Real-time Multislice-to-volume Motion Correction for Task-based Functional MRI at 7T

Winata, S. , Hoinkiss, D.C., Keith, G.A. , Al-Wasity, S. and Porter, D.A. (2023) Real-time Multislice-to-volume Motion Correction for Task-based Functional MRI at 7T. 39th Annual Scientific Meeting of the European Society for Magnetic Resonance in Medicine and Biology (ESMRMB 2023), Basel, Switzerland, 4-7 October 2023. pp. 65-66.

Full text not currently available from Enlighten.


Introduction 7 T MRI has higher inherent signal-to-noise ratio (SNR) than standard clinical field strengths that gives the potential for higher resolution imaging. The downside is further susceptibility to ghosting and blurring artefacts even to minimum motion. This is especially pronounced in longer acquisitions, where the effects of incidental motion are more visible [1]. Generally using long acquisitions and high resolutions, functional MRI (fMRI) protocols are prone to motion artefacts, which are typically corrected with retrospective motion correction [2]. Prospective, real-time motion correction can be performed to reduce these effects further [3]. The restricted environment in 7 T scanners (i.e. tighter head coils, narrower and longer bores) makes markerless, non-hardware techniques a compelling option. This abstract presents an implementation of the markerless Multislice Prospective Acquisition Correction (MSPACE) technique for 7 T task-based fMRI. This includes the use of the in-plane generalised autocalibrating partially parallel acquisitions (GRAPPA) [4]. This technique reduces spatial distortion effects in higher field strength EPI [5]. MS-PACE [6] is a prospective motion correction technique adapted from Prospective Acquisition CorrEction (PACE). In-plane and through-plane motion are estimated by registering a subset of equidistant 2D-EPI slices to a reference volume, differing from the volumetric registration in PACE [7]. This allows for sub-TR motion detection and higher temporal resolution of imaging system update. This method has previously been implemented at 3 T [8]. Methods The study was performed in a MAGNETOM Terra 7 T scanner (Siemens Healthineers, Erlangen, Germany) with a 1Tx32Rx head coil (Nova Medical, Wilmington, MA, USA) using an in-house developed GRE-EPI sequence on 10 healthy subjects (age 31 ± 9). The GRE-EPI fMRI protocol consisted of 3 scan groups: 2 resting; 2 left hand tapping; 2 right hand tapping. Motion correction was applied to 1 scan per group. The scan parameters were otherwise identical: voxel size 2 9 2 9 2mm3, resolution 96 9 96, GRAPPA factor 3, 60 slices, 110 volumes, echo spacing 580 ms, TR 4 s, TE 18 ms and total acquisition time 7m32s. The tapping scan block design (shown in Fig. 1) was transmitted with the aid of the PsychoPy software package [9]. Fig. 2 shows how the motion detection and correction pipeline operated. A 3-slice registration subset was used. Estimated motion parameters were subsequently fed back to the scanner and the imaging gradients were updated to account for these. The correction robustness was evaluated by retrospectively calculating rigid body motion parameters (3D translation and rotation) with the multislice to-volume method. Online and offline processing was done within the Image Calculation Environment (Siemens Healthineers, Erlangen, Germany) using ITK open-source image registration libraries. Results Fig. 3 compares the mean voxel displacement from each scan group across all subjects. Fig. 4 displays temporal SNR (tSNR) maps and percentage differences in tSNR (dtSNR) from the scans without and with real-time motion correction in each subject. These maps illustrate the temporal variance in noise and were calculated by comparing the mean signal of each voxel to its standard deviation over the time series. Discussion Fig. 3 demonstrates the consistent ability of the technique to correct for motion across all subjects. The technique worked in subjects with different propensity to move, from those who moved little to those who moved much more. It also was able to correct for head motion during the tapping scans, which are more inherently motion prone to incidental motion from the hand movements. Fig. 4 also correlates with these findings. When real-time motion correction is applied, there has been a net positive temporal SNR percentage improvement in a majority of the cohort's subjects. It is important however to note that each acquisition is separate and thus the motion patterns are variable. It has been demonstrated that the technique can correct for longer term motion components in 7 T task-based fMRI consistently across a cohort. Conclusion This study has evaluated an implementation of multislice-to-volume prospective motion correction for 7 T task-based fMRI and shown that the technique can consistently reduce the effects of long-term motion in a motion-propensity diverse cohort of subjects.

Item Type:Conference or Workshop Item
Additional Information:Abstract - available at Magnetic Resonance Materials in Physics, Biology and Medicine 36(Suppl 1): S65
Glasgow Author(s) Enlighten ID:Keith, Dr Graeme and Porter, Professor David and Winata, Mr Steven and Al-Wasity, Mr Salim
Authors: Winata, S., Hoinkiss, D.C., Keith, G.A., Al-Wasity, S., and Porter, D.A.
College/School:College of Medical Veterinary and Life Sciences > School of Psychology & Neuroscience
Research Group:Imaging Centre of Excellence
Related URLs:

University Staff: Request a correction | Enlighten Editors: Update this record