Evaluation of realistic layouts for next generation on-scalp MEG: spatial information density maps

Abstract

While commercial magnetoencephalography (MEG) systems are the functional neuroimaging state-of-the-art in terms of spatio-temporal resolution, MEG sensors have not changed significantly since the 1990s. Interest in newer sensors that operate at less extreme temperatures, e.g., high critical temperature (high-T c) SQUIDs, optically-pumped magnetometers, etc., is growing because they enable significant reductions in head-to-sensor standoff (on-scalp MEG). Various metrics quantify the advantages of on-scalp MEG, but a single straightforward one is lacking. Previous works have furthermore been limited to arbitrary and/or unrealistic sensor layouts. We introduce spatial information density (SID) maps for quantitative and qualitative evaluations of sensor arrays. SID-maps present the spatial distribution of information a sensor array extracts from a source space while accounting for relevant source and sensor parameters. We use it in a systematic comparison of three practical on-scalp MEG sensor array layouts (based on high-T c SQUIDs) and the standard Elekta Neuromag TRIUX magnetometer array. Results strengthen the case for on-scalp and specifically high-T c SQUID-based MEG while providing a path for the practical design of future MEG systems. SID-maps are furthermore general to arbitrary magnetic sensor technologies and source spaces and can thus be used for design and evaluation of sensor arrays for magnetocardiography, magnetic particle imaging, etc.

On-scalp MEG SQUIDs are sensitive to early somatosensory activity unseen by conventional MEG

“…Hence, while low-T c SQUIDs (hereafter called in-helmet sensors) demonstrate noise levels of ~3 fT/√Hz, the very best high-T c sensors show noise levels of ~10 fT/√Hz (Faley et al, 2006) and often near 50 fT/√Hz (Pfeiffer et al, 2019b). These differences in noise levels mean that a trade-off might be expected between on-scalp and in-helmet sensors, with on-scalp sensors promising the highest gain in signal-to-noise ratio (SNR) for superficial sources (Iivanainen et al, 2017;Riaz et al, 2017).…”

Detection of interictal epileptiform discharges: A comparison of on-scalp MEG and conventional MEG measurements

“…Today, characterization of functional connectivity within a small region is impossible using EEG or in-helmet -MEG (Schoffelen and Gross, 2009). However, it is possible that the improved source separation and neural signal amplitude of the on-scalp MEG measurement (Boto et al, 2016;Riaz et al, 2017) would allow not only identification of such subunits, but also characterization of network dynamics. These are of course issues that need to be further explored in future on-scalp MEG measurements on epilepsy patients.…”

“…Spatial resolution depends on the sensor spacing. A smaller sensor-to-sensor distance results in a better ability to distinguish between neural sources, compared to a greater sensor-to-sensor distance (Boto et al, 2016;Riaz et al, 2017). To address both the problems of sensor-cortex distance and that of the fix in-helmet system, as well as improve both neuroscientific and clinical applicability of MEG, systems where the sensors are flexibly placed directly on the scalp are under development (Borna et al, 2017;Boto et al, 2018;Iivanainen et al, 2019;Pfeiffer et al, 2019).…”

“…On-scalp MEG sensors comprise, amongst others, optically-pumped magnetometers (OPMs) (Budker and Romalis, 2007) and high-Tc SQUIDs (Zhang et al, 1993). Both of these on-scalp MEG sensor systems allow for a significant reduction of the sensor-cortex distance, as well as a rearrangement of the sensor layout geometry, thus increasing the signal-to-noise ratio and spatial resolution of the recorded neuronal activity (Boto et al, 2016;Iivanainen et al, 2017;Riaz et al, 2017;Schneiderman, 2014). Furthermore, placing the sensors evenly distributed on the scalp enables a more even sampling of brain regions.…”

A 7-channel high-Tc SQUID-based on-scalp MEG system

“…This can be partially mitigated by the signal gain and different spatial sampling that stem from coming closer to the head. As such, it is possible to obtain more information from the brain with high-T c vs low-T c SQUID technology, despite the inferior sensor noise levels of the former [7,8].…”