**NB  no figs at present **

Rotatory scanning for improved discrimination of very low frequency components of the magneto-encephalogram. Gardner-Medwin AR (1992). Biomagnetism '91: Clinical aspects. Ed. Hoke M., Erne S.N., Okada Y.C. & Romani G.L. Excerpta Medica: Elsevier, Amsterdam, pp.881-885

Rotatory scanning for improved discrimination of very low frequency components of the magneto-encephalogram
A.R. Gardner-Medwin
Department of Physiology, University College London, London WC1E 6BT, UK


 Clinical magneto-encephalography (MEG) signals that rise and fall with a time course of minutes would be of great interest, because they may arise from pathological ionic disturbances in the brain. Such signals have been recorded from anaesthetised rabbits during the ionic disturbance known as Spreading Depression [1]. Similar signals may occur in man in conditions such as concussion, stroke, and migraine, all of which involve profound and slowly changing disturbances of local neurological function. Direct evidence of ionic disturbances is not available from other non-invasive techniques in these conditions.
 Conventional MEG is poor at discriminating very slowly changing signals in the relevant frequency band: 0.001-0.1 Hz. Baseline stability can be poor, environmental noise is substantial, and screening is less effective and more expensive than at higher frequencies. Faster changing signals have been observed associated with Spreading Depression in vitro [2], but the conditions in vivo, and in clinical conditions, may tend to produce mainly the slowly changing signals [1]. Some signals have been recorded in clinical conditions [3]. Such data are not amenable to averaging to improve noise rejection, in the manner possible with experimental signals elicited by repeatable stimuli [1]. It would therefore be of great benefit to improve the baseline stability and artefact rejection in low frequency studies to assist comparison between signals in clinical situations and experimental tissue.


 Source movement is a well known strategy allowing measurement of DC fields [4,5]. The technique described here imposes a continuous rotatory scanning motion on the source or subject in a horizontal plane under the magnetometer (Fig.1). This converts horizontal gradients of DC fields into AC fields at the rotation frequency (typically 0.5-1 Hz). Noise levels and interference at this frequency are much less than at the frequencies of direct interest [6]. DC or slowly changing sources produce an extremely narrow frequency band of signals of interest, as shown with Fast Fourier Transform (FFT) analysis of the signals with steady calibration sources of different strengths (Fig. 2). This makes it possible to use special data processing techniques to extract the low frequency information from just this narrow band of the spectrum, thereby rejecting most of the noise at other frequencies.
 The technique is being developed using portable equipment, with the aim that it may be used, especially for clinical research, wherever there are the best combinations of patients and magnetometer facilities for particular projects.
 Up to 14 channels of magnetometer data and 2 channels of position data (monitoring the rotatory scanning in 2 perpendicular directions) are fed to a Data Acquistion Board (Microstar DAP 2400) in a portable computer (Mesh LCD-386). A correlation is carried out on the DAP board between each channel of magnetic data and each position channel. Since the position signals normally vary sinusoidally, each correlation is equivalent to extraction of a Fourier component of the magnetic signal at the rotation frequency. The correlation procedure is preferable, because it does not require prior knowledge of the rotation frequency and is tolerant of frequency variations and irregularities in the motion.

Figure 1. Rotatory scanning. A baseplate is supported on plastic balls and is driven by a motor at 3 m distance via a fibreglass rod and gears and belts (not shown). The radius of movement is 10 mm at a frequency 0.5-1 Hz. Note that the source maintains constant orientation: it does not spin around. All parts of the source and baseplate have the same amplitude of motion.

Figure 2. FFT spectra of SQUID magnetic signals with a 24 mm sensor coil 15 mm above the centre of the circular scanning trajectory (10 mm radius, 0.9 Hz) of a calibration current dipole. Four spectra are shown with the indicated dipole strengths. A small magnetic signal at the rotation frequency is seen even with no current, due to magnetic contamination.

 Substantial data compression is achieved on the DAP Board. For example, with 10 magnetometer channels, 4800 Bytes/s (17 MB/hr) of sampling data are reduced to typically only 66 Bytes/s (240 kB/hr) of output to the computer for display and storage. This includes all the essential information about the low frequencies: at the end of each rotation cycle, the computer is sent (for each channel) the mean directly measured field (in pT) and the two regression slopes (in pT/mm) representing the gradients of the field due to the moving sources, in the directions of the two position sensors.


 The AC field arises from horizontal gradients of the DC field. To a first (linear) approximation, which is adequate for most practical purposes, the AC field at the rotation frequency is proportional to the spatial gradients of the DC field due to sources on the moving platform. The DC field and the resulting AC field over a current dipole source are shown in Fig. 3. The second graph can be regarded as the steepness (regardless of direction) of the surface formed by the first graph.

Figure 3. a. The DC vertical field (Bz) on a horizontal plane at a height h above a current dipole of strength Q in the y direction. (Units: h for x,y and (?o/4?)Qh-2 for Bz). b. The peak-to-peak AC field as in (a), with horizontal rotatory scanning of radius a (<<h). (Units: (?o/2?)aQh-3). Graph (b) is 2a times the absolute magnitude of the gradient of (a).

 The peak-to-peak amplitude of the AC signals is comparable to the maximum amplitude of the DC signals. The ratio for a<<h is 5.2 a/h, and the peak AC and DC signals are equal for a/h=0.2 (e.g. 10mm radius, 50mm source distance).
 The AC fields measured with this technique have three notable features, apart from their benefit in noise reduction:
1. The maximum signal occurs directly over a dipole source.
2. The phase of the maximum signal directly indicates the dipole orientation.
3. There is improved rejection of signals due to distant magnetic contamination on the subject. This is because the field gradients that are measured fall off more steeply with distance than the fields themselves. This effectively increases the differential order of the gradiometer.


Trial tests with calibration sources confirm the qualitative improvements in noise rejection that can be achieved in low frequency studies. Fig. 4 shows the output regression signals due to a 0.04Hz calibration source, switched on for part of the time. These signals are essentially unaffected by interference that caused large shifts in the directly measured field (top trace). The ratio of the two regression signals indicates the dipole orientation. The X regression trace gives a clear measure of the dipole timecourse. The presence of a dead rabbit in the first part of the record produced only a small steady superimposed shift due to contamination. This trial was carried out with active cancellation of the Earth's magnetic field, but with no magnetic screening. The baseline stability of the regression records (referred to the size of dipoles that are detectable) is similar to that achieved with direct DC recording in a screened room [1], even though the ambient levels of interference were much greater.

Figure 4. Detection of a calibration dipole (monitored in the bottom trace) with rotatory scanning. At the start of the record, a dead frozen rabbit was positioned with its head directly under the sensor coil. The dipole source was 30 mm below the coil and 25mm off axis (next to the rabbit's ear). In the middle of the record, the rabbit was removed without disturbing the source.


The data shown here were obtained in the laboratory of SJ Swithenby, with help from him and K Fiaschi. I am grateful also to CN Guy for help and advice. Funded by the Wellcome Trust.


1 Gardner-Medwin AR, Tepley N, Barkley GL, Moran J, Nagel-Leiby S, Simkins RT, Welch KMA. Brain Res 1991;540:153-158
2 Okada YC, Lauritzen M, Nicholson C. Brain Res 1989;442:185-190
3 Tepley N, Barkley GL, Moran J, Simkins RT, Welch KMA. In: Williamson SJ, et al., eds. Advances in Biomagnetism. New York: Plenum, 1990;327-330
4 Cohen D, Kaufman LA. Circulation Res 1975; 36:414-424
5 Freake SM, Swithenby SJ, Thomas IM. In: Atsumi K, et al., eds. Biomagnetism '87, Tokyo: Tokyo Denki UP, 1988;434-437
6 Katila TE. (1990) In: Williamson SJ, et al., eds. Advances in Biomagnetism. New York: Plenum, 1990;19-32