Knowing the position (and orientation in the case of sensors measuring magnetic fields) is crucial to allow for estimation of neural sources from the data measured at the sensors.

Here we show an example of co-registering 4 OPM sensors placed in the Fieldline Alpha1 semi-rigid helmet to an MR of the participants brain.

First, we load modules we need

from OPM_lab.mne_integration import add_dig_montage, add_device_to_head, add_sensor_layout
from OPM_lab.sensor_position import OPMSensorLayout, FL_alpha1_helmet

import mne
from mne.viz import plot_alignment
from mne.utils._bunch import NamedInt

import pandas as pd

from pyvista import Plotter

The next step is to load both the raw OPM data, as well as the digitised points (i.e., the output after Digitising sensor positions)

opm_path = "path_to_your_data.fif"
raw = mne.io.read_raw(opm_path, preload=True)

dig_path = "path_to_the_dig_file.csv"
points = pd.read_csv(dig_path)

Next, rename the channels in the raw OPM data to correspond to the label of the sensor slot in the semi-rigid helmet

raw.pick(["00:01-BZ_CL", '00:02-BZ_CL', "00:03-BZ_CL", "00:04-BZ_CL"])

mne.rename_channels(
    raw.info, 
    {
        "00:01-BZ_CL" : "FL3", 
        "00:02-BZ_CL" : "FL10", 
        "00:03-BZ_CL" : "FL16", 
        "00:04-BZ_CL" : "FL62"
    }
)

After loading the data, any digitised head points, and EEG if any, is added to the Raw data object.

add_dig_montage(raw, points)

Define a variable with the depth measurements. It is important that the depth measurements are in the same order as the label input in the next code chunk!!

depth_meas = [40/1000, 47/1000, 44/1000, 40/1000] # mm converted to meter (order = 3, 10, 16, 62)

The next step is to determine the position and orientation of the OPM sensors relative to each other. This is done by initialising the OPMSensorLayout class, which takes a helmet template (in this case the FL_alpha1_helmet), labels of the sensors used and the corresponding depth measurements.

Important note: If you are using another type of sensor (e.g. not the default “FieldLine OPM sensor Gen1 size = 2.00 mm”) remember to specify it here using the coil_type flag.

sensor_layout = OPMSensorLayout(
    label=["FL3", "FL10", "FL16", "FL62"], 
    depth=depth_meas,
    helmet_template=FL_alpha1_helmet,   
    # coil_type=NamedInt("SQ20950N", 3024) # useful for plotting as orientation of OPMs (default coil types) are difficult to see on alignment plot

) 

The information about this subject-specific sensor layout which has been defined taking the depth measurement into account, is added to the data object using the code below. Here an example is shown using the a Raw object, but it works other MNE classes such as Epochs and Evokeds as well.

add_sensor_layout(raw, sensor_layout)

To move the sensor array to head space, a rigid 3D transform algorithm (Arun et al., 1987) is used to determine the rotation matrix R and translation vector Tthat aligns the OPM positions as described in the sensor array to the digitised OPM marks in head space. This transformation is described as the device to head transformation in MNE, and can be determined and added to the data object using the following code:

add_device_to_head(raw, points)

As this function relies on the sensor positions found in the data object, the order of using the functions is important.

To verify, that aligning the MR, head and sensor array coordinatesystems went well, we plot the alignment.

fig = plot_alignment(raw.info, meg=("sensors"), dig = True, coord_frame="head", verbose = True)    
Plotter().show()

After this step, MNE-python can be used to estimate the neural sources.