Correct differential diagnosis of different types of brain tumors is crucial for treatment planning, various anatomical and functional imaging methods have been proposed and applied for such purpose. However, the differentiation between metastasis tumor and glioblastom remains challenging and is much needed due to the prevalent and aggressive nature of glioblastom. These two types of tumors often have similar appearance under conventional contrasts, however a distinctive nature of glioblastom is the infiltrative growth in the surrounding tissues [1,2]. In this work, we use the measure of spatially matching 3D arterial spin labeling (ASL) and 3D amide proton transfer (APT) in the lesion proximal regions to differentiate metastasis and glioblastom. 3D CEST was implemented using the same spiral readout as 3D ASL, as illustrated in Fig.1, which allows spatially matching perfusion and APT acquisition to be made. The infiltration of glioblastom to the surrounding tissue may be illustrated using a MRS experiment: for metastasis tumor, MRS of in the lesion region shows elevated level of Cho as compared to NAA as expected, however such observation was not made in the MRS of the surrounding edema region (Fig.2a); whereas for glioblastom, MRS of both lesion and lesion proximal regions show elevated Cho level indicating the existence of tumor mass. The hypothesis is hence that the lesion proximal region may have different behaviors under perfusion and APT imaging. Total of 30 patients suspected with brain tumors were enrolled in this study with prior consent form obtained. In this cohort, 7 patients had confirmed metastasis tumor whereas 8 patients had confirmed glioblastom. Spatially matching T1 enhanced, T2, 3D ASL and 3D APT were obtained. Measurement of CBF and APT were performed in both the lesion and lesion proximal regions. The lesion region was defined by the enhanced region on T1 enhance image, whereas the proximal region was defined by the edema regions surrounding the lesion on T2 image (Fig.3). If no obvious edema was present, surrounding normal appearing tissue (within 0.5cm) was used as lesion proximal region. To reduce bias of varying individual CBFs, normalized CNF measure by taking the ratio of CBF of target region to that of cerebellum was used. Asymmetrical MTR was calculated for the quantification of APT at 3.5ppm.A representative set of images for metastasis tumor and glioblastom is shown in Fig.3. The T1 enhanced image and T2 image allowed the lesion and lesion proximal regions to be defined (contoured in white and red). Several observations can be made: firstly, in both cases high perfusion and high APT was observed in the lesion region (white) as expected; secondly, hypo-perfusion was mostly present in the lesion proximal region for metastasis tumor, whereas relatively higher perfusion was observed in in the lesion proximal region for glioma; thirdly and similarly, much lower level of APT was observed in the lesion proximal region for metastasis tumor, whereas relatively higher APT was seen in the lesion proximal region for glioma. In Fig.4, the overall perfusion and APT measurement for metastasis tumor and glioma are plotted. It is seen that although measurement in the lesion regions were mostly overlapping for the two types of tumor, but more distinctive signal levels were present in the lesion proximal regions, especially for APT measurement (mean 0.7% vs mean 1.8%). Both ASL perfusion [3] and APT [4] are sensitive means to detect tumor, and they are based on different foundations of detecting angiogenesis and cancer cell. Conjoint use of the two measure may improve the specificity in tumor differentiation. In this work, perfusion and CEST were applied in probing the abnormality in lesion proximal regions to differentiate metastasis from glioma, for glioma may infiltrates the surrounding tissues whereas metastasis has biological clear boundary. This approach was seen to be promising based on preliminary results, and especially APT measure showed quite distinctive measures for the two types of tumors. However, conclusive points are limited by the small sample size at current stage, due to the lack of pathological evidence for many patients enrolled. Further methodological improvement can be made by refining the quantification means used, such as considering the ratio between signal in lesion and lesion proximal regions.