Multichannel magnetoencephalography based on optical-pumped magnetometers

Magnetoencephalography (MEG) is a modern and highly demanded method of noninvasive neuroimaging in brain research and some medical applications, such as localization of epileptogenic sources, surpassing its more well-known sister method EEG in accuracy and depth of reconstruction of electrophysiological signal sources in the brain.

Despite a potentially wide range of applications, standard quantum interferometry-guided interferometry (SQUID) devices for MEG recording (Figure 1) are not widely available to most researchers because of the uniqueness of each device. They also have disadvantages, such as the long distance of the sensors from the head surface, which weakens the signal of cortical sources, the limited movement of the subject, the expensive maintenance and the stationarity of the apparatus.

At the Center for Bioelectrical Interfaces, we are developing an alternative MEG recording system based on sensors with different physical principles, including QuSpin sensors based on the optical pumping principle (OPM). OPM-sensors are devoid of many disadvantages inherent to SQUID-systems, each sensor has a compact body, which can be attached to a usual EEG cap, and it allows relatively free movement of the head during signal recording, as well as neater localization of cortical sources due to minimization of remoteness of sensors.


Fig. 1. Elekta-Neuromag MEG apparatus with SQUID sensors (left), QuSpin sensors, an example of their mounting on an EEG cap and recording from a subject, prototype MEG sensor (right).

The capabilities of the sensors can be demonstrated on simple experimental tests of recording functional brain rhythms - occipital alpha rhythm or a less pronounced sensorimotor rhythm. Both rhythms are well registered by sensors, and the simplest analysis of records of sensorimotor rhythm recorded from a small area of head surface allows to distinguish several sources with unique frequency-time characteristics, clearly see moments of rhythms synchronization and desynchronization (Figure 2), and such effects as beta-rebounding, which reveals remarkable spatial and frequency-time resolving abilities of sensors.

Fig. 2. Averaged envelopes of the frequency-time transformation of motor rhythms registered with the QuSpin OPM-sensors in two states of the subject - motor imagination (MI) and relaxation (REST).

Another important characteristic can be considered the possibility of their use in practical applications, such as in the brain-computer interface (BCI).   In our experiments we implemented an ideomotor BCI based on only 4 OPM sensors. The BCI is controlled by imagining hand movements, and the signals from the OPM-sensors are converted in real time into two commands to control the virtual phantom (movement and relaxed state). 

The result of the experiment was a rather intuitive control with minimal recognition errors after a certain number of trainings of the subject, possible even on a single sensor located in the correct area of the head. The conducted experiments speak about high quality of the signal and reduced requirements for spatial filtering and additional processing when using OPM-sensors, and outline further prospects for expanding the range of their application and increasing their accessibility to researchers.