Glioblastoma multiforme (GBM) and anaplastic astrocytoma are aggressive malignant primary-brain-tumors characterized by abnormal tumor vasculature. Inadequate tumor blood flow and diffusion-related limitations in oxygen supply results in hypoxic tumor-tissue subvolumes, whose response to conventional radiotherapy¹ and chemotherapy is typically poor. Early identification of hypoxic tumor areas in MRI may support therapy planning and monitoring. The administration of hyperoxic gas mixtures changes blood-flow, blood-volume, blood oxygenation, and concentration of dissolved-oxygen in tissues². The magnetic-susceptibility difference between diamagnetic oxyhaemoglobin and paramagnetic dissolved oxygen is expected to allow quantification of such changes via Quantitative-Susceptibility-Mapping (QSM), which attempts to derive the spatial distribution of local magnetic-susceptibility differences. In this pilot study the response of primary-brain-tumors to hyperoxic respiratory challenges was quantified using QSM.MRI Gradient-multi-echo images (FA=50, TE1=5.8ms, TE5=69.8ms, ΔTE=16ms, TR=74ms, voxel dimensions=0.87, 0.87, 2mm, matrix size=256x256x55) of three consenting primary malignant brain tumor patients were acquired on a 3T-MR-system. The patients inhaled medical-air (21% O₂), oxygen (100% O₂) and carbogen (95% O₂, 5% CO₂) gases. In previous clinically indicated MR-examinations, contrast-enhanced T1-weighted 3DMPRAGE sequence (FA=9, TE=2.6ms, TR=1670ms, voxel dimensions=0.49, 0.49, 0.9mm) images were acquired after administration of Gadolinium. QSM Multi-echo phase data was combined assuming a linear phase evolution³, followed by Laplacian unwrapping and background field removal (threshold 0.05)^{4, 5}. Quantitative-susceptibility-maps were generated by dipolar inversion of the corrected phase maps using an LSQR algorithm⁶. Susceptibility-difference values were referenced versus white-matter, because it is minimally affected by the hyperoxic challenges³. Co-registration Phase and magnitude data were coregistered to MNI-space via SPM12⁷, followed by registration of corresponding MPRAGE to 1^st echo magnitude data. The placing of regions of interest (ROI) on the images was strongly guided by contrast enhanced T1w-images for vital tumor tissue and T2w-images for edema and necrotic areas. Mean ROI susceptibility values measured under the different breathing regimes were tested for significant differences (paired two-tailed t-tests) between two groups showed same trends.All patients tolerated the exams well. In one patient, the front-part of the head-coil did not fit over the breathing mask, and was therefore not used, which resulted in lower image quality. Figures1-3 show images of Patients#1-3, respectively. The captions give information on tumor type, location of the lesion, and clinical outcome. Five lesions were evaluated in tumor areas (Group-1, Patient#1: Lesions-1-2, Patient#2: Lesion-3, Patient#3: Lesions-4-5) , with positive magnetic-susceptibilities (paramagnetism), and there was a significant reduction upon oxygen inhalation (-0.032±0.022ppm, p<0.031, range: -0.01 to -0.067 ppm), and an even larger decrease under carbogenic hyperoxia (-0.059±0.035ppm, p<0.021, range -0.04 to -0.1 ppb) (Fig.5). The additional susceptibility decrease from O₂ to carbogen inhalation in these lesions was -0.02±0.016ppm (p<0.025). In contrast, in three additional analyzed regions, assigned to edema (n=2) and necrosis (n=1) (Group-2), the magnetic-susceptibility moderately increased under O₂-hyperoxia (+0.016±0.003 ppm, p<0.011) and more strongly rose under carbogen-breathing (0.021±0.006 ppm, p<0.029). The additional susceptibility increase from O₂ to carbogen inhalation in the edematous and necrotic lesions did not reach statistical significance. Both QSM and corrected-phase images of Patient#2 depicted regions of high paramagnetism (high susceptibility) not visible on contrast-enhanced T1w-images, and only partially discernible on T2w-images (Fig.4). Higher magnitude of changes were observed for lesions in patient diagnosed with multifocal glioblastoma (Patients#1), while smaller changes were observed in the case of anaplastic astrocytoma (Patient#3). Interestingly, Patient#1, in whose two lesions the carbogen-inhalation induced strongest decrease of magnetic-susceptibility, had 3 months from MRI until tumor progression, whereas in Patient#2, where the challenge induced the smallest susceptibility changes, has not suffered from tumor progression, at the last follow-up after 4 months.An increase of blood oxygen saturation causes a reduction of corresponding apparent magnetic-susceptibility. Regions with a low baseline oxygen saturation but still functional perfusion, i.e., hypoxic regions, have a potential for larger susceptibility decrease than well perfused normoxic tissue. Strong negative susceptibility response to a hyperoxic challenge might be indicative for hypoxic, but not necrotic-tissue, potentially requiring treatment with a higher radiation dose for adequate response. In contrast, we found a small susceptibility increase in less perfused tissue hypothetically associated with larger extracellular volume fraction and higher fluid content that we tentatively attribute to dissolved paramagnetic molecular O₂⁸. The consistently larger magnitudes of the susceptibility changes observed under carbogenic hyperoxia than under O₂-hyperoxia may be rationalized by carbogen-induced vasodilatation and associated with higher tissue perfusion, in addition to the slightly higher arterial O₂ saturation levels. Whereas we believe the results presented here to be highly interesting and intriguing, more work with more patients and a careful clinical follow-up will be required to gauge the potential of QSM under respiratory challenge.