NelLab Logo

Neuroscience of Language Lab
at New York University


MEG System

The MEG system was provided by the Kanazawa Institute of Technology (Tokyo, Japan) as part of a collaboration with its Applied Electronics Laboratory. The same system is also developed and marketed by Yokogawa Electric Corporation (Yokogawa, Japan) as the MEG Vision. It is a recumbent system, with the dewar fixed in position. We refer to our system as having 160 channels, but in actuality it contains:

  • 157 axial gradiometers used to measure brain activity
  • 3 orthogonally-oriented (reference) magnetometers located in the dewar but away from the brain area, used to measure and reduce extramural noise offline
  • 32 open positions, of which we currently use 8 to record stimulus triggers

The MEG or SQUID (Superconducting Quantum Interference Device) sensors must be kept at superconducting temperatures in order for the system to operate. To achieve this end, the dewar requires ~100L of liquid helium per week.

Data from the MEG system and other equipment is acquired by a DAQ system assembled by Eagle Technology. The system has six DAQ PCs and one control PC. Each signal is digitized with 16-bit resolution. For raw data recording we use a sampling rate of 1 kHz. We can record with a higher sampling rate but the length of acquisition time that the machine is capable of is greatly reduced as the sampling rate increases. All inputs of the 192 channels can be displayed simultaneously by the six CRT monitors. The control PC runs Windows XP.


For all data acquisition and MEG system control and maintenance, we use the MEG laboratory software MEG160. This software is produced jointly by Yokogawa Electric Corp., Eagle Technology Corp., and the Kanazawa Institute of Technology (KIT). The principal author of the software has been Yasuhiro Haruta, of Yokogawa Electric Corp.

Noise reduction

For offline noise reduction using the three reference magnetometers, we employ the MEG laboratory software MEG160 (see Software/Recording), which uses algorithms developed at the Applied Electronics Laboratory at KIT, and specifically the Continuously Adjusted Least-Squares Method (CALM). This noise reduction procedure essentially eliminates any correlation that the data from the MEG sensors might have with any of the three reference magnetometers by removing any detected covariance. This is performed data point by data point, with a moving window typically of six seconds, such that for each data point the three seconds of data before and the three seconds of data after are used to determine the covariance. This process was designed to eliminate low frequency (<10Hz) noise at Keio University in Tokyo, Japan, where a KIT MEG machine is also located. There, as here in New York, large extramural magnetic noise from passing trains/subways can greatly disrupt MEG sensors.


The Room itself

The magnetically shielded room (MSR) is a product of Vacuumschmelze (Hanau, Germany). The shielding effect is provided by two layers of mumetal, the inner layer being 3mm and the outer layer being 2mm. Predicted shielding performance was of -60dB at 1Hz; actual performance exceeds this prediction. The exterior dimensions of the room are 2.9 x 3.5 x 2.9m, and the inner dimensions are 2.4 x 3.0 x 2.4m.

Active shielding

In addition to the offline CALM noise reduction system described above, we also have an active shielding system ("active compensation coils") designed by Vacuumschmelze and incorporated into the MSR. The system senses the incoming (extramural) signal and produces signals of compensating voltages to annul as much of that signal as possible. The system consists of:

  • 6 compensating coils, one on each surface of the MSR
  • A three-axis fluxgate sensor that detects the disrupting extramural noise
  • A controller that tunes the fluxgate sensor, digitizes the reference (input) signal, calculates the canceling (compensating) voltages, converts these voltages to analog signals and feeds these voltages to the compensating coils

Because the dewar/MEG sensors are not centered within the room, the precise locations of the compensation coils were adjusted by Dr. Jochen Bork of Vacuumschmelze so as to maximize the cancellation effect.

Other shielding

RF/EM shielding

To shield our cables proceeding from the MSR, we used Kitagawa Industries 300-series cable shielding.

To shield miscellaneous small parts, we used 3M copper foil tape (1mil/.00254mm), products 1181 and 1182 (1" and 2" wide, with single and double-sided conductive adhesive, respectively). Our local distributor for 3M was Bristol Tape Corp. (Fall River, MA).


We record the location of the marker coils and electrodes on the subjects, as well as three fiducial locations (the nasion, and the left and right preauricular points), before MEG recording takes place. Additionally, we record the shape of the subject’s head using a 3D laser scanning system, which generally allows 50,000-200,000 points to be collected. The hardware used for the digitization process is a Polhemus (Colchester, VT) FastSCAN COBRA 3D laser system. One receiver is placed on the subject during digitization and a transmitter is placed behind the subject. The use of the receiver on the subject permits accurate correction for subject movement.

We use MATLAB software to convert the file generated by the FastSCAN software to the format required by our analysis methods for source localization.

A Brain Vision (Morrisville, NC) system is used for the monitoring/recording of spontaneous EEG signal from ocular sites (used in the artifact rejection process) and for the recording of electric evoked potentials (eEPs) simultaneously with the magnetic evoked potentials (mEPs) that form the basis of the lab's research capabilities.

Brain Vision System

We have a DC 32-channel system with a DC-capable 32-channel digital EEG amplifier with a USB interface. The BrainAmp DC Amplifier and the PowerPack are set up inside the MSR during recording. These two devices (which are near the subject's feet, as far from the dewar as the cable from the EEG cap will allow) are connected to the BrainAmp USB2 Adapter via a fiber optic cable. This USB2 Adapter is outside the MSR and connected, via a USB cable, to the PC used for the EEG signal acquisition.

Electrodes/Electrode caps

For both spontaneous and evoked potentials, we use electrode systems from EASYCAP GmbH (Herrsching, Germany). All electrodes are sintered Ag/AgCl Multitrode flat disks. For evoked studies, we use custom-fabricated caps. These caps are standard in that they contain 29 Multitrode B18-LU-150-MEG compatible electrodes in a standard 10-20 system array, as well as 3 Multitrode B18-HU-200-MEG electrodes for monitoring ocular signals, one Multitrode B18-LU-150-MEG ground electrode, and one Multitrode B18-LU-150-MEG Ref electrode.

Recording notes

For both artifact monitoring and full cap evoked recording, we use two references, located on the left and right mastoid. The right mastoid is typically the site used as reference during data acquisition; electrode signals are then re-referenced to the left mastoid site during offline averaging.

Electrode impedances are kept below 6kΩ. To assess impedance, we use the EIM-107 Prep-Check Plus multi-lead impedance meter from General Devices (Ridgefield, NJ). We have found this to be a very handy device, being easily programmable to test at various thresholds and in various sequences.

Another helpful tool has been ASP (Irvine, CA) CIDEX OPA, which we use for disinfecting electrodes and caps. This solution is much less volatile and corrosive than other disinfecting agents, and does not require a specially vented fume hood for use.

Bioamplifier System for EOG recordings

A bioamplifier system is used for the monitoring/recording of EOG signals (from ocular sites, used in the artifact rejection process) simultaneously with magnetic evoked potentials (mEPs).

Bioamplifier headbox

Our 24-channel bioelectric amplifier and headbox are products of SA Instrumentation Co. (Encinitas, CA). The amplifier is electrically and magnetically isolated, and run from a DC power source. The headbox accepts loose-lead electrodes. We can also utilize electrode caps for this system inside the MSR.

Much of the video system was provided either wholly or partially by Nancy Kanwisher and her lab at MIT. Special thanks to Nancy and her father, and Liu Jia; thanks also to Dr. Alison Harris, formerly an undergraduate working in Nancy's lab, and now a PhD recipient from Harvard University in the Vision Sciences Lab.

Image generation

Our visual stimuli are computer-generated. The images generated on the computer monitor are mirrored by an InFocus LP850 projector. The LP850 uses Digital Light Processing (DLP) technology developed by Texas Instruments (Warren, NJ). This technology immediately transmits the image drawn on the LCD within the projector to its mirror-and-lens system, meaning that there is very little lag between the receipt of image information and transmission of the image (a large concern for the temporal accuracy of MEG experiments). The native resolution of this projector is 800 x 600, and our experiments either employ this resolution or most likely a 1024 x 768 resolution.

Image resolution

We have replaced the original lens system with several wide-angle camera lenses to produce a focused image size of approximately 18cm x 18cm (7" x 7") at a distance of 24cm (9.5") from the participants. The distance of the image from the projector is approximately 2.4m (8'). The image also enters the MSR horizontally, and must be reflected down. We accomplish this using a first surface mirror from Edmund Scientific (Barrington, NJ). This mirror is suspended from the ceiling and is fixed at a 45° angle; the incoming image hits the mirror and is reflected 90° straight down.

The image comes to rest on a ground glass "screen." The screen is held in a frame built by Nancy Kanwisher's father out of old fiberglass supports from a windsurfer. It isn't pretty, but it's functional. What is slightly unusual about the setup is that the participant is essentially viewing the back of the screen, rather than the front of the screen (and what we usually think of when we see anything project onto a screen). This means that the image is natively "backwards"; in order to fix this we use the rear projection option on the LP850 to "flip" the image.

The audio system is not especially fancy, but the routings are complex: we must be able to get auditory stimuli to the participant in the MSR, to speak to the participant through the audio system (the MSR is rather strongly sound-dampening), to hear the participant outside the MSR, and, if the experiment calls for a participant voice response, to use the participant's voice as a trigger.


To talk to the subject, we use a standard unidirectional microphone that must be turned off at the start of experiments to prevent feedback. To hear the subject, we have installed an Audio-Technica (Stow, OH) ATM10a Omnidirectional Condensor Microphone using a pass-through in the MSR; since the microphone is functionally inside the MSR, a condenser mike is necessary to eliminate RF noise.


Input from the subject mike is fed into a Eurorack (Behringer; Willich, Germany) UB502 mixer preamplifier. Here we adjust the gain, in particular boosting the amplification of the subject mike. The mike we use to communicate to the subject is fed into a MACKIE (Woodinville, WA) 1202-VLZ3 Premium mike/line mixer. There are six inputs and four outputs, and the levels on each can be adjusted separately. Also feeding into the MACKIE mixer are the lines for sound generated by the stimulus computers for auditory experiments and experiments providing auditory feedback.


Output from the MACKIE mixer is sent to three destinations. The first is a mix of the experimenter's mike and the stimulus computer's output that goes to the subject, via E-A-Rtone (AEARO Company; Indianapolis, IN) 3A insert earphones. Two 50 Ohm amplifiers for the E-A-Rtone system located just outside the MSR, and sound is fed to the subject via silicon air tubes approximately 1.5m (5') in length. The second is the stimulus output that goes to the outside speakers where the experimenter is located. Finally, the auditory stimuli may be sent directly to the trigger box to record stimuli onset on a trigger line of the MEG system. We can use two additional channels to record any combination of the subject, the experimenter, and the computer-generated stimulus, if, for example, we should want to examine off-line the response produced by the subject.

Subject output is also sent directly and uniquely from the Eurorack UB502 mixer preamplifier to a set of close by speakers so that we can monitor the subject without hearing the computer output. We also have an OptoAcoustics (Or Yehuda, Israel) fiber optic microphone that we use to record voice onset triggering, by running it through an amplifier and then directly to the trigger box via a BNC line.


At present, all of our experimental stimuli are administered from an Apple iMac computer utilizing Psychophysics Toolbox, which operates with MATLAB software (The MathWorks, Inc.; Natick, MA), or from a PC running Presentation (Neurobehavioral Systems, Inc.; Albany, CA) or DMDX software. We also have an Apple Mac Mini running PsyScope X software. For the iMac and PC, stimuli triggering is run through a StimTracker Model ST-100 (Cedrus Corporation; San Pedro, CA). For PsyScope, all timing for experiments, along with all responses made during experiments, is managed by a Response Button Box developed by ioLab Systems. Both of these devices have millisecond resolution accuracy, and are not susceptible to the same drag that Mac and PC system "clocks" are. Responses by the subject are triggered either by voice (See the OptoAcoustics system [link]) or button-press from a Current Designs (Philadelphia, PA) fiber optic system. The actual peripheral response buttons in the MSR are two 2-button response pads (a total of four buttons) utilizing a photodiode system that sends signals directly to the stimulus computer via a USB plug. The interface between the buttons and the computers is a 904 FIU series photodiode box.


Stimuli are normally administered via Psychophysics Toolbox, which operates with MATLAB or, if from the PC, via Presentation or DMDX software. We use this for both auditory and visual experiments. PsyScope X is set up on the Mac Mini but is rarely used. The PC can also run Psychophysics Toolbox but currently this software cannot utilize the response pads in the MSR.