Comparison of beamformer implementations for MEG source localization

Abstract

Beamformers are applied for estimating spatiotemporal characteristics of neuronal sources underlying measured MEG/EEG signals. Several MEG analysis toolboxes include an implementation of a linearly constrained minimum-variance (LCMV) beamformer. However, differences in implementations and in their results complicate the selection and application of beamformers and may hinder their wider adoption in research and clinical use. Additionally, combinations of different MEG sensor types (such as magnetometers and planar gradiometers) and application of preprocessing methods for interference suppression, such as signal space separation (SSS), can affect the results in different ways for different implementations. So far, a systematic evaluation of the different implementations has not been performed. Here, we compared the localization performance of the LCMV beamformer pipelines in four widely used open-source toolboxes (MNE-Python, FieldTrip, DAiSS (SPM12), and Brainstorm) using datasets both with and without SSS interference suppression. We analyzed MEG data that were i) simulated, ii) recorded from a static and moving phantom, and iii) recorded from a healthy volunteer receiving auditory, visual, and somatosensory stimulation. We also investigated the effects of SSS and the combination of the magnetometer and gradiometer signals. We quantified how localization error and point-spread volume vary with the signal-to-noise ratio (SNR) in all four toolboxes. When applied carefully to MEG data with a typical SNR (3-15 ​dB), all four toolboxes localized the sources reliably; however, they differed in their sensitivity to preprocessing parameters. As expected, localizations were highly unreliable at very low SNR, but we found high localization error also at very high SNRs for the first three toolboxes while Brainstorm showed greater robustness but with lower spatial resolution. We also found that the SNR improvement offered by SSS led to more accurate localization.

Keywords: Beamformers; EEG; LCMV; MEG; Open-source analysis toolboxes; Source modeling.

Fig. 1

Simulation of evoked responses. a) The 25 simulated dipolar sources (green dots) in the source space (grey dots), b) Simulated evoked responses of a dipolar source at five strengths and the field patterns corresponding to the peak amplitude (SNR in parenthesis). The dipole was located at (−19.2, −71.6, 57.8) mm in head coordinates.

Fig. 2

MEG data simulation workflow (details in Suppl. Fig. 1).

Fig. 3

The dry phantom. (a) Outer view, (b) cross-section, (c) positions of the employed dipole sources, (d) phantom position with respect to the MEG sensor helmet, and (e) position and rotation of the phantom during one of the moving-phantom measurements (Dipole 9 activated).

Fig. 4

Surfaces that bound the source space used by each toolbox. a) Sagittal, b) coronal, and c) axial views of the bounding surfaces in MNE-Python (grey), FieldTrip (lavender), DAiSS (SPM12) (mint) and Brainstorm (coral). d) Transparent view of the overlap and differences of the four surfaces (color indicates the outermost surface).

Fig. 5

The pipeline for constructing an LCMV beamformer for MEG/EEG source estimation. A similar pipeline was employed in all four packages.

Fig. 6

Localization error (a) and point-spread volume (b) as a function of input SNR for raw and SSS-pre-processed simulated datasets. The markers size indicates the true dipole amplitude.

Fig. 7

Localization error (a) and point-spread volume (b) as a function of input SNR for phantom data recording in a stable position. The markers size indicates the true dipole amplitude.

Fig. 8

Localization error (a) and point-spread volume (b) as a function of input SNR for data from the moving phantom.

Fig. 9

Source estimates of human MEG data. (a) Localization difference from the reference dipole location for raw and tSSS-preprocessed data. (b) Peaks of the beamformer source estimate of tSSS-preprocessed data. From left to right: visual stimuli presented to left (triangle) and right (square) upper and lower quadrant of the visual field (the two axial slices showing all sources); somatosensory stimuli to left (triangle) and right (square) wrist; auditory stimuli to the left (triangle) and right (square) ear. Reference dipole locations (yellow and orange circles).