To assess time-series SNR and residual aliasing levels in accelerated EPI acquisitions at 7 T using different autocalibration signal (ACS) acquisition schemes for GRAPPA reconstruction in a task commonly used in fMRI for presurgical planning which elicits head motion and changes in ΔB₀.The conventional means to acquire the autocalibration data required for accelerated EPI with the GRAPPA parallel imaging reconstruction is either in a single shot (SS) or, for high GRAPPA factors and resolution, in a ‘consecutive-slice’ segmented (‘multi-shot’ or MS) scheme. This has been shown to be sensitive to motion and ΔB₀ changes due to respiration¹. Gradient-echo (FLASH)-based² and ‘consecutive-segment’ (FLEET)¹ ACS acquisition schemes have been shown to be temporally more robust, but to our knowledge have only been tested with resting-state fMRI. Some tasks used in preoperative functional localization of motor and speech areas^3,4 induce motion and ΔB₀ changes during the time series which may result in reduced time-series SNR (tSNR) (and consequently BOLD sensitivity), and enhanced residual aliasing (i.e. ghosting). The ACS acquisition schemes could be expected to perform differently in such circumstances. In this study, we assess tSNR and ghost levels using SS, MS, FLASH and FLEET ACS schemes in fMRI acquisitions with a chin task.7 T fMRI was acquired in 7 healthy volunteers with 4 ACS schemes (SS, MS, FLASH² (acquired with Siemens WIP #676b) and FLEET¹) while performing 20 runs (5 with each scheme; 56 volumes per run) of a chin task in a ABABABA (A: rest; B: task) block design, consisting of opening and closing the mouth with a target frequency of ~0.5 Hz. Twelve resting state (rs) runs (3 with each scheme; 40 volumes per run) were additionally acquired in 4 of the participants. Participants were instructed to stay still during rest and ACS acquisition periods. Image resolution was 1.5 mm isotropic, GRAPPA 2 and TE/TR=23/2500ms. Forty-four reference lines were acquired in each protocol. Five dummy excitations were acquired before the FLEET ACS acquisition and a FLEET flip angle of 10° was used. Functional runs were slice timing and motion corrected with FSL. Time-series SNR and Nyquist ghost-to-signal levels were calculated for both rs and chin tasks in all 4 acquisition schemes, registered to the first volume of the middle run with FSL, and averaged over runs. tSNR values were all averaged over the brain and ghost levels were assessed by calculating the ratio of the average signal in an ROI containing the aliasing area to the average signal of the brain image. Average tSNR and ghost levels were calculated across volunteers in all 4 schemes for both rs and chin tasks, and differences across acquisition schemes were assessed via Wilcoxon paired t-tests at p<0.05. tSNR and ghost group results were registered to a 7 T EPI template in MNI space with FSL.tSNR and ghost level group results are shown in Figure 1. Higher average tSNR values were obtained for rs than in the chin task for all acquisition schemes. FLASH and FLEET yielded the highest average tSNR values in rs and in the chin task. There were no significant differences in tSNR between ACS schemes in rs. In the chin task, however, there were significant differences between all pairs of schemes other than between FLASH and FLEET: {SS-MS: p=0.03; SS-FLASH: p=0.02; SS-FLEET: p=0.04; MS-FLASH: p=0.02; MS-FLEET: p=0.02; FLASH-FLEET: p=0.09}. Lower but similar average ghost levels were found in rs for FLASH and FLEET but these differences were not significant across methods in both rs and chin tasks.tSNR values were higher in the chin task with FLASH and FLEET ACS than with SS and MS. This effect was significant, while the slightly better performance of FLASH and FLEET in the rs was not significant, suggesting that it is important to assess the performance of parallel imaging reconstruction stability in a task which elicits motion as well as in the resting-state acquisitions. There was no significant difference in residual aliasing in both rs and chin tasks, suggesting that all schemes yield a similar level of ghosting in a motion-eliciting task.FLASH and FLEET acquisitions yielded significantly higher average tSNR values compared to SS and MS in the chin task and no significant residual aliasing enhancement in both rs and chin tasks. Since BOLD sensitivity is proportional to tSNR, this study demonstrates the importance of using acquisition methods in fMRI that are robust to motion in applications that use motion-evoking paradigms (e.g. presurgical planning for localization of motor or speech areas), whilst keeping the residual aliasing at an acceptable level.