Spatiotemporal signal space separation for regions of interest: Application for extracting neuromagnetic responses evoked by deep brain stimulation

Abstract

Magnetoencephalography (MEG) recordings are often contaminated by interference that can exceed the amplitude of physiological brain activity by several orders of magnitude. Furthermore, the activity of interference sources may spatially extend (known as source leakage) into the activity of brain signals of interest, resulting in source estimation inaccuracies. This problem is particularly apparent when using MEG to interrogate the effects of brain stimulation on large-scale cortical networks. In this technical report, we develop a novel denoising approach for suppressing the leakage of interference source activity into the activity representing a brain region of interest. This approach leverages spatial and temporal domain projectors for signal arising from prespecified anatomical regions of interest. We apply this denoising approach to reconstruct simulated evoked response topographies to deep brain stimulation (DBS) in a phantom recording. We highlight the advantages of our approach compared to the benchmark-spatiotemporal signal space separation-and show that it can more accurately reveal brain stimulation-evoked response topographies. Finally, we apply our method to MEG recordings from a single patient with Parkinson's disease, to reveal early cortical-evoked responses to DBS of the subthalamic nucleus.

Keywords: magnetoencephalography; source leakage correction; spatiotempotal signal separation.

FIGURE 1

The effect of ROI size on signal recovery for spherical and cubic ROIs. Upper panel: (a) highlights the setup of the simulation with 100 randomly oriented sources on the surface of the blue sphere at a distance of 0.04 m from the expansion origin. The sensor distance from the origin, R is fixed at 0.08 m. Grey spheres represent spherical boundaries of the ROIs that were tested (b) the effect of ROI radius on signal power and cumulative signal power. When the ROI does not encompass the simulated sources a small proportion of the signal power is recovered and there is little dependence on L. Signal power recovery and the dependence on L increase as the ROI radius approaches R. Lower panel (c and d) is as per upper panel, except for a cubic rather than a spherical ROI. For cubic ROIs, we selected the dimensions of the side such that the body diagonal was equal to the diameter of the corresponding sphere in (a).

FIGURE 2

Application of spatiotemporal signal separation approaches to reconstructing evoked responses in a phantom recording. Left: Topographies of the simulated dipole at the four different DBS settings (no stimulation, 5 Hz monopolar DBS, 20 Hz monopolar DBS and 130 Hz monopolar DBS), after pre‐processing the data with one of four different approaches (standard pre‐processing, tSSS, ROI‐tSSS sphere and ROI‐tSSS cube) are shown. The colour bars represent field strength measured in femtoteslas (fT). Plots to the right show the reconstructed time courses of the simulated sinusoid, for each DBS setting and each pre‐processing approach. The ROI‐tSSS sphere and ROI‐tSSS cube approaches reproduce the dipole topography well across all stimulation conditions.

FIGURE 3

The mean squared error (MSE) of the simulated dipole topography, computed between different DBS conditions and the no stimulation condition is plotted for all four pre‐processing approaches in the phantom recording. The lowest MSE across all stimulation conditions is achieved by the ROI‐tSSS sphere and the ROI‐tSSS cube approaches. The ROI‐tSSS approaches offer the greatest benefit at clinically deployed DBS frequencies of 130 Hz.

FIGURE 4

(a) The ROI‐tSSS approach is applied to detect the evoked response to monopolar 5 Hz DBS of the right subthalamic nucleus in a patient with Parkinson's disease. The upper panel shows the sensor level evoked response (averaged across trials and channels after baseline correction relative to a 20 ms window [−0.05 to −0.03 s] prior to the onset of the stimulation pulse at time 0) after constructing a 5 cm region of interest centred on the right motor cortex at MNI co‐ordinates 37–18 53; the lower panel shows the corresponding topography of the evoked response peaks at 2.5 and 4.2 ms. (b and c) Right hemispheric source level evoked response amplitudes at the times of the two peaks are extracted using LCMV beamforming and projected onto a cortical mesh after applying tSSS in (c) and ROI‐tSSS in (d). In (d), for each grid point, the ROI‐tSSS sphere algorithm (with a 5 cm radius) was applied prior to beamforming. There are focal frontal, temporal and parieto‐occipital regions demonstrating high‐amplitude evoked responses.