Our aim was 1) to comparatively assess the response to visual and motor activation in phase- and quantitative susceptibility data (fQSM^{1, 2}) versus in traditional-BOLD-fMRI data and 2) to use angiography/venography data to selectively exclude regions of draining-veins for better identification of activated brain tissue. We also tested whether fQSM data allows identification of the “visual-(resting-state)-network”.MRI 2D gradient-echo-EPI (TR=3s, TE=25ms, FA=85, SENSE=3.5, voxel-dimensions=1.25, 1.25, 1.3mm, recon size=176x176x34 (Volunteer-2: 176x176x38)) images of two consenting volunteers were acquired on a 7T-MR-system(Philips, Achieva). T1w- Anatomical-Scan: 3D-inversion-recovery gradient-echo, TR=8.2ms, TE=3.79ms, FA=8, voxel-dimensions=0.94x0.94x1mm, Phase-Contrast-Angiography: 3D-gradient-echo, TR=35.9ms, TE=7ms, FA=10, voxel-dimensions=0.898x0.898x0.9mm. Paradigm³ The quarter-fields of the visual-cortex were stimulated via 16s of flickering, started with 9s of fixation; 10 blocks of upper-left/lower-right (ULLR) and upper-right/lower-left (URLL) color-changing wedges were presented over 200 scans. Subjects’ attention was assured by a simple button-response task to any contrast alteration of the fixation point. Phase-Data-Processing Time series data were first registered to the MNI-template. Phase data were unwrapped via the Laplacian-approach⁴, and background-field-corrected via V-SHARP⁵ (Rmax=10vx, threshold=0.1). Quantitative susceptibility maps were generated by dipolar-inversion of the corrected phase maps using the LSQR algorithm⁶. Functional-Phase-Processing The sign of QSM data was inverted, to match the sign of an activation change with that in BOLD-fMRI, and the minimum value over all-time series both for QSM and SHARP data was subtracted from each pixel for compatibility (non-negative values only) with SPM12⁷. SHARP and QSM images were normalized to MNI-space with a 1.8mm isotropic voxel-size (87x105x76 voxels) and further smoothed with a 2.5mm-Gaussian-kernel. Each data set was subject to GLM analysis with the Canonical Hemodynamic Response Function (HRF) as basis function. The phase-data processing was repeated for different SHARP threshold values (0.04, 0.06, 0.08, 0.1 and 0.12). For the QSM reconstruction, the SHARP data set with the largest number of activated voxels was used. A temporal-bandpass-filter, 0.01-0.11 Hz, was applied to the fQSM data, and results were compared with those obtained without the filtering (Fig.2). The statistical analyses were repeated for fQSM data after masking blood-vessels (Fig.3c). Grey-Matter-Parcellation was done with the T1w-anatomical images using FreeSurfer⁸ (Fig.4a). Resting-State-Analysis Resting-state experiments were performed with identical sequence parameters as the task-experiment and Independent-Component-Analysis (ICA) was applied to fQSM and BOLD-fMRI data9.Image data of volunteer-1 at various processing steps and ULLR-activations-maps overlaid on T1w-Anatomy are shown in Fig.1. Regions of apparent activations in the motor cortex of fQSM and BOLD-fMRI data sets are shown in Fig.2. For volunteer-2, SHARP, fQSM and temporally-filtered fQSM activation maps overlaid on EPI-magnitude data are shown in Fig.3. SHARP data obtained with a threshold of 0.1 gave the largest number of activated voxels for both visual-tasks (Fig.3, insert top right). There were more ULLR-activated voxels in QSM- than in SHARP-derived maps (Fig.3a-b), and in band-pass filtered fQSM [0.01-0.1 Hz] than in non-filtered fQSM (Fig.3b, c). Fig.4 a) and b) comparatively show fQSM ([0.01 0.1 Hz]) and BOLD-fMRI-derived activation maps for both visual tasks. The temporal signal evolution during the visual-tasks showed similar trends for SHARP and fQSM (Fig.3 – green plots), vs. BOLD magnitude data (Fig.4). The relative magnitude variations were approximately 8% for BOLD-fMRI, 2% for fQSM in veins, and 1.5% for fQSM in GM. The vessel mask applied to the QSM data prior to spatial smoothing is shown in Fig.4c). The blocking of larger veins from the activation analysis did not eliminate all areas of apparent fQSM activation, in particular, many regions in seemingly excellent alignment with gray-matter structures survived the masking, as, is evidenced by comparison with gray-matter structures in the visual cortex identified by GM-parcellation (Fig.5a). Results of the resting-state IC-analysis of the fQSM data and the correspondingly identified visual network are shown and compared to results the visual-task (Fig.5b-c). The spatial mismatch between the location of actual brain-tissue neuronal activation and corresponding electric-signal-sources and the location of activation detection by traditional-BOLD-fMRI is an inherent weakness of fMRI, which, in part, has physiological causes, such as down-stream displacement of venous-blood with activation- and neurovascular-coupling induced increased oxygen-saturation levels in combination with low temporal resolution. However, beyond physiological causes, the mismatch is exacerbated by distant field and field-gradient variations provoked by local saturation and susceptibility changes, which affect distant signal phase- and T2*-weighted signal-magnitude changes in non-activated-volumes. We believe that the results presented here support the hypothesis that fQSM may significantly contribute to alleviation of such latter problems and, besides yielding, from the same data-set, information complementary to that from BOLD-fMRI, may even have the potential to become instrumental in new research areas, e.g., cortical-layer specific function imaging¹⁰.