The relationship between magnetic and electrophysiological responses to complex tactile stimuli

Abstract

Background: Magnetoencephalography (MEG) has become an increasingly popular technique for non-invasively characterizing neuromagnetic field changes in the brain at a high temporal resolution. To examine the reliability of the MEG signal, we compared magnetic and electrophysiological responses to complex natural stimuli from the same animals. We examined changes in neuromagnetic fields, local field potentials (LFP) and multi-unit activity (MUA) in macaque monkey primary somatosensory cortex that were induced by varying the rate of mechanical stimulation. Stimuli were applied to the fingertips with three inter-stimulus intervals (ISIs): 0.33s, 1s and 2s.

Results: Signal intensity was inversely related to the rate of stimulation, but to different degrees for each measurement method. The decrease in response at higher stimulation rates was significantly greater for MUA than LFP and MEG data, while no significant difference was observed between LFP and MEG recordings. Furthermore, response latency was the shortest for MUA and the longest for MEG data.

Conclusion: The MEG signal is an accurate representation of electrophysiological responses to complex natural stimuli. Further, the intensity and latency of the MEG signal were better correlated with the LFP than MUA data suggesting that the MEG signal reflects primarily synaptic currents rather than spiking activity. These differences in latency could be attributed to differences in the extent of spatial summation and/or differential laminar sensitivity.

Figure 1

An example of a somatosensory MEG response. Simultaneous stimulation of right D1 and D2 evoked responses in the left hemisphere. The response waveforms at three stimulation rates are shown. From top to bottom, the interstimulus intervals (ISIs) were 2s, 1s and 0.33s. The filter range was 1–100 Hz. Each curve shows one channel's time course. The first response peak appeared at 19 ms (arrow). While the latency of this peak did not change across ISIs, the amplitude decreased with increasing stimulation rate.

Figure 2

An example of coregistration of MEG and electrophysiological signals. Amira software was used to digitally reconstruct and "reslice" both MRI and histological sections, such that the locations of MEG signal sources and electrophysiological recording data could be identified with precision. A, C, E, Sagittal, horizontal, and coronal views, respectively, obtained using MRI in Case 24056. The yellow dot indicates the signal source identified using MEG. Blue, magenta, and green lines in panels A and B indicate planes of section shown in panels C-H. B, D, F: Sagittal, horizontal, and coronal views, respectively, that were digitally reconstructed from block face images obtained during histological sectioning of the brain. G: Horizontal view of the brain in Case 24056, digitally reconstructed from MRI data. Note that this plane of section matches the plane of section used for histological sectioning shown in panel H. H: Block face image of the brain taken during histological sectioning. The white box in panel H indicates the location of the photomicrograph shown in I. Arrowheads in panels B, D, F, and H indicate a single electrode track from recordings in the same location as the signal source identified using MEG. I: Digital photomicrograph of the section adjacent to the block face image shown in panel H, reacted for cytochrome oxidase. Asterisks indicate penetrations at which the receptive fields shown in panel J were obtained. The receptive fields at recordings in this location were identical for both MEG and electrophysiological recordings. The location of this penetration corresponds to the signal source identified using MEG. Digitally "resectioning" the MEG and histological data allowed us to match these data sets with high fidelity. Banding pattern in panels B, D, and F is due to lighting variations during photography of individual block face images.

Figure 3

Comparison of relative amplitude of MEG, LFP, and MUA data. The relative response intensities for 1s ISI and 0.33s ISI were normalized to the 2s ISI signal amplitude. For all data sets, there is a significant change in the amplitude of response, regardless of the measure, with a change in the ISI of the stimulus, which constitutes different rates of stimulation. The relative amplitude of the MEG data was significantly different from the relative amplitude of MUA (p < 0.001), but not from the LFP (p > 0.05) relative amplitude.

Figure 4

Comparison of peak latency of MEG, LFP and MUA. The absolute response peak latency with respect to stimulus onset was different for the three measures at each stimulation rate. Both original and adjusted latencies (the original peak latency minus the maximum time shift introduced by the cut-off frequency of the filter) were shorter in LFP than in MEG recordings. Both were longer than the peak latency of the MUA burst.

Figure 5

An example of the rate effect in LFP and MUA recordings. For MUA recordings, the same threshold was set for all trials recorded from the same cortical site. A spike was counted when the voltage was over the threshold level within a 1 ms bin. In A, raster plots of raw MUA data were collected over 100 trials from a single recording site, which is at the depth of 4000 μm in the central sulcus. Each dot represents a single spike. B shows the average LFP waveforms over 100 trials in the first 40 ms after the onset of the stimulus (as shown in the grey box in the inset LFP waveform at 2s ISI) at each stimulation rate recorded from this site. C shows the post-stimulus time histograms of the same data used in A. The green, red and blue lines in B and bars in C represent data at 0.33s, 1s and 2s ISIs respectively.

Figure 6

Scatter plots for the regression analysis. The response amplitudes at 0.33s and 1s ISIs are normalized to the response amplitude at the 2s ISI for each measurement method. Each plot shows the averaged amplitudes in one measurement method against another. The averaged amplitudes in one measurement metric recorded from three hemispheres are combined to make the scatter plots. The amplitudes at the 2s ISI are removed from the plots, since data were normalized to these values. A trend-line is fitted based on the six paired data points from the three hemispheres in each plot. The slope of the trendline and R² for the fit is shown in each plot.