Magnetic encephalograms (MEGs) are methods based on measuring very weak magnetic fields (several orders of magnitude weaker than the Earth's magnetic field) caused by electrical activity in the brain.
MEG, researchers face the complex task of understanding which areas of the brain are active when measuring only sensors around the head.
this problem is called "anti-problem" and fundamentally there is no universal solution: any set of measurements can be explained by numerous different configurations of neuro-activity sources in the cortical cortical layer.
in order to make meG application practical, special mathematical methods are used to convert sensor signals into cortical activity maps.
these methods can be divided into two categories.
, as part of the so-called "global" approach, narrows down many possible solutions to the problem based on general a priori assumptions about brain activity.
these constraints, the researchers looked for source distributions in the corties that could explain the measurement data.
"local" approach, including the algorithms described in the paper (ReciPSIICOS), designed to find separate sources before creating complete images of brain activity.
ReciPSIICOS uses an adaptive beam-forming device (BF) - a method for processing sensor measurements that detects activity signals in target neuron populations.
this, it tries to mute signals from other sources, rather than mute signals from all other sources, as the Global method does, but only signals that are currently active.
this method provides a higher fidelity in the activity visualization than the Global method when only the active signal is suppressed.
, however, the method can also suppress target signals generated by neuron integrations activated at the same time as neuron populations in other brain regions.
in real life, this correlation reflects the interaction between neuron populations, an inherent property of the brain, and researchers must look for ways to overcome this barrier.
information about the active neuron population and its interacting properties is encoded in a special co-variance matrix, which can be calculated from sensor data.
beamforming algorithm uses this matrix to determine which sources should be suppressed.
strictly speaking, this approach only works if the source does not interact: information about this interaction is also included in the correlation matrix and negatively affects the performance of the beam-forming algorithm.
using observed data models and correlation matrix models, researchers have developed a mathematical algorithm that removes information about source interactions from related matrices.
, they extended the application of beam formation methods to environments with synchronous neuron sources and provided the necessary precision in the group of neurons that visualize interactions.
" brain magnetogram technology combines the ability to record precise aspects of the evolution of neuron activity time and the ability to locate potentially high fidelity of active neuron populations.
The first feature comes from the recording of electrical activity, which changes significantly faster than the hemodynamic reactions used by functional magnetic resonance imaging and require complex mathematical methods in order to achieve high accuracy in spatial positioning, said Dr. Alexey Ossadt, author of the study.
To evaluate the performance of the algorithm, the researchers first generated a data set of signals actually received by analog sensors and tested four methods: two types of ReciPSIICOS and two previously developed algorithms (linear constraint minimum variance (LCMV) beam-former, and minimum paradigm estimation (MNE) methods).
LCMV and both ReciPSIICOS methods work without correlation between signals, but When correlations exist, ReciPSIICOS is better able to handle tasks than its predecessor.
under pressure tests for positive modeling accuracy, the results are similar: ReciPSIICOS has been shown to be less sensitive to the inaccuracy of the models used, which is unavoidable in practice.
scholars have also demonstrated the operability and high performance characteristics of the new method through some actual MEG data sets, which are characterized by the presence of synchronous neurons, which cannot be properly processed by traditional beam formation algorithms.
source: Researchers find a way to increase spatial resolution in brain activity visualization Original source: Aleksandra Kuznetsova et al, Modified covariance beamformer for solving MEG inverse problem in the environment with correlated sources, NeuroImage (2020). DOI: 10.1016/j.neuroimage.2020.117677