Introduction

According to the West China Glioma Center (WCGC) and Central Brain tumour Registry of the United States (CBTRUS), metastasis and glioma are the most frequent malignant brain neoplasms, and the incidence rate of both diseases has been increasing in recent decades [1, 2]. The ability to distinguish metastases from glioma is critical because the therapeutic planning and follow-up of these tumours are vastly different [3]. If metastasis is suspected, high-dose Gd should be used to elucidate other metastases in the brain [4], and other parts of the body should be examined. For metastasis therapy, nonsurgical treatment is preferred over conventional surgical resection [5]. However, when glioma is suspected, various advanced MRI techniques are required to detect the true tumour margin, and surgical resection is the first choice [6]. Usually, differentiation between the two neoplasms depends primarily on clinical history, particularly in cases with multiple lesions. However, if the patient’s clinical history is uncertain, or the lesion is solitary, the distinction of these neoplasms may be difficult [7]. On conventional MRI, there is substantial overlap between the images of solitary brain metastases (SBMs) and glioblastomas (GBMs), as both manifest similar signal features and contrast enhancement patterns. Previously, investigators worldwide used DWI, PWI, MRS and ASL perfusion imaging to differentiate between these tumours [8,9,10,11]. However, the distinction between these two types of tumours still has limited diagnostic specificity.

Amide proton transfer-weighted (APTw) MR imaging was invented by Zhou and van Zijl et al. and is a novel molecular MRI technique based on chemical exchange-based saturation transfer (CEST), which was designed to detect amide protons of endogenous low-concentration mobile proteins and peptides in tissue [12,13,14]. This technique does not require the use of exogenous contrast agents and is applied to imaging through the exchange between amide protons of mobile proteins and peptides and protons of bulk water. The APTw signals are mainly related to cell density and endogenous mobile proteins and peptides [15,16,17,18]. APTw imaging has emerged as a valuable tool for grading brain tumours [18,19,20] and other cancers that occur in the prostate [21], breast [22, 23] and neck [24, 25]; for distinguishing tumour recurrence from radiation necrosis [26], distinguishing pseudo-progression from true progression in gliomas [27] and differentiating between primary central nervous system lymphomas and glioma [15]; and for characterising cerebral ischaemia [28,29,30]and Parkinson’s disease [31]. However, this technique has not been used to compare SBMs and GBMs. Therefore, the purpose of this study was to evaluate the usefulness of APTw MRI in distinguishing SBMs from GBMs at the molecular level based on their APTw values.

Materials and methods

Study population

The institutional review board of our hospital approved this study, and each patient signed the consent form. From May 2011 to October 2016, 94 patients with SBMs or GBMs were recruited and received conventional and APTw MRI. Data from six patients were excluded because of insufficient image quality due to motion artefacts. Therefore, 88 patients were included in this study. Forty-five patients (33 males, 12 females; age range, 30-74 years; mean age, 56.5 ± 9.2 years) had SBMs confirmed by histopathology (n = 39) or clinical diagnosis (n = 6). The primary sites of cancers included the lung (n = 28), breast (n = 6), kidney (n = 4), colon (n = 4), liver (n = 2) and the lymphatic system (n = 1), and 43 patients (29 males, 14 females; age range, 18-71 years; mean age, 44.8 ± 13.8 years) had GBMs confirmed by histopathology.

MR imaging

All patients underwent a conventional brain MRI sequence and two-dimensional (2D) APTw sequence on a 3.0-T MRI scanner (Achieva 3.0 T; Philips Medical Systems, Best, The Netherlands) before clinical intervention. A body coil was used for radiofrequency (RF) transmission, and a 16-channel head coil was used for signal reception. Conventional MRI sequences included T1-weighted [repetition time (ms)/echo time (ms), 400/20], T2-weighted [repetition time (ms)/echo time (ms), 2800/105] and fluid-attenuated inversion recovery [FLAIR, repetition time (ms)/echo time (ms)/inversion recovery time (ms) 8000/204/2200]. Then, a contrast agent (Gd-DTPA; 0.2 ml/kg body weight; Magnevist; Bayer Schering, Guangzhou, China) was injected through the median cubital vein, and gadolinium-enhanced T1-weighted images were acquired. Other imaging parameters were as follows: field of view, 240 × 240 mm2; slice thickness, 5 mm; gap, 2 mm; matrix, 512 × 512.

The 2D fat-suppressed, fast spin-echo APTw pulse sequence was performed at one T2-weighted image slice showing the maximum area of the tumour before the Gd-T1-weighted image. For saturation, a pulse-train radiofrequency saturation (duration time =800 ms; inter-pulse delay =10 ms; power level =2 μT) was used. APTw imaging was performed with a multi-offset (offsets =0, ±0.25, ±0.5, ±0.75, ±1, ±1.5, ±2, ±2.5, ±3, ±3.25, ±3.5, ±3.75, ±4, ±4.5, ±5 and ±6 ppm) multi-acquisition protocol. The protocol was repeated eight times at an offset of ±3.5 ppm to increase the signal-to-noise ratio of the APTw images. In addition, an image that did not exert a saturated pulse was acquired for the signal normalisation, and an image that exerted a saturated pulse at the offset of 15.6 ppm was acquired to calculate the conventional MTR value. The detailed imaging parameters were as follows: sensitivity-encoding factor = 2, repetition time (s), 3; echo time (ms), 11; field of view, 240 × 240 mm2; section thickness, 6 mm; matrix, 128 × 128; voxel size, 1.65 × 3.15 × 6.00 mm3. The total acquisition time was 192 s.

Image analysis

The APTw raw data were analysed using interactive data language written in the programme (IDL; Research Systems, Inc., Boulder, CO, USA). First, the normalised saturated signal intensity curve (Ssat/S0, where Ssat and S0 were the signal intensities obtained with and without selective saturation, respectively) as a function of saturation frequency offset, usually called the Z-spectrum, was calculated. The B0 field inhomogeneity effect was corrected, as described previously [18]. To reduce the contributions from conventional magnetisation transfer contrast and direct saturation of bulk water, a B0-corrected Z-spectrum was used to analyse the magnetisation transfer ratio asymmetry (MTRasym) as follows: MTRasym = Ssat(-offset)/S0-Ssat(+offset)/S0. The APTw signal was calculated as MTRasym (3.5 ppm).

Images analysis was performed by two experienced neuroradiologists (L.W. and S.J., who have 11 and 10 years of experience in neuroradiology, respectively). The tumour enhancement type and oedema index (EI) were recorded. Five regions of interest (ROIs) were distributed within the Gd-enhancing tumour area, as well as within the peritumoral brain zone (PBZ; peritumoral T2w hyperintense area, 0.5-2.5 cm from the enhancing tumour area to avoid partial volume contamination) (Fig. 1), based on the Gd-T1WI and T2WI co-registered with the APTw image [32]. Large cystic cavities and haemorrhagic components were always excluded. The size of each ROI was 15 pixels. In addition, the contralateral normal-appearing white matter (CNAWM) was analysed (Fig. 1). For each patient, the APTw values for all ROIs were recorded. Then, the maximum APTw value (APTwmax), minimum APTw value (APTwmin) and mean APTw value (APTwmean) were determined. To reduce the effect of patient age [33] and other potential experimental errors [34], we also recorded the relative APTw (rAPTw = APTw - APTw CNAWM): rAPTwmax, rAPTwmin and rAPTwmean.

Fig. 1
figure 1

Example of the placement of ROIs. Five ROIs were placed in the Gd-enhancing tumour part (black circle), five ROIs were placed in the peritumoral T2 hyper-intensity area (white circle), and one ROI was placed in the contralateral normal-appearing white matter (red circle) based on the co-registered traditional images. The necrosis-related image artefact (white arrow) can also be seen

Full size image

Pathological data acquisition

The pathologic images were analysed with image analysis software (Image-Pro Plus, version 6.0, Media Cybernetics, Silver Spring, MD, USA). One neuropathologist, blinded to the MRI features, reviewed the haematoxylin and eosin (H&E)-stained biopsy samples from the tumour core and the PBZ of 39 patients with SBM and 43 patients with GBM. The pathological type and tumour grade were recorded according to 2007 WHO classification of tumours of the CNS. The presence of infiltrating cells in the PBZ was also estimated.

Statistical analysis

Statistical analysis performed was statistical analysis software packages (SPSS19.0 and MedCalc v13.1.2.0). The patient’s gender was compared between SBMs and GBMs using the chi-square test. Comparison among APTwmean values for the tumour core, PBZ and CNAWM was performed using a one-way analysis of variance (ANOVA) test, followed by multiple comparisons using the Tukey test, to analyse the statistical significance of differences. Comparison between patient’s age, APTwCNAWM, APTwmax, APTwmin, APTwmean, rAPTwmax, rAPTwmin or rAPTwmean values for SBMs and GBMs was performed using an independent-samples t-test. Receiver-operating characteristic (ROC) curves were generated for each parameter to assess the areas under the curves (AUCs) and to determine optimal cut-off values according to Youden’s index for the discrimination of SBMs and GBMs. P < 0.05 was considered a statistically significant difference for all tests.

Results

There was no gender difference between the SBM and GBM groups (P = 0.545). The patient age was higher in the SBM group than in the GBM patients (P < 0.001).

Pathological data

In biopsy samples from the PBZ, the tumour cells were present in 29/43 samples; by contrast, 7/39 samples were present in the SBM group.

Comparison of conventional and APTw image features

Forty-five cases of SBMs (45 lesions) were located in the cerebral parenchyma (n = 36) and in the cerebellar hemisphere (n = 9). Eight cases showed mild oedema, 26 cases showed moderate oedema, and 11 cases showed severe oedema. On the Gd-T1WI, tumours showed ring-like (n = 13), nodular (n = 22) or patchy (n = 10) enhancement. The tumour cores showed hyperintensities (relative to the CNAWM) on the APTw images, and the hyperintense areas on APTw images were approximately equal to the lesions shown on the Gd-T1WI. For the 43 cases of GBMs (43 lesions), the lesions were located in the cerebral parenchyma (n = 35), brainstem (n = 2) and cerebellar hemisphere (n = 6). Seven cases showed mild oedema, 30 cases showed moderate oedema, and 6 cases showed severe oedema. In the Gd-T1WI, the tumours showed ring-like (n = 24), nodular (n = 11) or patchy (n = 8) enhancement. Compared with the CNAWM, the lesions identified by the APTw image were larger than those identified by the Gd-T1WI. Figure 2 shows an SBM patient, and Fig. 3 shows a GBM patient.

Fig. 2
figure 2

Results for a 54-year-old female with solitary brain metastasis (white arrows). The primary tumour was lung adenocarcinoma. A-C The tumour was located in the right parietal-occipital lobe with moderate peritumoral oedema. It appears hypointense on T1WI, hyperintense on T2WI and ring-like contrast enhanced on Gd-T1WI. D An APTw image shows a relatively homogeneous hyperintense mass (tumour core, APTWmax =3.55%, APTWmin =2.87%, APTWmean =3.34%; peritumoral oedema, APTWmax =1.72%, APTWmin =1.57%, APTWmean =1.63%; APTWCNAWM = 0.74%). E H&E-stained sections (×200) of the tumour core show many dense, aggregated cells with atypical nuclei arranged in glandular, nest and cord patterns. F H&E-stained sections (×100) of the peritumoral oedema show the degeneration of nerve cells and the proliferation of glial cells, but no tumour cells are found

Full size image
Fig. 3
figure 3

Results for a 67-year-old male with glioblastoma. A-C The tumour is located in the right occipital lobe with mild peritumoral oedema. The tumour appears hypointense on the T1WI, hyperintense on the T2WI and ring-like contrast enhanced on the Gd-T1WI. D An APTW image shows a heterogeneously hyperintense mass (tumour core: APTWmax =3.87%, APTWmin =2.96%, APTWmean =3.61%; peritumoral oedema: APTWmax =2.09%, APTWmin =1.63%, APTWmean =1.89%; APTWCNAWM =0.57%). E H&E-stained sections (×200) of the tumour core show many multinucleated atypical tumour cells with relatively abundant cytoplasm. F H&E-stained sections (×100) of the peritumoral oedema show proliferating glial cells and a number of cells with atypical nuclei

Full size image

Quantitative APTw image analysis

The APTw value in the CNAWM was significantly lower in SBMs than in GBMs (0.49% ± 0.18% vs. 0.54% ± 0.14%, P = 0.046). The APTw value in the tumour core of SBMs and GBMs was significantly higher than in the CNAWM (2.76% ± 0.71% vs. 0.49% ± 0.18%, P < 0.001 and 2.94% ± 0.67% vs. 0.54% ± 0.14%, P < 0.001). The APTwmean in the PBZ of SBMs and GBMs was significantly higher than in the CNAWM (1.23% ± 0.23% vs. 0.49% ± 0.18%, P < 0.001 and 1.71% ± 0.34% vs. 0.54% ± 0.14%, P < 0.001). The APTwmax, APTwmin, APTwmean, rAPTwmax, rAPTwmin or rAPTwmean in the tumour core showed significant differences between SBMs and GBMs (P > 0.05). By contrast, all parameters in the PBZ were lower in the SBMs than in the GBMs (P < 0.001) (Table 1 and Fig. 4).

Table 1 APTW parameter values of the tumour core and peritumoral brain zone
Full size table
Fig. 4
figure 4

Comparisons of the amide proton transfer-weighted (APTw) parameter (APTwmax, APTwmean, APTwmin, rAPTwmax, rAPTwmean, rAPTwmin) values measured in the peritumoral oedema for GBMs and SBMs; APTw values in the SBM group are significantly lower than those in the GBM group (P < 0.001)

Full size image

ROC analyses of the APTw parameter values in the PBZ are summarised in Table 2. The APTwmin had the highest AUC of 0.905. When the APTwmin of the PBZ was less than 1.21%, the lesion was diagnosed as an SBM. The sensitivity was 84.4%, the specificity was 86.1%, and the accuracy was 85.2% (Table 2).

Table 2 ROC analysis results of APTw parameter values in the peritumoral oedema
Full size table

Discussion

In this study, the gender distribution of patients with SBMs was not significantly different from that of patients with GBMs, although these two tumour types have been shown to be more common in males. Tan et al. [35] found that there was no age difference between the SBM group and high-grade glioma group, but our result shows the SBM patients were significantly older than the GBM patients. As in a previous study [33, 34], the white matter volume was observed to decline after 50 years of age because of demyelination, and the APTw effect due to the myelin basic protein shifted from the oligodendroglia cytoplasm to the myelin sheath. Therefore, we speculated that this was the reason that the APTw parameter value of the CNAWM was lower in the SBM group.

Whether the tumour core can be reliably discriminated between SBMs and GBMs with conventional MRI, even with several advanced MRI techniques such as DWI, PWI, MRS, DTI and ASL, is still controversial [3, 8, 11, 35,36,37]. Previous studies have shown that APTw MRI can identify the tumour core mainly related to the mobile amide proton content and cell density, which was not relevant to Gd enhancement [16, 18]. This finding is particularly important in the context of the glioma grade, because 10% of patients with glioblastoma may demonstrate non-enhancement on Gd-T1WI [15], which may miss the central nervous system damage and nephrogenic systemic fibrosis caused by Gd deposition during re-examination [38, 39]. Here we use APTw imaging to distinguish between SBMs and GBMs. Compared with the CNAWM, both SBMs and GBMs exhibited tumour core hyperintensity on APTw images. Quantitative APTwmin, APTwmax, APTwmean, rAPTwmin, rAPTwmax or rAPTwmean values of the tumour core were not significantly different (P > 0.05). It is well known that as malignant tumours SBMs and GBMs exhibit rapid cell proliferation in the tumour core [35] and that the cell density and mobile proteins and peptides are higher than those of the CNAWM. However, the difference in proton content between SBMs and GBMs is expected to be negligible. An investigation of the tumour core using APTw MRI cannot provide useful information to distinguishing SBMs from GBMs.

The peritumoral brain zone (PBZ) is defined radiologically as the non-enhancing brain area of several centimetres in width surrounding the tumour in Gd-T1WI. Usually, this area shows hyper-intensity on T2WI and FLAIR images because of vasogenic oedema [40]; as the blood-brain barrier breaks down, capillary permeability increases, a pressure gradient from the vasculature to the extracellular environment is formed, and many intravascular proteins penetrate into the extra-cellular space [41]. In this study, the APTw value in the PBZ of SBMs and GBMs was significantly higher than in the CNAWM (P < 0.001). It is interesting to note that the values of APTwmin, APTwmax, APTwmean, rAPTwmin, rAPTwmax or rAPTwmean in the SBM group were significantly lower than in the GBM group (P < 0.001). In metastasis brain tumours, PBZ T2 hyperintensity mainly reflects vasogenic oedema [7, 35]. In the GBMs, it is hypothesised that specific cells and inflammatory cells infiltrate into the PBZ [40, 42,43,44] and that the PBZ T2 hyperintensity reflects more than vasogenic oedema. Technically, the APTw effect is due to the shift of the amide proton from mobile proteins and peptides to the water proton [12]. In the PBZ of GBMs, other than intravascular proteins, because of the infiltration of tumour cells and inflammatory cells, the tumour-related molecules and inflammatory mediators are also retained in the extra-cellular space of the PBZ [41, 45], and the mobile protein and peptide content may be higher than in SBMs. The pathological examination of PBZ supports this hypothesis. Notably, the APTw value in the PBZ of SBMs was significantly higher than in the CNAWM. We speculate that, as the intravascular proteins penetrate into the extra-cellular space of the PBZ, the amide proton content exceeds that in the CNAWM [41].

According to ROC curve analysis, the APTwmin had the highest AUC (0.905) and accuracy (85.2%) for distinguishing between SBMs and HGGs. For the 88 participants in this study, the accuracy for differentiating SBMs from GBMs was 51.6% (by Y.D., a junior radiologist with 3 years of experience in brain imaging, blinded to the pathology reports, based on conventional MRI sequences) and 79.5% (according to official radiology reports, performed by senior neuro-radiologists in our department). Compared with conventional MRI, protein-based APTW imaging can improve the accuracy for discriminating between SBMs and GBMs.

On conventional MRI images, although glioma cells infiltrate into the PBZ, the blood-brain barrier does not break down, and enhancement is not observed in Gd-T1WI [18]. However, the cell density and mobile proteins and peptides in the PBZ increased; in the PBZ of GBMs, the APTw value was higher closer to the tumour core than the APTw value in the CNAWM; thus, the APTw MRI, which is based on amide protons, may provide new insights into the evaluation of the extent of GBM cell infiltration in the PBZ on a molecular level.

However, this study has several limitations. First, the sample size of the metastasis group mainly involves the metastasis of lung cancer. A larger sample size that includes metastases of more primary cancers is required for more reliable results. Second, the APT-weighted sequence covered only one slice, and the MRI signal changes in other tumour regions could not be evaluated in the present study. Third, in the current study, the voxel size of the APT-weighted sequence was relatively large (due to the small APT effect), which may have led to homogeneous APTw hyperintensities due to the large partial volume effect.

In conclusion, APT-weighted MRI can differentiate between solitary brain metastases and glioblastoma, which indicates that comparisons in APTw parameters in the peritumoral oedema may be useful for differentiating between SBMs and GBMs. The APTwmin has the highest sensitivity, specificity and accuracy. As a safe and completely non-invasive imaging technique, APT-weighted MRI could be a supplement to improve the preoperative diagnosis of SBMs and GBMs in the clinic.