Enabling ambulatory movement in wearable magnetoencephalography with matrix coil active magnetic shielding

Abstract

The ability to collect high-quality neuroimaging data during ambulatory participant movement would enable a wealth of neuroscientific paradigms. Wearable magnetoencephalography (MEG) based on optically pumped magnetometers (OPMs) has the potential to allow participant movement during a scan. However, the strict zero magnetic field requirement of OPMs means that systems must be operated inside a magnetically shielded room (MSR) and also require active shielding using electromagnetic coils to cancel residual fields and field changes (due to external sources and sensor movements) that would otherwise prevent accurate neuronal source reconstructions. Existing active shielding systems only compensate fields over small, fixed regions and do not allow ambulatory movement. Here we describe the matrix coil, a new type of active shielding system for OPM-MEG which is formed from 48 square unit coils arranged on two planes which can compensate magnetic fields in regions that can be flexibly placed between the planes. Through the integration of optical tracking with OPM data acquisition, field changes induced by participant movement are cancelled with low latency (25 ms). High-quality MEG source data were collected despite the presence of large (65 cm translations and 270° rotations) ambulatory participant movements.

Fig. 1.

The matrix coil active magnetic shielding system. (a) System is arranged in a bi-planar geometry and each plane contains 24 square coils arranged in an overlapping grid pattern. The blue and red colours highlight the $4\times 4$ grid of coils and the overlapping $3\times 3$ (excluding the central coil) set of coils. The black dots represent the positions of OPM sensors in a 3D printed helmet placed at the centre of the coils. (b) Photograph of a single plane of the constructed coil system. c) the distributed windings of one plane of a bi-planar coil designed to generate a uniform magnetic field in the x-direction (see (a)) over a $40\times 40\times 40$ cm³ volume at the centre of the two planes with deviation <5% from the target magnetic field. Such designs have been used in previous OPM-MEG experiments. Red and blue denote regions of opposing current flow. The second plane (not shown) features the same windings, but the current directions are reversed with respect to the first plane. (d) The currents applied to one plane of the matrix coil to generate the same field as the coil shown in (c), note the similarity between the two designs in terms of current distribution and direction of current flow. The current directions in the second plane are again reversed.

Fig. 2.

System schematic and nulling procedure flowchart. (a) Overview of OPM-MEG system with matrix coils. The system is housed in a magnetically shielded room. Coil panels are placed either side of a participant wearing a helmet that contains OPMs. A control PC is used to read incoming data from the OPMs and optical tracking cameras then combine these data to calculate and apply nulling currents. A separate PC is used during MEG experiments to provide (in this case) auditory stimulation to a participant via a set of speakers. Trigger channels are used to synchronise the stimuli to the OPM recording. (b) Flowchart of continuous field nulling process.

Fig. 3.

Compensating magnetic field drifts on a fixed array of OPMs. (a) The position of the helmet containing 15 triaxial OPMs with respect to one plane of the matrix coils. The helmet is deliberately positioned off centre in the x, y and z directions to showcase the flexibility of the matrix coil system to generate the required compensation fields at multiple locations between the planes (helmet centre of mass (x,y,z) = (0.15, −0.24, 0.18) m). (b) Magnetic field drifts recorded by the OPMs over a period of 600 s. Each colour represent an individual channel. For the first 300 s the matrix coil system was not active, for the final 300 s the matrix coil compensated changes in the uniform field and field gradients over the helmet. (c) The current applied to each of the 48 coils throughout the experiment. Each colour represents an individual coil current. Note the similarity between the profile of the field changes in the first half of the experiment and the current variations in the second half of the experiment. (d) The power spectral density of the data measured by the OPMs with the coils off (red) and with the coils on (blue). The low-frequency noise level is reduced by the matrix coil, but an increase in noise is seen at higher frequencies. The black dashed line indicates 15 fT/√Hz, the noise floor of the OPMs used in the experiments.

Fig. 4.

Compensating magnetic field changes on a moving array of OPMs. (a) Magnetic field changes recorded by the OPMs over a period of 120 s whilst a seated volunteer made controlled head movements. For the first 60 s the matrix coil system was not active, for the final 60 s the matrix coil compensated changes in the uniform field and field gradients over the helmet. (b) The current applied to each of the 48 coils throughout the experiment. (c) The change in the position of centre of mass of the helmet during the experiment. (d) The rotation of the helmet about the centre of mass during the experiment. Note the size of the movement remains similar throughout the recording, but the amplitude of the artefact is significantly reduced by the matrix coil in the second half of the recording. (e) The power spectral density of the data measured by the OPMs with the coils on (blue) and with the coils off(red). Compensating a stronger and more quickly varying field results in a large artefact at 40 Hz (the rate at which coil currents are updated). The black dashed line indicates 15 fT/√Hz.

Fig. 5.

Dry current dipole phantom and sensor gain experiments. (a) Photograph of the dry-type current dipole phantom used in these experiments. The phantom coil is an isosceles triangle wound around a (grey) plastic former with base 5 mm and height 65 mm. An empty (blue) OPM casing is attached to a (clear) Perspex cylinder into which the phantom coil is glued. (b) Photograph of the phantom inside the OPM-MEG helmet. The empty OPM casing can be inserted into any slot in the helmet, the base of the coil is 30 mm beneath the inner surface of the helmet. (c) Each of the 23 triaxial OPMs (black dots) used in the experiment are shown around the template brain, which was used to design the helmet surface. Red arrows indicate the three channel orientations for each OPM. The slot into which the phantom was inserted is shown in blue. (d) Plot of the field amplitude measured on a single channel (indicated by the green dot and arrow in (c)) due to a 200 nA, 23 Hz phantom signal, Results are shown for four experimental cases. Note that during the experiment in which the helmet moved but the matrix coils were not active a large range of field offsets and phantom signal amplitudes are recorded due to sensor gain changes as the field experienced by the OPM moves away from zero. (e) Zoomed in plot of data for low field offsets (0 - 200 pT) to show consistency of measured source amplitude in the other three experiments.

Fig. 6.

Dry phantom source reconstruction results. The reconstructed source position and orientations are shown for the dipole fitting and beamformer analysis for three cases, black circles show OPM positions and the blue dot highlights the slot into which the phantom was inserted. The first case is when the helmet was static and the matrix coils were switched off, referred to here as the ground truth. In the beamformer analysis, this voxel and approximate source orientation were consistent with the ground truth for all experiments other than trials in positions 2 and 3 during the experiment where the helmet was moving without the matrix coils active. Inset arrows show zoomed in source positions and orientations for comparison between conditions.

Fig. 7.

Range of motion and magnetic field change recorded from a participant undergoing ambulatory movements. (a) Histograms of the change in position of the centre of mass of the OPM-MEG helmet relative to its mean position. The three histograms show left-right, up-down and forward-backward movement, respectively. For all plots, blue and red denote the experiments with and without the matrix coils active. (b) Histograms of the change in orientation of the centre of mass of the OPM-MEG helmet relative to its mean orientation. The three histograms show pitch, yaw and roll, respectively. We note that only a short amount of time was spent at the large pitch and roll orientations, and there is a correspondingly low count rate at the extremes of the corresponding histogram x-axis. (c) Histogram of the change in magnetic field measured by each channel of the OPM array relative to the value at the start of the recording. This shows that the matrix coil significantly decreases field variations which are induced by sensor movement. (d) The power spectral density of the data measured by the OPMs in each recording. Compensating large field changes induced by participant movement results in a further increase in both the broadband noise and the 40/80 Hz coil artefacts. The black dashed line indicates 15 fT/√Hz.

Fig. 8.

Source reconstruction of MEG data collected from a participant undergoing ambulatory movements with and without matrix coils active. (a) Data from trials in which the participant was pressing the button with their right index finger with the matrix coil active. (b) Data from trials in which the participant was pressing the button with their left index finger with the matrix coil active. (c) and (d) show data from right and left-handed trials during the experiments where the matrix coil was off. For each case: (i) Beamformer images of the spatial signature of beta band modulation (thresholded to 80% of the maximum value) shows activity in the contralateral sensorimotor cortices as expected (channel positions (black dots) and orientations (red arrows) shown inset to (a)). ii) Time-frequency spectra of a virtual electrode at the peak of the beamformer image showing movement related desynchronization (blue) in the beta band and post movement beta rebound (yellow). (iii) Power spectral density of the virtual electrodes reconstructed using a triaxial beamformer and a radial-only beamformer. Note how the triaxial beamformer significantly reduces the impact of the 40 and 80 Hz artefacts generated by the matrix coil.