, including the cerebrospinal fluid (CSF). The tumor mass identified by the hyperintense APT area was in size similar to that identified by Gd-enhanced T₁w images. This is expected because the cerebral metastasis generally shows a clear edge. On the other hand, the stroke lesion (5 days after onset) was hypointense on the APT_w images (Fig. 5e–h), as also expected.

Fig. 5.

Conventional and APT_w MR images of a patient with lung cancer metastasis (a–d) and a patient with stroke at 5 days post-onset (e–h). The RF saturation power used for APT imaging was 2 μT. The tumor (solid arrow) is hyperintense, while the stroke (open arrow) is hypointense on the APT_w images. The hyperintense signal in the basal ganglia region (black arrow) contralateral to ischemia on the APT_w image (h) is an artifact.

In this study, the APT_w MR images were displayed by rainbow colors (instead of grayscale) and a display window (−5%, 5%) was used for the RF saturation power of 2 μT. As illustrated in Fig. 6 (the same brain tumor patient as Fig. 5a–d), a proper window should be selected to improve display effects of APT_w images. Here, two interesting things should be pointed out. First, when proper different display windows were used (green for the CNAWM), the resulting APT_w MR images corresponding to the three RF irradiation power levels applied (1, 2, and 3 μT) became very similar. Second, there existed some CSF-related artifacts around the tumor edge on the APT_w image with the saturation power of 1 μT (Fig. 6c), which were not visual on the APT_w images with both 2 (Fig. 6b) and 3 μT (Fig. 6f).

Fig. 6.

APT_w MR images (a–d, f) at different saturation power levels and for different display windows and Gd-T₁w MR image (e) for a patient with cerebral metastasis (the same patient as Fig. 5a–d). The tumor (red bold arrow) shows hyperintense on all of the APT_w images. However, the lesion appears diffuse on the APT_w image at 1 μT (a,d) due to the presence of CSF artifacts (white thin arrow).

DISCUSSION

The APT approach has successfully been applied to the imaging of brain tumors (16ȃ18) and strokes (14,19–21). However, care should be taken when interpreting the observed APT effects that in principle originate from backbone amide protons associated with endogenous mobile, cytosolic proteins and peptides. To reduce the interference of the coexisting conventional MT and direct saturation effects, APT imaging data is generally processed using an asymmetry analysis. However, one should keep in mind that these concomitant saturation effects are not completely removed and the conventional MTR asymmetry (13,31), ${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$, still remains in the APT-MRI signal. Namely, only an apparent APT-MRI signal is measured in our APT experiments. Quantitatively, according to Eq. [2], the APT-MRI signal measured in vivo, MTR_asym(3.5ppm), consists of the APTR and the conventional MTR asymmetry, ${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$. Based on a two-pool exchange model (37), the APTR depends primarily on mobile amide proton concentration and amide proton exchange rates (which are related to tissue pH). However, the origin of the conventional MTR asymmetry, ${\text{MTR}}_{\text{asym}}^{\prime }=\left(3.5\text{ppm}\right)$, is not completely clear, which may include the inherent asymmetry of the solid-phase macromolecular MT effect and possible intramolecular and intermolecular nuclear Overhauser effects (NOE) of aliphatic protons of mobile macromolecules and metabolites. It has been shown previously that like the z-spectrum, this residual non-APT component is evidently power dependent (31). Consequently, the experimentally obtained APT signals are complicated, which depend not only on the properties of tissue but also on the experimental parameters applied. Presently, there is no effective way to separate amide proton concentration, pH, and conventional MTR asymmetry effects in APT-MRI signals; moreover, it is very likely that the conventional MTR asymmetry depends on tissues or pathological states. In spite of these, our previous experiments on animals (14) and humans (17) have shown that the difference between the MTR_asym spectra for different pathological states (at the same power level) becomes very small at larger offsets (e.g., >6 ppm). This suggests that the magnitude of ΔMTR_asym(3.5ppm) may be dominated by the APTR, rather than by the conventional MTR asymmetry. Therefore, for convenience, APT_w images may also be called APT images.

Particularly in the case of brain tumors, increasing MTR_asym (3.5ppm) in the tumor with respect to the normal brain tissue can be attributed to increasing protein and peptide content, increasing pH, or decreasing conventional MTR asymmetry. Note that the potentially decreasing conventional MTR asymmetry in the tumor means decreasing |MTR′_asym (3.5ppm)| or increasing MTR′_asym (3.5ppm), where MTR′_asym (3.5ppm) is negative. However, our data in this and previous (15,17,22) studies have clearly shown a visible APT effect in the offset range of 2–4 ppm, downfield from the water resonance, in the MTR_asym spectra. Moreover, the increasing APT effect in the tumor, compared to the normal brain tissue, is often maximized at the offset of 3.5 ppm, which corresponds to the composite resonance frequency of mobile protein and peptide backbone amide protons at 8.25 ppm in the proton MR spectrum (35). Further, because often only a small intracellular pH increase is detected in the tumor, with respect to the normal brain tissue (38), this increasing APT effect in the tumor can be attributed primarily to increasing protein and peptide content (2,34). Similarly, in the case of strokes, the MTR_asym (3.5ppm) abnormalities observed in the lesion may be associated with several factors. At the hyperacute stage, there is a drop of pH (~0.5 pH units (36)) in the lesion. The APT-MRI hypointensity observed may be dominated by the pH reduction because the changes in the protein and peptide content and the convention MTR asymmetry should be minimal, as described previously in our animal study (14). At the acute and subacute stages, as studied in this work, it is very likely that there is still a decreasing pH in the stroke region, resulting in the lower APT-MRI signal (namely, MTR_asym(3.5ppm) ). However, the observations may be contaminated by other possible factors, such as changes in mobile protein and peptide content and conventional MTR asymmetry. Particularly, unlike in the case of brain tumors, the potentially decreasing conventional MTR asymmetry in the stroke (namely, decreasing |MTR′_asym (3.5ppm)| or increasing MTR′_asym (3.5ppm); compared to the normal brain tissue) during the acute and subacute stages may increase the MTR_asym (3.5ppm) intensity in the lesion, thus potentially decreasing the APT image contrast between the lesion and the surrounding brain tissue in the APT_w image.

The selection of appropriate (optimized and feasible) experimental parameters, particularly the RF saturation power, would be very important in APT imaging. The RF power dependence of the CEST effect (including APT) has previously been investigated in vitro (39) and in vivo (40). It is worthwhile to note that when performing studies in vivo, the conventional MT effects should be taken into account. Interestingly, the APT signal in vivo in the animal brains shows an RF power dependence characteristic similar to in vitro, which initially increases with increasing RF irradiation power levels but subsequently decreases. Currently, except for an MTR asymmetry study in a broad offset on healthy subjects (31), the previous APT or CEST contrast investigations of RF power dependence were limited to phantoms or animals, which were performed in animal scanners. It is worth noting that the optimal RF irradiation scheme obtained from the animal scanners is not applicable to humans on clinical scanners. For the clinical applications of APT imaging, the RF irradiation scheme (saturation time and power) is typically limited by scanner hardware and SAR regulations, particularly when the body coil is used for the RF transmission. As discussed at the beginning of this section, the APT-MRI signal measured in vivo is very complicated. Therefore, it is very difficult to establish an exact theoretical model to optimize the RF saturation power for APT imaging. In this study, based on the experimental data of brain tumors and strokes, we expect to propose a practical, best possible RF saturation power that not only results in large APT image contrast but also stays within the SAR limits.

According to the experimentally obtained MTR_asym (3.5ppm) data from the normal human brain (Fig. 1c), the APT-MRI signals are nearly nulled at 2 μT, because MTR′_asym (3.5ppm) (which is negative) and APTR cancel each other out. Although this result is hard to understand, the finding that 2-μT power nulls normal tissue may help standardize the experimental parameters for clinical applications. On the one hand, as shown in Fig. 5d, this RF irradiation power of 2 μT would lead to a positive MTR_asym (3.5ppm) value in brain tumors, whereas both the tumor periphery (which is typically hyperintense on the T₂w images) and the surrounding normal-appearing white matter are almost nulled. Namely, the tumors would show hyperintensity, compared to CNAWM, on APT_w images. On the other hand, the 2-μT irradiation power would cause a negative MTR_asym (3.5ppm) signal in the stroke, at least for hyperacute ischemic stroke patients, where there is a drop in pH (~0.5 pH unit (36)). As shown in Fig. 5h, the lesions, compared to CNAWM, would be hypointense on APT_w images. Therefore, the use of 2-μT irradiation power in APT imaging best allows radiologists to differentiate between these two entities in the clinic.

Our experimental results (Figs. 2–4) show that the apparent APT signals, MTR_asym (3.5ppm), in brain tumor, stroke, and CNAWM all increase with increasing RF radiation power levels (1 to 3 μT; all P < 0.05). However, the APT-MRI contrast between brain tumor and CNAWM, namely, ΔMTR_asym(3.5ppm), decreases slightly from 1 to 3 μT (insignificantly between 1 and 2 μT or between 2 and 3 μT; marginally significantly between 1 and 3 μT, P = 0.049), while the APT-MRI contrast between stroke and CNAWM increases (more negative) within this RF power range (significantly between 1 and 2 μT or between 1 and 3 μT; insignificantly between 2 and 3 μT). In all tumor and stroke cases, 2 and 3 μT have similar APT image quality (insignificant contrast difference, see Fig. 4b; less artifacts). When a proper display window is used (green in the CNAWM), the resulting APT_w images become very similar (Fig. 6). Because 3 μT would cause more RF energy absorption in tissues, it is obvious that the irradiation power of 2 μT should be used in APT imaging. In the case of brain tumors, the APT contrast seems higher at 1 μT than at 2 and 3 μT. However, the more negative background in the APT_w images at 1 μT than at 2 and 3 μT may complicate the explanation of the images. This is because the APT signal of the CSF is always approximately zero (independent of the power), which is almost the same as the APT signal of the tumor at 1 μT (Fig. 4a). The higher APT signal in CSF than in CNAWM may cause an erroneous diagnosis, when the lesion is close to large CSF regions, such as the ventricles (Fig. 6a and d). Finally, in the case of strokes, the APT contrast at 1 μT is quite low (Fig. 4b). Consequently, the 1-μT RF saturation power may not be used for the APT studies of both brain tumors and strokes.

CONCLUSIONS

We have demonstrated the feasibility of APT imaging of stroke patients on a clinical 3T scanner and experimentally determined the RF saturation power dependence of APT image characteristics in normal brain, brain tumors, and strokes on clinical scanners. When an irradiation power of 2 μT is applied, the tumor and stroke lesions show hyperintensity and hypointensity, respectively, compared to the CNAWM, while most normal (or normal-appearing) brain areas, including the CSF, are iso-intense on the APT_w images. Therefore, the RF irradiation power of 2 μT can make the interpretation and differentiation of these two types of brain lesions straightforward and should be useful for clinical applications of APT imaging in the brain.