Magnetoencephalography

How MEG Works

As previously described, MEG measures magnetic fields associated with neural activity. These signals are extremely weak: The largest neuro-magnetic fields, such as the spontaneous alpha rhythm, are only approximately 10-12 tesla (T). Magnetic fields evoked by stimulus presentation are usually even smaller, in the range of tens to hundreds of femtotesla (fT; 10-15 T). For comparison, the magnetic field of the earth is about 30 to 50 microtesla (μT; 10-6 T), roughly a billion times greater than these evoked neuromagnetic signals. Recording these weak neuromagnetic signals requires highly sensitive superconducting quantum interference device (SQUID) sensors capable of detecting tiny magnetic fields. A single MEG machine typically contains hundreds of SQUIDs arranged in a helmet shape inside a cryogenic dewar, a container filled with liquid helium to maintain the superconductivity of the SQUIDs. Sensors are designed for maximal sensitivity to the nearest source of interest, the brain, excluding magnetic fields located further away. This equipment can also be enclosed in a magnetically shielded room to further exclude ambient magnetic noise from sources such as electrical devices, radio frequency signals, and the Earth's magnetic field.

Recorded MEG data represents changes in magnetic field strength as a function of time. Perception research often focuses on evoked magnetic fields in response to stimulation, calculated by averaging together tens to hundreds of trials. These evoked responses can be studied either in isolation or in conjunction with source localization. Source localization methods attempt to solve the “inverse problem” of estimating neural sources from the signal recorded at the scalp. Mathematically, the inverse problem has an infinite number of solutions: Any number or combination of neural sources can produce a given signal measured at the scalp. Therefore, MEG cannot achieve the precise spatial localization of neuroimaging techniques, such as functional magnetic resonance imaging.

However, additional constraints based on anatomy and known brain functionality, as well as mathematical techniques (e.g., beamforming), can allow better estimates of the origins of MEG signals.

...