MEASUREMENT OF SLOWLY CHANGING MAGNETIC SOURCES IN THE BRAIN
A.R. Gardner-Medwin1, T. Elbert2 and Z. Kovalik2
1Dept. Physiology, University College London, London WC1E 6BT, UK and
2Inst. Exptl. Audiologie, University of Münster, Germany
Slowly changing magnetic sources in the brain, rising and falling with a time course of minutes, can result from pathological ionic disturbances in the brain due, for example, to Spreading Depression (1). This paper describes progress and issues that arise in the application to human subjects of a technique (3,4) devised to improve the resolution of such very low frequency (0.001-0.1Hz) MEG recordings. It would be of value to obtain improved records from patients with conditions such as concussion, stroke, and migraine (2).
The signal to noise ratio in this ultra-low frequency band is mainly limited with conventional recording by two factors. Firstly, baseline stability is poor due to environmental noise (traffic, lifts etc.). Screening is less effective than at higher frequencies. Secondly, human subjects often have significant ferromagnetic sources in their bodies (particularly fingers, lungs, gut, dental implants) producing significant low frequency signals with even small body movements.
Technique. Source movement is well known as a strategy enabling measurement of DC magnetic fields. The present technique employs a continuous circular scanning motion of the subject in a horizontal plane under the magnetometer. The subject's mattress moves back and forth sinusoidally by 24mm in both horizontal directions (x,y) so that each part of the subject executes a circular motion 24mm in diameter. This converts horizontal gradients of DC fields into AC fields (Fig.2a) at the rotation frequency, typically 0.25 Hz in man. The AC signals are unaffected by slow baseline drift. Their amplitude and phase in relation to the motion, derived once per cycle for each recording channel with a digital processing board, gives a measure of the size and distribution of the field gradients due to DC sources (3). Measured noise levels and interference at this frequency are roughly 30 times less (in terms of amplitude per root bandwidth) than at 0.001 Hz, and the signal to noise ratio for measuring very slow signals is correspondingly improved.
Rejection of signals due to contamination. The potential signals of interest due to sources within the brain are of the order of tenths of 1pT, with gradients of tenths of 1pT/cm or less (1). Signals 100 or 1000 times larger are often encountered in the neighbourhood of contamination sources. Large signals from the teeth probably cannot be eliminated, rendering subjects with certain types of dental work unusable for low frequency field measurements.
Fig.1. Calculated maximum signals due to magnetic dipoles at varying horizontal distances from a perfectly balanced vertical magnetometer. Full lines: 2nd order gradiometer (baseline 65mm) as at Milton Keynes; dashed lines: 1st order gradiometer (baseline 50mm) as at Münster. Light (upper) lines: direct field measurement; heavy (lower) lines: gradient measurements with horizontal scanning movements. For each distance (x) the dipole is assumed 100mm below the sensor coil, with orientation giving the greatest signal. Values are normalised with the signal for a dipole at x=0. Calibration measurements made on the two instruments agree within a factor of 2 with the calculated curves.
Sources in the chest, abdomen and limbs are sufficiently distant that the signals from them can be sufficiently attenuated at the head. These sources move with the scanning motion in the same way as the head, so their effects are not eliminated as they are for stationary sources by the scanning technique. If the contamination sources are strictly constant, they contribute constant field gradients that constitute a baseline on which slow changes due to disturbances in the brain would be visible. However, slow changes in the contamination sources due to respiration, patient movement due to discomfort with prolonged immobility and gut movement are inevitable and their effect must be minimised. Fortunately the signals fall off rapidly with distance, at a rate that depends on technique (Fig. 1). With a 2nd order gradiometer (full lines) the sensitivity to ferromagnetic contamination falls off more rapidly than with a first order gradiometer (dashed lines). With both instruments use of the rotatory scanning technique (heavy lines), rather than direct DC measurements, produces a more rapid fall off of the contamination signals. The best combination is a 2nd order gradiometer with the scanning technique (heavy full line), for which the signals fall off approximately as distance-6. This combination would minimise the problem of artefact from body contamination.
Measurement of fields due to light adaptation in the eye. The continuous current generated by the pigment epithelium behind the retina changes with a slow time course following an increase in light level, rising to a peak about double its baseline level in the dark approximately 8 minutes after light onset and then falling gradually (5, 6). The current leaves the eye through the cornea, generating a positive corneo-scleral potential. The rapid changes during eye movements of the associated magnetic fields have been measured (7), and it has been shown that these changes vary in amplitude with the same time course in light adapatation as the electrical measurements. The field itself, with stable eye position, has approximately the same amplitude and time course as the signals anticipated from ionic disturbances in superficial brain tissue. It was first measured in Milton Keynes using the scanning technique with a single channel 2nd order gradiometer in an unscreened environment in collaboration with S. Swithenby and K. Fiaschi (4). We have now made measurements with a 37-channel BTi magnetometer (1st order gradiometers, in a mu-metal shielded room).
We used a subject selected for low contamination (out of 6 tested). The Dewar vessel was positioned with ca. 15mm clearance over the side of the head with sensor coils centred over the approximate sites shown in Fig. 2B (mean head positions during the scanning motion). Fig. 2A shows samples of the field variations during the scanning cycles. For each sensor the absolute value of the field gradient (in the direction with maximum gradient) was calculated in each cycle from the peak to peak AC amplitude. These measurements are shown throughout the session in Fig. 3A along with the mean field during each cycle (Fig.3B). The subject was dark adapted for 10min before the start of recording, with light onset 3min after the start. The gradients increased to a maximum 50%-100% above baseline levels depending on position, about 7min after light onset and then declined towards baseline. The largest changes were measured at the sensors over the temple (Fig. 2B). The time course is similar to the established time course of the changes in corneo-retinal current after light onset (5). The mean fields during each cycle (Fig.3B) correspond to measurements made with conventional techniques, since they are hardly affected by the scanning motion. They exhibit drift and fluctuations both before and after light onset that obscure the changes due to light adaptation. The gradient measurements (Fig. 3A) are unaffected by these shifts, except immediately after light onset when the electrical disturbance due to switching the lights on caused a shift to saturation on some channels. The saturation was rectified 2min later by resetting the BTi magnetometer system (open triangles).
Fig. 2. Magnetic recordings during light adaptation. A: DC records of
fields at 37 sensor sites as shown in (B), as the subject was scanned in
a 24mm diameter circular track (anti-clockwise, 0.25Hz) under the magnetometer.
Records are shifted relative to their mean (dotted line). B: Vector representations
of the field gradients measured from records as at (A), showing maximum
gradient measured in individual cycles (peak-to-peak AC signal /
scan diameter) and its orientation (direction of increasing upward vertical field) inferred from the timing of the maxima as in (A). Each vector starts from the point corresponding to the centre of its sensor coil. The vector orientation for the central sensor was measured by correlation of the signal with measurements of the x,y horizontal movement on a separate computer that performed on-line analysis but could only accept signals from a few channels. Vector orientations for other channels are inferred from the timings of maxima relative to these measurements on the central channel. Channels correspond in A,B and in Fig. 3 with rotation in Fig. 3B to bring those indicated by * into register.
The magnetometer coils measure approximately vertical fields (ignoring, for purposes of interpretation, the slight tilt of peripheral coils in the BTi instrument). The gradients of the vertical field can be measured with the scanning technique in any direction in the horizontal (x,y) plane. Fig. 3A shows only the absolute horizontal gradients, along the direction for which they are greatest, equivalent to ?((dB/dx)2+(dB/dy)2). The orientation information is retained in a vector presentation (Fig. 2B), showing the gradients superimposed for several cycles around the time of peak amplitude following light onset. The fanning of these vectors (due to the variability of successive measurements) shows graphically the direction and amplitude of the steepest gradient at each sensor position, as it were "looking" in the direction of increasing upward magnetic field. The pattern of the largest vector gradients around the temple is approximately that over a clockwise loop of current in this region, as for example due to current out of the eye, upward over the forehead, back through skull and brain. The pattern suggests that a larger fraction of the current leaving the cornea returns via long pathways in this upward direction than through the lower nasal and oral regions, possibly due to different tissue resistivities. Much of the current may return via shorter pathways within the orbit itself, which by virtue of the near symmetry of the patterns would have little magnetic effect.
Theoretical considerations suggest that the ideal instrument for mapping
DC sources would include a large array of 2nd order gradiometers. The data
show however that even 1st order gradiometer arrays are effective, with
the scanning technique, for monitoring and mapping field gradients due
to slowly changing magnetic sources.
Fig. 3. Field measurements during light adaptation. A: Absolute values of horizontal gradient of the measured (approximately vertical) fields during each cycle of the scanning movements of the subject. Gradients were inferred as described at Fig.2. B: simultaneous measurements of the mean field during each cycle, shifted relative to an arbitrary baseline. Time of light onset is shown by a full triangle on the top records. At this time most channels showed a large DC shift and many went into saturation due to the electrical transient. The saturation was corrected 2min later (open triangle). For Fig. 3B arbitrary shifts were made in the records at these two times to bring them on scale and approximately to equivalent heights. No such adjustments were made to the gradient measurements in Fig.3A, which were only transiently affected by the DC shifts, except where they were rendered unmeasurable (shown as zero) on the saturated channels.
Acknowledgements. Supported by the Wellcome Trust. Much of the development of the scanning technique was carried out with collaboration of S. Swithenby and K.Fiaschi (Dept. Physics, Open University, Milton Keynes, UK).
1. Gardner-Medwin AR, Tepley N, Barkley GL, Moran J, Nagel-Leiby S, Simkins RT, Welch KMA (1991) Brain Res 540:153-158
2. Barkley GL, Tepley N, Nagel-Leiby S, Moran JE, Simkins DO & Welch KMA (1990) Headache 30: 428-434
3. Gardner-Medwin AR (1992) In: Biomagnetism '91. Clinical aspects. Hoke M, Erne SN, Okada YC, Romani GL eds.
4. Gardner-Medwin AR (IN PRESS) In: Proc. Symposium om Migraine - Basic Mechanisms and Treatment, ed. A. Lehmenkühler, Urban & Schwarzenberg, München
5. Arden GB & Kelsey JH (1962) J.Physiol.161: 189-204
6. Nilsson SEG (1991) In: Heckenlively J.R. & Arden G.B., eds.: Principles and practice of clinical electrophysiology of vision. Mosby Year Book pp. 328-333
7. Katila T, Maniewski R, Poutanen T, Varpula T & Karp PJ (1980) J.appl.Physics: 52:2565-2571