MEG Coregistration

Contrarily to EEG, MEG sensors are not attached to the subject's scalp. The head of the subject can move rather freely inside the helmet, unfortunately. This is critical when one wants to average out several trials extracted from multiple acquisition runs because sensors are not at the same locations with regards to the brain inbetween runs. This causes at best, loss of sensitivity in the data average. We have developed a registration approach that will allow you to coregister multiple MEG datasets to a single sensor helmet. The location of a given sensor in that helmet is obtained from the average of this sensor's locations across runs. Some runs may have excluded some 'bad sensors', some may have not. Hence, the registered data sets will be estimated on the set of sensors that are common to all acquisition runs. From our simulation studies, coregistration correction is accurate if movement amplitudes between runs/subjects remain below 3 cm on average.

Recent MEG systems include real-time head-tracking solutions. Using our approach, you will be able to correct offline MEG recordings from all movement artifacts, within an acquisition run!

Here are the steps you are going to take:

1. Select the datasets you want to coregister to a common MEG helmet
2. Run the coregistration process

Selection of MEG datasets to coregister

From the DataManager window, right-click over a dataset you want to add to the coregistration process and select 'Add to process candidates'.

The coregistration technique is one of BrainStorm's processes. We will soon detail what a process is in BrainStorm's sense. In short, this is how we would like to run batch procedures on multiple datasets (including averaging, source estimation and ... coregistration).

A new window appears 'Batch process candidates' (see screenshot below) :

Keep on adding all the files you want to coregister to the same MEG helmet. Click on ' Done' in the 'Batch process candidates' window when finished.

You may add all datasets from a study by right-clicking over a study name in the DataManager and select 'Add all datasets in folder to process candidates'.

A new window pops-up and this is the 'Process Launcher':

Select Co-register MEG datasets in the list of Available processes; this latter will show into the Duty List on the right (see figure, above). Note that there are some other duties available... More nice features from BrainStorm probably ;)

Click on Launch!

The coregistration process will now run onto your preseleted MEG datasets.

Several window will pop up as the process is gently going through the data:

1) Sensory arrays - average sensor array in green, shows the multiple MEG sensors arrays from your selected datasets and the resulting average helmet will show up in green (see below)

2) One window per selected MEG file is created (see below). This will help you evaluate whether things have been running properly and how different from the original data, the coregistered outcome could be.

The figure displays the original and co-registred time series on all sensors with the average distance between the sensor helmets (here 1.38 cm). The surface topographies of data are also shown at the time instant of maximum surface data intensity. The Corr index is the average empirical correlation between all original/co-registered time series (here: 89.60%)

Go back to the DataManager, and click the Refresh button.

The newly-created datasets are stored in subfolders named \coregistered placed within the respective original data folders. The data files themselves have a _R_ in their filename which recalls they are 'registered'.