Suppression of the extra cranial lipid signal poses challenges for MRSI at 7T. Aside from restricting coverage to avoid exciting peri-cranial lipids using PRESS localization, most lipid suppression methods are based on either short TI inversion recovery (STIR), or spatially/frequency selective saturation of the lipid resonances followed by their dephasing (1,2). In this work, a surface gradient ‘crusher’ coil (3) was implemented to improve the lipid suppression in a volumetric proton echo-planar spectroscopic imaging (EPSI) sequence design at 7T.All experiments were performed using a Philips Achieva 7T scanner equipped with a dual-transmit system and a 16-channel receive coil. The crusher coil (Fig. 1) placed inside the 16-channel receive coil (3) was interfaced to one of the scanner’s (unused) B0 shim amplifiers which provided a pulsed current of 7A. A single, unbalanced pulse (Fig. 2) applied during the sequence TE was used to generate local inhomogeneous magnetic fields in the outer layer of the head, and therefore dephase the coherence of the lipid signals. Both gradient echo MRI and spectroscopic imaging datasets were acquired. The EPSI sequence adapted from (4) was comprised of 4 CHESS pulses for water suppression (bandwidth 200 Hz), and an inferior saturation band followed by a slice-selective spin-echo with EPSI readout. Phase-encoding was achieved in 2 directions to provide with full 3D coverage. The water signal was acquired by an interleaved, small flip angle (20°) minimum TE gradient-echo EPSI readout. The 80 mm slab excitation was prescribed in an orientation parallel to the anterior-posterior commissure line. Scan parameters include TR/TE 1710/35 ms, acquisition matrix 50x50x18, FOV 280x280x180 mm3, nominal voxel 5.6x5.6x10 mm3. The spectral bandwidth for EPSI readout was 2466 Hz. Sequence SAR was < 46% of the FDA limit. The EPSI scan time was 25 min. All spectra were processed using MIDAS (5). Two methods of lipid suppression were compared: one using a STIR sequence (TI 250 ms), or using the pulsed crusher coil. To calibrate the required depth of crushing, gradient-echo MR scans (Fig. 3) were performed with different duration pulses.Gradient echo images without (Fig. 3A) and with application (Fig. 3B,C) of the crusher coil pulses in a normal volunteer (age 32 years, male) show the lipid suppression performance. With a 2 ms pulse (Fig. 3C), it can be seen that there is appreciable suppression of cortical brain tissue, and it was determined that a pulse duration of between 0.5 to 1.0 ms gave optimal performance. Three EPSI spectra at 7T in a normal volunteer (age 35 years, female) are displayed in Figure 4 from both central and peripheral voxel locations using a 0.75ms crusher pulse duration. Lipid contamination was reduced by a factor of approximately 35 compared to the scan without lipid suppression. Well-resolved peaks from NAA 2.01 ppm, tCr 3.03 ppm, tCho 3.19 ppm and tCr 3.92 ppm are observed with good SNR from the nominal 0.3 ml voxels. Field inhomogeneity was less than 20 Hz over the whole volume of the EPSI slab. With respect to no lipid suppression, STIR reduces lipid contamination by a factor 1.5, the crusher coil by a factor 35 and both methods combined (STIR + crusher coil) by a factor 72 (Fig. 5).Use of a surface crusher coil for lipid suppression at 7T has a number of advantages compared to conventional lipid suppression methods; no RF pulses are involved, so sequence SAR is reduced. There is no dependence of lipid T1 (unlike the STIR method) and also no dependence on variations in RF pulse flip angles (which can affect both STIR and OVS methods). Unlike the IR method, brain magnetization is also unaltered, so long as the crusher pulse is properly calibrated. The minimum sequence TR is also reduced somewhat, since there is no TI (or OVS) time period, the crusher pulse being applied during the sequence TE. The crusher coil does require additional hardware and has to be placed inside the receive coil array; however, with correct design, the coil has little effect on both transmit B1 level or receiver coil SNR. The crusher coil also has fixed geometry, which, in combination with the varying head sizes and shapes, may result in variable performance in different subjects. Therefore it is important to carefully position each subject within the coil, and run rapid gradient-echo images in order to calibrate the optimum pulse duration and/or amplitude. Crusher coils have previously been implemented for conventional single-slice MRSI at 7T (3); this abstract demonstrates that the method can be readily adapted to use for volumetric EPSI at 7T with extended brain coverage.