Reducing motion artefact in high resolution 7T scans using a new head stabilization 'MinMo' device

2306 

Abstract Submission 

Jyoti Mangal¹, Simon Richardson², Yannick Brackenier¹, Raphael Tomi-Tricot^1,3, Fred Dick², Martina Callaghan⁴, David Carmichael¹

¹School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, ²Division of Psychology and Language Sciences, University College London, London, United Kingdom, ³MR Research Collaborations, Siemens Healthcare Limited, Frimley, United Kingdom, ⁴Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College, London, United Kingdom

Jyoti Mangal  
School of Biomedical Engineering and Imaging Sciences, King's College London
London, United Kingdom

Simon Richardson  
Division of Psychology and Language Sciences, University College London
London, United Kingdom

Yannick Brackenier  
School of Biomedical Engineering and Imaging Sciences, King's College London
London, United Kingdom

Raphael Tomi-Tricot  
School of Biomedical Engineering and Imaging Sciences, King's College London|MR Research Collaborations, Siemens Healthcare Limited
London, United Kingdom|Frimley, United Kingdom

Fred Dick  
Division of Psychology and Language Sciences, University College London
London, United Kingdom

Martina Callaghan  
Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College
London, United Kingdom

David Carmichael  
School of Biomedical Engineering and Imaging Sciences, King's College London
London, United Kingdom

High-resolution quantitative brain MRI is important for neuroanatomical research and its alteration in pathology[1]. Both long scan durations and ultra-high field facilitate its acquisition, but they increase scan sensitivity to motion causing visible artefacts. Many methods reducing the impact of motion have been proposed[2], however at high field, motion remains a challenging problem due to interactions with the B0 field[3]. We therefore aimed to limit the occurrence of motion, thereby reducing motion artefacts, during long-duration high resolution scans using a head stabilisation device called the MinMo[4] (Fig 1). This was tested using long (10min) and very long (20min) scan durations using (A) linear and (B) random-checkered DISORDER sampling patterns[5] respectively, the latter to enhance motion sensitivity. Our main aim was to evaluate the effect of the MinMo on image quality visually and quantitatively using normalised gradient squared (NGS)[6] within the brain.

Fig. 1(a) shows the schematic diagram of the unloaded MinMo device with components of the head frame labelled on it for the closed configuration.

Data acquisition: To determine whether the MinMo effectively reduced motion artefact, we obtained whole-brain data at resolution 0.6mm³ using an optimized multi-echo GRE (MEGRE) protocol with TR=30ms. Data was acquired in 5 healthy volunteer (HV) subjects. For each HV, 2 sets of 2 MEGRE scans were obtained. The overall acquisition scheme is shown in Fig.1(b). For one of the sets, the HV was loaded in the MinMo device, and for the other set, conventional foam pads following standard radiographer practice were used. In each set, one long scan was acquired with linear cartesian sampling IPAT=2x2, time of acquisition Tacq=10:40[min:s], and one, longer scan was acquired with random-checkered DISORDER sampling pattern in the phase-encoding direction, IPAT=2x2, phase/slice oversampling=0.44/0.41 with Tacq=21:30[min:s]. The HVs were asked to lie still and were all accustomed to volunteer for MR research studies. Reference low-resolution (6mm³ iso) fully sampled scans were acquired each time the HV was loaded/unloaded in the device and were used in the reconstruction step to compute the coil sensitivity profiles using ESPIRIT[7]. All acquisitions were done on the 7T scanner (MAGNETOM Terra, Siemens Healthcare) 32ch receiver, 8ch PTx.

Reconstructions were done on down-sampled data (1mm³) for all HVs using the conjugate-gradient SENSE[8] method. Normalised gradient squared, one of the metrics with the closest correspondence to visual image quality assessment, within whole brain extracted volumes[9] was calculated. Image quality was compared between the MinMo Off and MinMo On cases for both the long-duration scans. MATLAB2023b was used for all reconstructions and analysis.

Fig. 2(A) and (B) shows representative zoomed-in images for the 5 HVs reconstructed for the long scan and the very long scan with MinMo Off and On. Fig. 2(C) and Fig2(D) shows the bar plots for the NGS values calculated on the brain volumes shown in Fig. 2(A) and (B) respectively. For most cases, using MinMo improved the image quality (given by increase in NGS). The p-value for NGS done using a paired t-test between MinMo On and Off cases was reported to be 0.0042 (significant).

Preliminary investigation suggests that using the MinMo device improved image quality (increased NGS) in compliant subjects during most of the long duration (10-20 minutes) scans obtained. We used a 10-minute scan duration as it is a typical duration used commonly in MPM studies[10]. The ~20min scan was obtained aiming to maximise motion sensitivity owing to its long duration as well as motion sensitised phase encoding ordering scheme. Motion correction techniques such as aligned-SENSE[11] can be used to aid motion correction but were not yet unexplored.

Motion Correction and Preprocessing ²

Anatomical MRI ¹

Acquisition

HIGH FIELD MR

MRI

MRI PHYSICS

STRUCTURAL MRI

Other - Quantitative MRI; Motion Correction; High resolution MRI