Group Analysis in MNE-Python of Evoked Responses from a Tactile Stimulation Paradigm: A Pipeline for Reproducibility at Every Step of Processing, Going from Individual Sensor Space Representations to an across-Group Source Space Representation

Abstract

An important aim of an analysis pipeline for magnetoencephalographic data is that it allows for the researcher spending maximal effort on making the statistical comparisons that will answer the questions of the researcher, while in turn spending minimal effort on the intricacies and machinery of the pipeline. I here present a set of functions and scripts that allow for setting up a clear, reproducible structure for separating raw and processed data into folders and files such that minimal effort can be spend on: (1) double-checking that the right input goes into the right functions; (2) making sure that output and intermediate steps can be accessed meaningfully; (3) applying operations efficiently across groups of subjects; (4) re-processing data if changes to any intermediate step are desirable. Applying the scripts requires only general knowledge about the Python language. The data analyses are neural responses to tactile stimulations of the right index finger in a group of 20 healthy participants acquired from an Elekta Neuromag System. Two analyses are presented: going from individual sensor space representations to, respectively, an across-group sensor space representation and an across-group source space representation. The processing steps covered for the first analysis are filtering the raw data, finding events of interest in the data, epoching data, finding and removing independent components related to eye blinks and heart beats, calculating participants' individual evoked responses by averaging over epoched data and calculating a grand average sensor space representation over participants. The second analysis starts from the participants' individual evoked responses and covers: estimating noise covariance, creating a forward model, creating an inverse operator, estimating distributed source activity on the cortical surface using a minimum norm procedure, morphing those estimates onto a common cortical template and calculating the patterns of activity that are statistically different from baseline. To estimate source activity, processing of the anatomy of subjects based on magnetic resonance imaging is necessary. The necessary steps are covered here: importing magnetic resonance images, segmenting the brain, estimating boundaries between different tissue layers, making fine-resolution scalp surfaces for facilitating co-registration, creating source spaces and creating volume conductors for each subject.

Keywords: MEG; MNE-Python; analysis pipeline; good practice; group analysis; minimum norm estimate (MNE); tactile expectations.

Figure 1

An example sequence of the experimental paradigm is shown. The annotations on the bottom show the coding used throughout for the different events of interest. Stimulations happened at a regular pace, every three seconds. When omissions occurred, there were thus six seconds between two consecutive stimulations.

Figure 2

Cookbook for performing the Minimum-Norm Estimates for a single subject.

Figure 3

ICA components corresponding to eye blinks (ICA 000 and ICA 0001) and heart beats (ICA 028).

Figure 4

Epochs and event-related field for channel MEG 1812 for condition standard 3. The colouring indicates the field strength for each epoch. Two evoked responses can be seen after about 60 and 135 ms respectively.

Figure 5

Source space. Sources are restricted to the cortex. Yellow dots mark equivalent current dipoles on the cortical surface.

Figure 6

Transformation. The positions of the head, skull, brain, and helmet sensors after the transformation.

Figure 7

Noise covariance matrices. As can be seen the covariance between magnetometers is greater than between gradiometers. This can be explained by magnetometers being more sensitive to far away sources than gradiometers are.

Figure 8

Spatial distribution of neural activity at 56 ms for standard 3 for sub-01: There is some spread, but there is a clear activation of the contralateral sensory cortex. Values are dSPM-values. These are current estimates normalized with the noise-covariance. The cortex is shown inflated with gyri darker than sulci.

Figure 9

Grand average butterfly plot for standard 3 showcasing the SI (56 ms) and SII (135 ms) components.

Figure 10

Spatial distribution of neural activity at 56 ms for grand average of standard 3: There is some spread, but there is a clear activation of the contralateral sensory cortex. Values are dSPM-values. These are current estimates normalized with the noise-covariance. The cortex is shown inflated with gyri darker than sulci.

Figure 11

A t-value map for standard 3 vs. non-stimulation at 56 ms. The cortex is shown inflated with gyri darker than sulci.