Quantitative susceptibility mapping (QSM) is an emerging post-processing technique that uses gradient echo phase images to calculate the tissue magnetic susceptibility and it was found as a valuable method to characterize diseases such as Parkinson’s disease (PD) ^1,2. It was also shown that Susceptibility Weighted Imaging (SWI) is useful to evaluate neurodegenerative diseases, but susceptibility quantification is hindered by the non-locality and orient dependency of high-pass filtered phase images ^3,4. For retrospective studies, QSM might be obtained from available magnitude and vendor-provided high-pass filtered phase images of SWI. The purpose of this study is to estimate to what extent current QSM values may be affected when using vendor-provided high-pass filtered phase data. We analyzed the influence of different high-pass filtered phase images on the susceptibility assessment and compared the results to quantitative susceptibility estimation using a background field removed phase data.All MRI experiments were performed at 3 Tesla (Siemens Trio, Siemens Medical, Erlangen, Germany) using a 32-channel head-coil. Flow compensated 3D multi gradient echo images were acquired on five volunteers and seven PD patients with the following parameters: TR = 44ms, TEs = 4.6, 11.2, 18.0, 24.8, 31.6, 38.4ms, α = 20°, FOV = 200 x 200mm², matrix = 256 x 256, resolution = 0.75 x 0.75mm², slice thickness = 1.5mm, number of slices = 80. Magnitude and phase data were stored on the scanner. Post processing of the data at TE = 18ms was performed in Matlab (Matlab 2014a, The MathWorks) in the following steps: (1) phase unwrapping, (2) high-pass filtering using different filter strengths (ranging from 0.01 to 0.3 of 3dB cutoff point, (Fig. 1)) and (3) susceptibility estimation using the iLSQR ⁵ method (an algorithm for sparse linear equations and sparse least squares) available from the STISuite toolbox ⁶. As reference, QSM post-processing was performed using a background field removal method (the V-SHARP algorithm ⁷) and susceptibility quantification applying iLSQR. Region of interest (ROIs) were draw in the red nucleus (RN) and the substantia nigra (SN) and the results for the two subject groups were compared for the different high-pass filtered images.Fig. 1 shows the eight different high-pass filter used in this analysis (from h=0.01 (the weakest) to h=0.3 (the strongest)). Fig. 2a shows the original phase, Fig. 2b the phase after phase unwrapping. The purple rectangle in Fig. 2c displays the SN/RN region of the brain for which the results of the remaining sub-figures of Fig. 2 are shown. Fig. 2d-g shows the phase images after different high-pass filter are applied. With increasing filter strength from h=0.025 to h=0.3 the phase is reduced for larger tissues (like SN and RN), the contours of the object remain. A similar trend can be seen for the calculated susceptibility using these high-pass filtered images (Fig. 2h-k). The results of the ROI analysis in SN and RN (Fig. 3a) can be seen in Fig. 3b and c, respectively. The calculated susceptibility values are in both tissues decreasing with increasing high-pass filter strength. While for the RN no significant difference was found between the PD and the control group, the SN showed a significant difference between both groups for susceptibility calculation using the background field removed phase and a high-pass filter up to h=0.05. Fig. 4 shows a comparison between a high-pass filtered phase image provided by the vendor (Fig. 4a) and using a high pass filter h=0.05 (Fig. 4b) on two different subjects. These images appear visually similar, though actual filter parameters used by the vendor are not known.This study evaluated the effect of different high-pass filtered phase images on the outcome of the susceptibility estimation for two different brain tissues (SN and RN) in PD and controls. The analysis showed, that the quantification using high-pass filtered susceptibility estimation resulted in consistently lower susceptibility values, but significant differences in the susceptibility could be detected up to a certain filter strength (here h=0.05 for the SN). For analyzing susceptibility differences in a retrospective study, where only the magnitude and the high-pass filter phase data is available, the high-pass filter strength needs to be evaluated and the effect of the filter needs to be considered.The increasing strength of the high-pass filter results in consistently lower susceptibility values in SN and RN. Up to a certain filter strength (here h=0.05) differences in susceptibility can still be distinguished. However, as the vendor provided images visually appear close to our h=0.05 images, retrospective QSM reconstruction using vendor-provided SWI data will likely yield fair estimate of susceptibility.