Researchers interested in quantitative susceptibility mapping (QSM) and the appropriate choice of reference tissues.Exploiting the quantitative nature of QSM is desired in many applications. Since QSM is only able to quantify magnetic susceptibility in relation to a reference value rather than in absolute terms, a suitable reference tissue has to be found to be able to compare different subjects and stages of disease. Susceptibility values of a reference tissue are ideally independent of age and disease. Previously, in vivo susceptibility maps were referenced to cerebrospinal fluid (CSF) (1,2) or white matter (3,4), e.g. internal capsule (5). In this study, we assess the suitability of 12 regions in the brain to serve as reference region for QSM.25 melanoma patients with brain lesions (32-77 years; mean age 55.3 years) and 25 without brain lesions (28-80 years; mean age 57.4 years) were measured 151 times in total. The patients were staged in each measurement according to the burden of metastatic disease in the brain: 0 for no brain lesion, 1 for full regression due to therapy, 2 for partial regression, 3 for stable lesion status, 4 for progression and 5 for strong progression in size and/or number of lesions. All patients were scanned at a 1.5 T whole-body MR system (Magnetom Symphony, A Tim System, Siemens Healthcare) with a 12-channel head-matrix coil during routine clinical workup with a clinical protocol including a T1-weighted spin-echo (pre- and post-contrast), a T2-weighted TSE, a diffusion weighted EPI, a T2-weighted FLAIR and a fully flow-compensated 3D gradient-echo sequence. Imaging parameters for the 3D-GRE were (pre-contrast): flip angle=15°, TR=49 ms, TE=40 ms, acquisition matrix=320x250x72, voxel size=0.75x0.88x1.9 mm³, readout bandwidth 80 Hz/pixel, partial parallel imaging (GRAPPA) with an acceleration factor R=2 and 24 reference lines. 8 patients with lesions were additionally measured at a 3 T whole-body MR-system (Biograph mMR, Siemens Healthcare) using a similar clinical imaging protocol. Phase images were combined using the vendor-provided adaptive combine method. Brain masks were generated from the magnitude images using FSL-BET (6). Phase images were unwrapped using a Laplacian-based phase unwrapping (1,7,8). The background field was removed with V-SHARP (7,8) (with kernel size up to 25 mm). Susceptibility maps were calculated using the iLSQR method (1,9) and iLSQR-parameters recommended for effective removal of streaking artifacts and accurate quantification of susceptibility were used (9). Susceptibility maps of patients with intra-metastatic bleedings were calculated using superposition (10) to minimize artifacts. 12 regions in the brain (see Fig. 1 and its caption, where region labels are defined) were identified as suitable reference regions and drawn on the magnitude images and susceptibility maps of the first measurement (at the 1.5 T scanner) of each patient using The Medical Imaging Interaction Toolkit (MITK) (11,12). All measurements of each patient were co-registered to this first measurement using affine registration in FSL-FIRST (13).Figure 2 shows that csf_post and ci2 are the regions with the smallest standard deviation from the mean susceptibility value of all patients for each region. The mean susceptibility for csf_post is 0.012±0.014 ppm and for ci2 -0.062±0.016 ppm. All other CSF volumes, csf_ant and csf_sup, show comparably small standard deviations. The regions ci2 and csf_post show virtually no age dependence (Fig. 3a,b), whereas most other regions show a weak trend (Fig. 3c,d). Although there is no clear dependence on disease progression observable, the brain nuclei (Fig. 4c), followed by the regions of the corpus callosum (Fig. 4d) seem to depend more on the used staging than the other regions.Although, white matter is known for orientation dependence of susceptibility contrast (14,15), both ci2 and csf_post fulfill optimal criterions for a reference for QSM of the brain: susceptibility values do not vary strongly between subjects and don’t seem to be age- or disease-dependent regarding spread of metastases in the brain and treatment or metastasis-growth-induced bleedings. Susceptibility values in this study were not or only slightly dependent on the burden of metastatic disease, a fact that may differ in other diseases. Disease-related changes in susceptibility have been observed in the red nucleus (16) and the caudate nucleus (5) in multiple sclerosis (MS). Even susceptibility of CSF may vary in MS as CSF ferritin level changes have been reported in a recent study (17). So far, no cerebral pathologies are known to correlate with iron depositions in the internal capsule (18). Therefore, especially the posterior limb of the internal capsule, but also the atrium of the lateral ventricles may prove to be suitable reference regions for QSM applications in a wide range of cerebral pathologies.