magnetoencephalography

Summary

In physics, any electric current is accompanied by the production of a magnetic field. Ion currents flowing inside the brain and nerves also generate magnetic fields. The superconducting quantum interference device (SQUID) has been developed in the last 30 years, and magnetoencephalography (MEG) uses this highly sensitive detector to measure the very weak magnetic fields at the surface of the scalp. MEG can record the magnetic fields generated by thousands of synergistically active neurons and monitor activity in different areas of the brain without causing damage to the brain.

Operation method

magnetoencephalography experiment

Materials and Instruments

DC Superconducting Quantum Interferometer Cryogenic Cooling Tank Magnetic Field Shielding Chamber

Move

move I. Maintenance steps

There are two different types of maintenance: routine maintenance and special maintenance in case of imperfections or malfunctions in the MEG system.

1. Routine Maintenance

In most modern MEG systems, routine MEG maintenance is performed by weekly renewal of the liquid helium in the cryostat, as well as computer software and hardware checks (data backup).

2. Specialized Maintenance

Special maintenance is required when the shielding quality of the MSR deteriorates. Magnetic fields or mechanical stresses can adversely affect the highly permeable housing of the MSR (e.g. moving large instruments in close proximity, deformation due to machinery).

In this case, we have to de-magnetize the MSR to optimize the performance of the magnetic material in the only device.

This process can be accomplished by placing a coil around the housing of the MSR (so that the MSR becomes part of the coil) and applying an alternating current (e.g., 50 Hz or 60 Hz mains) to the coil, so that the current slowly rises to a peak of 400-800 amperes and then decreases gradually to zero.

Subject Preparation 1. Magnetic field purification

The subject must not be wearing any artifacts or magnetic contaminants that could cause any kind of movement (breathing, heartbeat, etc.). Check if the subject is wearing any of the following items:

Metal objects on clothing such as zipper fasteners, buttons, hooks, etc.

One metal object on a bra or corset/corset (usually worn by females, these items can generate magnetic contamination with the female subject's breathing)

A wallet, keys, and jewelry made of non-precious metals

Dentures or implants that sometimes become ferromagnetic.

Magnetic deposits in the lungs or encapsulated magnetic objects inside the body (e.g., debris left over from injuries)

The only way to reduce this type of interference is to demagnetize the subject. During demagnetization, the subject is exposed to an alternating magnetic field that decreases to O In our experience, commercially available coils used to demagnetize television tubes can accomplish this (e.g., Bemstein Model 2-505, Werkzeugfabrik Steinriicke). The coil is rotated centrifugally in the horizontal plane in front of the subject at a rate of IHz^. after 5?8 revolutions, the coil is slowly moved away from the subject in a helical motion until it is 2 m away from the subject, at which point the coil has rotated about 20 revolutions, the coil is tilted at an angle of 90. and the current is stopped.

2. Determination of the head coordinate system - placement of markers

In many cases, the purpose of MEG measurements is to find the location of a biomagnetic field source based on anatomical features of the head.

Therefore, a link must be established between the coordinate localization system of the MEG sensor and the coordinate system of the head (built on anatomical landmarks) and, ultimately, with the imaging system (e.g. MRi system). There are two main steps:

(1) The datum reference point for the MEG system is first determined by scanning the scalp surface or by means of a 3D scanner (e.g. IsotrakII, Pdhemus, see Appendix). Through the fitting process, the scanned scalp surface (or datum) can be matched to the anatomical contour of the head (or to the datum) established from the MR image.

(2) Another step is the use of markers placed on the subject's head. During MEG measurements, markers with small coils are used. When these coils are energized, they produce magnetic dipoles that can be localized by the MEG meter, and the markers are filled with the appropriate material during the MR or PET image scanning process.PET scanning does not allow for the direct localization of anatomical structures with millimeter accuracy, but it is possible to combine the PET scanning image with the MEG coordinate system by using the appropriate markers.

The following is the examination procedure:

Three markers were placed at three points at the root of the nose, bilaterally in front of the ears, and fixed using double-sided tape. The markers consisted of 3 coils, each with a precise geometric shape (printed circuit) (BeckeretaL1992); and were secured to each of the three faces of a small obtuse prism (DiekmannetaL1995). The other markers consisted of single planar coils. In order to fix the transformation of the coordinate system on a computer, these coils need to be placed on the midline of the head (e.g., top of the head, occipital region, etc., and the coils can be aligned with the EEG electrodes). A low-frequency alternating current (e.g., 8 Hz) can be used to gradually apply voltage to the marker coils, or all coils can be pressurized at the same time using alternating currents of different frequencies. The current strength should be stable but adjustable. Starting with a small current strength (e.g. 0.1 mA) gradually increase the coil current strength, starting with the sensor closest to the marker coil and ending with the sensor furthest away from the marker coil, so that eventually all coils are exposed to the optimal signal (high signal-to-noise ratio without increasing the maximum slew rate of the SQUID). With the help of these measurements, the coils can be identified by means of a constant repetitive step similar to that used to locate the source of a biomagnetic field. With these measurements, the position of the coils can be determined by repeating steps similar to those used to locate biomagnetic field sources (see the appendix "Data analysis: inversion algorithm"), but the localization of a biomagnetic field source has to be established first by establishing precise assumptions about the magnetic field sources (magnetic dipoles defined by the currents and the geometry of the coils), the number of sources, and their relative orientations (defined by the geometry of the prisms).

For magnetic resonance imaging, the cylindrical hole in the center of the marker (depth 5 , diameter 6 ) was filled with a similar 10 mmol/L CuSO4 solution as a contrast. For PET imaging, this hole should be filled with a positronic radioactive solution.

3. Preparation for Simultaneous Use of EEG Recording and EEG Electrodes

It has already been pointed out that using MEG together with EEG is a very attractive approach in many cases.

In some clinical settings, this combination is very common.The EEG electrodes must be non-magnetic and not cause magnetic interference, and the electrodes should be disk electrodes, which minimize the distance between the head and the cryostat.

The number and placement of electrodes should be varied depending on the purpose of the diagnosis and study. If surface source localization is performed simultaneously with a uniform physical model using a combination of MEG and EEG (see Appendix "Data Analysis: Inversion Algorithms"), the head recording sites must be sampled with sufficient density.3 This may require repositioning of the standard 10-20 EEG recording system on a 2- to 3-cm grid.

Ideally, the EEG electrode must have a coil when used in conjunction with the MEG (Beckeretal.1992). With the aid of this coil, it is possible to determine the position of the head electrodes by a method similar to the electrode marking described above. Once the electrode position has been determined, the shape of the head can then be determined and this data can be combined with the corresponding MR data to allow for the interconversion of the MEG and MRI coordinate systems (coordinate system determined by 3D scanning).

4. Subject position

Any movement of the subject will cause the magnetic field detected by the MEG sensor to change. Therefore, the subject's head must remain relatively stationary with respect to the MEG system. For example, by means of vacuum pads and inflatable balloons (used in helmet-mounted cryocoolers). The subject should be comfortable in a stationary sitting or lying position, avoiding magnetic beds and chairs It must be ensured that the subject's body is in the most comfortable position while undergoing the test in order to minimize the possibility of subject movement. Depending on the subject's task and body position during the measurement, a vacuum pad should be used to keep the subject's body still and stable.

5. Connection of Stimulus Presentation Instruments and Monitoring Equipment

Stimulus presentation and monitoring instruments inside the shielded room must not generate their own magnetic fields, and connections to instruments outside the shielded room must not affect the electrode function of the SQUID. In particular, cables connected to the shielded room must be thoroughly filtered to minimize radiation frequencies and ground lines must be avoided.

In order to understand the importance of these issues, it must be noted that electromagnetic interference (EMI) generated by radiated frequencies can seriously degrade the signal-to-noise ratio (SNR) of SQUIDs, and can even cause them to stop working. In fact, high-frequency fields have two negative effects on SQUIDs: (1) The Josephson effect, the principle on which the SQUIDs work, will be seriously interfered with, resulting in unstable operation or even a complete stop of operation. The limited bandwidth of the readout electronics cannot be synchronized with the rapidly changing high-frequency field, which will not lock the system. To avoid EMI, each low-frequency wire (power or signal) passing through the MSR must be guided through a pass-through filter adjusted to the appropriate bandwidth and sufficient current/voltage. High-speed signals and data lines like triggers must be routed through the light to achieve a wide bandwidth and EMI safe range.

Ground lines (multi-pole ground connections) can cause serious interference to measurement signals. The amplitude of the interfering signals can exceed the brain's magnetic field signal by about two orders of magnitude. Therefore, like every component in a complex electronic system, the ground shield must be star-shaped rather than polygonal. This requires a connection between the insulator's chassis and the MSR's enclosure to avoid direct grounding of the insulator. For instruments above the frequency of the wire (walkie-talkies, televisions) it is possible to weaken the signal connection by inducing a transformer or closing the capacitor circuit, thus cutting off the ground.

6. Special devices

Visual stimuli must be projected into the magnetic field shielded room through small holes in the wall of the shielded room, or the subject may be allowed to receive the stimuli presented by a monitor outside the shielded room through the small holes, or the stimuli may be presented through a fiber optic system.

The auditory stimuli must be conducted to the subject's ear through a non-magnetic conduit, which may be IOmm diameter PVC (polyvinyl chloride) tubing, with one end of the conduit connected to the shielded chamber and the other end fitted with an adjustable hose to connect to the subject's earhole. The tubing outside the shielding chamber should be connected to a small electronic acoustic transducer. Normal headphones or amplifiers cannot be used as they produce large magnetic field artifacts. For simple signals, piezoelectricbuzzers can also be used if the energizing cable is far from the MEG's transducer.

For tactile stimuli, barometric, hydrodynamic, or voltage stimulators can be used if the stimulation device (in close proximity to the subject) does not contain ferromagnetic material. For sending nociceptive stimuli, light pulses can be emitted through laser pulsers and glass fibers outside a shielded room. Somatosensory stimuli can also be sent via conventional electrical stimulators, but the cables conducting the stimulation current are carefully twisted.

Intercom and piezoelectric loudspeakers can be used inside the shielded room, but their connecting cables have to be kept at a considerable distance from the sensors.

Television cameras for monitoring subject behavior must be packaged in boxes shielded from electric and magnetic fields.

Instruments, monitors (e.g., C ○ 2) and interventionbatteries required by the patient are to be installed outside the shielded room and connected to the patient through the cables and plastic conduits described above.

II. Measurement instructions 1. Select data recording parameters

Before starting MEG recording, the parameters of the data collection program must be set. Parameter settings for filtering and sampling frequency should be taken into account when recording changes in the organism's function over time:

The upper and lower frequencies (comerfrequencies) should be set according to the desired frequency of the signal. High-pass and low-pass filters should be set to improve the signal-to-noise ratio, and high-pass filters should limit the sampling frequency to control the amount of data collected. Brain magnetic field signals typically contain components below IkHz. Normal spontaneous brain magnetic activity, mid- and long-latency, and epileptic-like magnetic activity ranges from about 0 to 70 Hz, while short-latency magnetic fields and fields generated by multiple neuronal coactivation may contain components with frequencies in the hundreds of Hz range.

The sampling frequency should satisfy the sampling principle and avoid confusion (see Chapters 35 and 45). The sampling frequency is basically set according to the uppercomerfrequency, but the filterste print ness above this frequency should also be taken into account. The sampling frequency for recording spontaneous brain activity is about 200-500 Hz, whereas the compoundactionfields of action potentials from peripheral nerves may require up to several kHz.

Data can be recorded continuously or only segments of data related to the stimulus event can be recorded. Continuous recordings can be analyzed offline using different epochs, but such recordings require large enough storage (hardware) space. Discontinuous recording requires less storage space but limits subsequent data analysis. The experimenter must specify whether the raw or preprocessed data (e.g., magnetometer vs. gradient coil signals) should be retained or averaged according to an online algorithm. In clinical applications, the last recording method is bear enough to fulfill the requirements in some cases. However, for research purposes, it is recommended to record single scans (singlesweep). The above methods can be used only if the researcher has the appropriate software for recording and analyzing the data and does not have to write a computer program to analyze the data himself.

2. Calibration

Each part of the MEG recording requires several technical measurements to achieve a fixed standard and to minimize noise interference. When the subject is ready to collect physiologic data, calibration must be performed. Depending on the MEG system used by the researcher, calibration should include the following steps:

Determination of a software gradient coil coefficient

Calibration of an active shielding instrument

Measurement of the head position according to the sensor's coordinate system

This is a necessary step in all systems that use 3D scanning or magnetic labeling (see above, Determination of Head Coordinate System and Marker Placement). This step must be performed very carefully, as all inferences about the location of the 3D source in the brain are based on this step. The subject must keep his/her head still from the beginning of this step to the subsequent measurements. If head movement is detected during the measurement, it is best to stop the current measurement and redo the head position determination. If the experimental design does not allow for stopping in the middle, the head position must be measured after the experiment has been performed and the results compared with the head position before the measurement to check whether the head movement is within the pre-determined range. Even if the range of head motion was not estimated in advance, the head position should be tested at the end of the experiment to determine if the head remained stationary during the experiment.

3. Appropriate biometrics

Biological measurements are more dependent on experimental and diagnostic purposes and there are some basic recommendations for such measurements. Whenever possible, conditions permit, make on-line visual estimates of the quality of the recorded data. As long as the experimenter controls the MEG measurement environment, he must ensure that no movement of ferromagnetic material occurs during the measurement (e.g., elevators). In order to identify artifacts due to the subject's movements, it is preferable to control the subject's behavior by monitoring the subject's behavior via television or by keeping a laboratory assistant in the shielded room. In clinical applications, it is desirable to keep a monitor with the patient to prevent epileptic or claustrophobic episodes.

III. Data Analysis 1. Pre-processing

One of the first important steps in MEG data analysis is to preprocess the data. The purpose of preprocessing the data is to improve the signal-to-noise ratio (SNR). The MEG data collected for a given experiment consists of a series of time series. Therefore, all of the techniques used to improve the SNR of the time series, especially those used in EEG data acquisition, can be used in MEG data analysis.Digital filtering is used to limit the bandwidth of the data to a specific "band of significance," or to eliminate things like low-frequency drift or powerline components.

That is, averaging over N time periods improves the SNR by l/N1/2 (see Chapter 45). In practice, the increase in SNR is smaller because the response of biological systems will change over time (e.g., fatigue or adaptation processes), and much of the noise itself is of biological origin (spontaneous EEG activity, heartbeat activity, etc.), which may be non-Gaussian and spatially and temporally correlated.

The elimination of spatially coherent components (e.g., artifacts produced by eye movements) is commonly used when dealing with biological noise. This 'filter' is implemented (in a broad sense) by recording the noisy signal (e.g., blinks) in the vicinity of the noise source using different leads and identifying, and subtracting from the signal data, the relevant biologically interfering activity in the individual leads (for the corresponding algorithms, see Widrow1985, Abraham-FuchsetaL1993).

Huotikinen et al. (1995) developed a method for eliminating or reducing biological noise without the use of separate lead recordings of "noise". This method is based on the signalspaceprojection method. The principle of the method is as follows: a time period is labeled using maximum noise estimation, and noise (e.g., blinks) is identified from the signal leads by this labeled time period. This time period represents a multidimensional noise vector with a specific wave amplitude and direction in the signal space. In principle, within this time period, if a noise component that appears in the signal is less than the maximum estimate, the component of that signal component that has the same vector direction as the noise is removed.

The SNR can also be reduced by the temporal "timejitter" caused by the signal (or trigger signal) associated with the evoked stimulus (i.e., by using random time intervals in a range), which can be improved by estimating the stimulus-response latency before averaging the overlap, e.g., by the matched-filter technique (Whalen 1971). Whalen 1971).

deWard (1981) and Bertrand et al. (1990) describe more sophisticated methods to overcome the shortcomings of direct averaging. However, reducing noise during the recording of data is the preferred method for obtaining the best SNR.

2. Directional surveys and magnetic topography

After the preprocessing preparations have been completed, it is possible to determine the distribution of the magnetic field at each sampling and to present these distributions graphically (all commercial systems include graphing programs). Typically, these graphs can depict the distribution along the surface of the head, or the contours of the static magnetic field density on a sphere fitting the head, or on a plane tangent to some point of the head. They allow a direct presentation of the results of the measurements and an initial estimation of the source of the magnetic field and its corresponding temporal variations.

For example, Figure 37-5A presents the magnetic field induced by an electric shock to the left median nerve, with data sampled at nine different latency ranges over a matrix of 22 sensors on the surface of the right hemisphere's central region. The time series in Figure 37-5A represent 2 rough dipole magnetic field patterns, one generated around I8ms and the other around 28ms. Based on the results, it can be roughly estimated that there are two ECDs in approximately the same two regions, one pointing positively toward the y-axis (18ms) and the other pointing negatively toward the axis (28ms).

Note, however, that as the magnetic field pattern becomes more complex (Fig. 37-5B), it becomes more difficult to estimate the source of the magnetic field based on visual inspection alone.

3. Solving the Inverse Problem (1) Magnetic field source model selection

The selection of a magnetic field source model is as critical a step in localization as the selection of a head or torso model ^ In addition to the magnetic field topography, physiological constraints must be taken into account. For example, a patient with focal epilepsy has a brain that produces distributional spikes, and at some point in time a magnetic field with dipole characteristics appears; localizing this dipole is best done with a single ECD.

Similarly, at least two ECDs should be used to study stimuli that induce bilateral activity, but in addition to physiological considerations, the choice of model should also take into account speed, computer configuration, software functionality, and the purpose of the experiment.

If only one temporal feature is analytically significant (the peak of the wave at the time of a single epileptic spike issuance), we only need to decide whether or how many discontinuous ECDs are needed or whether the form of the magnetic field permits the use of distributed source analysis (described in detail below).

If the entire time series of a given activity is to be analyzed, multiple models need to be considered, and we restricted the models to 3 categories:

a) SingleMovingDipoleModel. If the topography of the magnetic field shifts from time to time, but remains dipolar, unless physiologically exhibiting a different pattern, the bioelectric current source should use a single ECD, whose parameters (position, moments) are determined by fitting the magnetic field separately within each time event. Thus, in a sense, the position of the dipole represents the center of gravity of the magnetic field generating source, which changes with the time course. This model is widely used in scientific research and medicine (AineetaL1998). It is important to note that the fitting of a single dipole mentioned above is often referred to as a moving dipole fit, since the position of the dipole changes with each repetition of the fitting process.

b) Fixed dipole and rotating dipole models. If no individual dipole features are shown in the magnetic field topography, or if the physiological data show multiple dipoles active together at the same time, it is generally assumed that the magnetic field sources have a fixed position and pointing, and that only their intensities change with time.

Alternatively, the current pointing can be made to change with time. These two models are called fixed dipole and rotating dipole models, respectively. In addition to providing information about the position of the dipoles, these models make it possible to specify the changes in the intensity of these dipoles over time and, in the case of the rotating dipole model, to estimate the mode of rotation of the dipoles. The use of such wedges presupposes a high SNR; moreover, it is often difficult to determine the appropriate number of dipoles. Finally, if the distance between the two sources is too close for a given sensor we cannot use a fixed dipole model. Under the typical sensor-source distance conditions of current instruments (4 cm), it is difficult to distinguish between sources that are 2 cm or closer. Fixed and rotating dipole models were first proposed by Scherg (1984) to analyze EEG data (Scherg 1990; Mosheretal.1992), and this approach has now been successfully used in evoked potential and magnetic field activity analysis.

c) Current distribution models. The purpose of these models is to determine a current density distribution that explains the measured data, rather than to localize several individual ECDs. As mentioned above there is no unique solution to this inverse problem. Therefore, physical or mathematical methods must be used to limit the number of possible solutions. Based on these restrictions, several source-distributed solutions to the inverse problem arise. The first method is the traditional minimumnormestimate (see HSmSlainenetal.1993 for an early review),. This method requires that the Euclidean or L2 criterion of the current source be minimized. Later fine tuning based on other limitations can be divided into several categories: LQRETA (lowresolutiontomography, translated by Pascual-Marquietal.1994); MFT (Ioannidesetal.1995); CCI (Fuchsetal.1995); FOCUSS (Focalundertomography, translated by Pascual-Marquietal.1995); FOCUSS (Focalundertomography); FOCUSS (FOCUSS); FOCUSS (FOCUSS); FOCUSS (FOCUSS); FOCUSS (FOCUSS). FOCUSS(Focalunderdeterminedsystemsolver, Focusing Algorithm, Translator's Note: GorodnitskyandRao1997);VARETA(Valdes-Sosaetal.1998).Robinson and Rose(1992) introduced a method for estimating the current distribution by null homo til wave to estimate the current distribution.

All methods must be used with caution. They are not individually designed algorithms for solving inverse problems. Any of these methods will have varying degrees of ghostimage, producing a blurred image at the point source. At the same time, conventional minimum-paradigm estimation methods provide the best estimation of magnetic field sources close to the sensor-minimum-normsolutions can be used (1) if there is a lack of, or only a small amount of, limiting paddle information and a lack of reasonable assumptions, and (2) if the purpose of the study can be accomplished by estimating the region of maximum current density. Minimum-normsolutions are used to analyze the late component of evoked activity, to localize epileptic foci, and to locate pathological brain rhythms (Grummichetal.1992;Bamidisetal.1995;Ioannidisetal.1995).

(2) Selection of volumetric models

The choice of volumemodel (head, arms, legs, torso) can have a significant impact on the accuracy of localization; this is especially true for analysis of MEG and EEG data fused (or EEG data analyzed separately). MEG source localization data for scalp surfaces can often be optimized using a simple sphere model if the surface cranium can be well fitted to a sphere. Only when the shape of the frequency bone or cortical surface deviates significantly from a sphere can a real head model significantly increase the accuracy of the fitting results, which requires more complex programs, more powerful computers, and more time. There are also finiteelementmodels, which have the advantage of taking into account the different electrical conductivity of the cortical surface (brain tissue distributed along the course of the nerve fibers has a higher electrical conductivity than brain tissue spanning the nerve fibers). Therefore, the following factors need to be considered when choosing a volumetric model: (i) the assumed source of data generation; (ii) the accuracy of localization; and (iii) the availability of computational tools. There are three volumetric models:

a) simplegeometricalmodels. Spheres, spherical shells, infinitehalfspace. The advantage of these models is that the positive problem can be solved analytically, and this method is computationally faster. This model must be used if (i) the estimation accuracy is only required down to the centimeter, and (ii) the shape of the scalp in the region where the source of the magnetic field is located can be well estimated with a sphere (e.g., the central region of the cortex of the main motor and somatosensory areas, or the optic cortex of the parietal lobe). Simple geometric models are used when computational speed is more important than computational accuracy, or when large amounts of data need to be processed with limited computational power. Which geometric model is most appropriate needs to take into account the characteristics of the data being analyzed:

MEG data. Depending on the shape of the brain being measured, a simple sphere model with the best match to the local curvature can be chosen. It is preferable to obtain limbic data of the cerebral cortex based on the measured MR data of the subject. If MR data are not available, we can fit the head to a sphere either from the head contour curves obtained by the 3D scanner or from the EEG electrode positions determined in the magnetic field; the sphere model should be 1.5 cm smaller than the actual radius, taking into account the thickness of the scalp and skull. Fusion of MEG and EEG data (or EEG data only). A 4-shell sphere model should correspond meaningfully to the cerebral cortex (innermost layer), cerebrospinal fluid, skull, and scalp, respectively. In our experience, the use of concentric sphere models serves the purpose. Data on the shape of the edges of each layer should be obtained from the MR images of each subject, and in the absence of individual data, the contour of the subject's head should be estimated by using 6?7_thickness in place of the scalp and skull, and 2 mm in place of the cerebrospinal fluid.

Neuromagnetogram tracing (magnetoneurography). The torso or limbs can be fitted using an infinite half-space model, and the surface of the torso or limbs should be tangent to the edge between the air and the body.

b) Reali

For more product details, please visit Aladdin Scientific website.