Measurement of Somatosensory Evoked Magnetic Fields Using Adjustable Magnetoresistive Sensor Array

An adjustable helmet-style magnetoresistive (MR) sensor array, a room-temperature magnetic ﬂux sensor, was developed to demonstrate simultaneous multipoint measurement of the somatosensory evoked magnetic ﬁeld (SEF). The SEF was measured in three healthy subjects using an array of 30-channel MR sensors placed on the left hemisphere of the head with median nerve stimulation in the right wrist and averaged over 8,000 measurements. An M20 component (maximum amplitude of 725 ± 257 fT, peak latency of 20.5 ± 0.45 ms), considered to originate from the primary somatosensory cortex, was observed at an approximate latency of 20 ms of the magnetic ﬁeld waveform in all cases. The phase inversion observed around C3 corresponded to the palmar area of the primary somatosensory cortex on the contour map of the magnetic ﬁeld at the M20 peak. The MR sensor, a relatively inexpensive and easy-to-use magnetic sensor that does not require a zero-ﬁeld environment, was successfully used for simultaneous multipoint SEF


Measurement of Somatosensory Evoked Magnetic Fields Using Adjustable Magnetoresistive Sensor Array
Tetsuro Tatsuoka*, Shigenori Kawabata, Jun Hashimoto, Yuko Hoshino, Kensuke Sekihara, Life Fellow, IEEE, Tomohiko Shibuya, Yoshiaki Adachi, Member, IEEE, and Atsushi Okawa Abstract-An adjustable helmet-style magnetoresistive (MR) sensor array, a room-temperature magnetic flux sensor, was developed to demonstrate simultaneous multipoint measurement of the somatosensory evoked magnetic field (SEF).The SEF was measured in three healthy subjects using an array of 30-channel MR sensors placed on the left hemisphere of the head with median nerve stimulation in the right wrist and averaged over 8,000 measurements.An M20 component (maximum amplitude of 725 ± 257 fT, peak latency of 20.5 ± 0.45 ms), considered to originate from the primary somatosensory cortex, was observed at an approximate latency of 20 ms of the magnetic field waveform in all cases.The phase inversion observed around C3 corresponded to the palmar area of the primary somatosensory cortex on the contour map of the magnetic field at the M20 peak.The MR sensor, a relatively inexpensive and easy-to-use magnetic sensor that does not require a zero-field environment, was successfully used for simultaneous multipoint SEF measurements in humans and may be a promising option to realize a reasonable magnetoencephalography application device.Index Terms-biomagnetic sensors, magnetoencephalography, magnetoresistive element, somatosensory evoked magnetic fields.

I. INTRODUCTION
AGNETOENCEPHALOGRAPHY (MEG), which records weak magnetic fields derived from electrical activity in the brain, has been widely used in clinical applications such as focal epilepsy and brain science research as it demonstrates excellent temporal and spatial resolution and can noninvasively identify active areas from outside the body [1]- [3].
Conventional magnetoencephalographs (MEGs) use superconducting quantum interference device (SQUID) magnetometers as magnetic sensors, which require cooling with liquid helium, resulting in high equipment and maintenance costs [4], [5].
MEG or evoked MEG, which relies on room-temperature magnetic sensors such as optical pumping magnetometers (OPMs) or tunnel magnetoresistance (TMR) sensors, was recently reported [6], [7].Room-temperature magnetic sensors offer a high degree of freedom in terms of arrangement as they T. Tatsuoka was with the TDK Corporation, Tokyo, Japan and is currently with METOOL Inc., Tokyo, Japan.
S. Kawabata, J. Hashimoto, Y. Hoshino, K. Sekihara, and A. Okawa are with Tokyo Medical and Dental University, Tokyo, Japan.
T. Shibuya is with the TDK Corporation, Tokyo, Japan.
do not require a cryostat to store liquid helium.The sensors are placed in close contact with each participant.As the magnetic field originating from the magnetic source decays with the square of the distance, shortening the distance from the signal source improves signal-to-noise ratio [8], [9].The magnetic sensor developed in this research utilizes the magnetoresistive (MR) effect.
In this study, we conducted simultaneous multipoint measurements of the somatosensory evoked magnetic field (SEF) using a sensor array consisting of multiple magnetic sensors.The device utilizes MR effect elements and an adjustable fixture that allows the sensors to be positioned in close contact with the head.

II. METHOD
The SEF was measured in a magnetically shielded room (MSR) using a helmet-shaped sensor array with 30 MR sensors (Nivio xMR sensor, TDK, Tokyo, Japan).
Figure 1 shows the appearance and noise density spectrum of the magnetic sensor, and Table I lists the main characteristics of the MR sensor.
The sensor had a cuboidal shape, 12 mm square, and 74 mm in length, with a noise density ranging from 3 pT Hz -0.5 at 1 Hz.
The MR element is a giant magnetoresistive (GMR) element.The sensitivity axis of the sensor was located along the longitudinal direction of the cuboid, and the magnetic field was recorded in the positive direction from the end face of the sensor to the cable.The sensor consisted of a four-element Wheatstone bridge to reduce the influence of temperature drift.
Inside the sensor, a rod-shaped ferrite flux concentrator was magnetically coupled with the MR elements to improve the sensitivity.A compensating coil was wound around the flux concentrator to expand the dynamic range and improve the linearity through feedback control.The structure that realizes a compact magnetic sensor is patented [10].
The sensor array is characterized by a structure wherein multiple sensors are radially mounted on a helmet-shaped sensor holder fixed on an aluminum frame.The position of each sensor can be adjusted in the radial direction.The internal Y. Adachi is with the Kanazawa Institute of Technology, Ishikawa, Japan.*Corresponding author, T. Tatsuoka (e-mail: tatsuoka@metool.co.jp).Color versions of one or more of the figures in this article are available online at http://ieeexplore.ieee.orgM dimensions of the sensor holder were 188 mm in the left-right direction, 104 mm in the superior-inferior direction, and 238 mm in the anterior-posterior direction.The position of each sensor could be adjusted within approximately 20 mm by sliding it along the axial direction inside the cylindrical structure attached to the helmet-type sensor holder.The sensor's position could be fixed using immovable grub screws located at four points on the side as shown in Fig. 2.
Thirty sensors were placed corresponding to the left hemisphere from the top of the head, such that the center was near C3 in the international 10-20 method as shown in Fig.  were determined using a spherical calibration coil [11] for SQUID-based whole-head MEG.
The room wherein the measurements were conducted was magnetically shielded by two layers of permalloy and two layers of aluminum with an attenuation of −32 dB at 1 Hz.
The subjects wore marker coils at five points on the left side of the head to position it relative to the sensor array, and their heads were inserted into the sensor array in the supine position as shown in Fig. 4.
The operator adjusted the sensor positions such that all sensor tips were in close contact with the subjects' heads.
The median nerve was electrically stimulated at the right wrist joint in three healthy subjects, and the SEF was measured from the area centered around C3 of the international 10-20 method, the contralateral side of the stimulation (with an averaging count of 8000).The demographic data of the participants are presented in Table II.The measurement and signal processing conditions are listed in Table III.The filter settings were decided referring to the recommendation from the International Federation of Clinical Neurophysiology (IFCN) for somatosensory evoked potentials (SEP) [12], with high-pass filtering below 3 Hz and low-pass filtering of 2 kHz or higher.
After the SEF measurement, a sinusoidal current was applied to the marker coils to generate a known magnetic field.The sensor array detected the magnetic field from the marker coils, whose positions were estimated via magnetic source analysis to reveal the position and orientation of the subject's head relative to the sensor array.Upon completion of the measurements, the subject removed their head from the helmet, and the calibration coils were placed in the helmet.Magnetic field signals from the calibration coils were recorded, whereupon the 3D position coordinates of each sensor were identified based on the calibration coil.The position of the marker coil was estimated based on the information about the sensor position.
This study was approved by the Ethical Review Committee of Tokyo Medical and Dental University Hospital.(Approval number: M2021-178)

III. RESULTS
The SEF measurement results from the three subjects are shown in Fig. 5.The plot on the left indicates superimposed waveforms of magnetic field signals from all sensors.The plots in the middle shows the waveforms of the sensors that included the highest positive/negative peaks around a latency of 20 ms.The peak at about 20 ms in latency was observed in all three subjects.The latency and amplitude of the peak were 20.5 ± 0.45 ms and 725 ± 257 fT, respectively.
The contour map on the right side indicates the magnetic field distribution at the peak latency in the observation area illustrated at the top of Fig. 5.The contour map of the magnetic field was arranged as if looking at C3 from the front, with the parietal side at the top, left temporal side at the bottom, frontal side to the left, and occipital side to the right.

A. Validity of the M20 component as the observed signal
On the contour map of the magnetic field at about 20 ms in latency shown in Fig. 5, a phase reversal was observed in all cases, wherein the signal was distributed from outward to inward of the head on the parietal side and inward to outward on the left temporal side.This suggests that this component originates from a current from the occipital to the frontal area near C3 and is considered as the M20 component originating from the primary somatosensory cortex.This is consistent with the physiological findings in preceding studies [13]- [15] that the M20 component is estimated as a forward current dipole in the hand area of the primary somatosensory cortex localized in the posterior wall of the central sulcus.

B. Sensor placement with respect to the head surface for superior stability
In the third subject, the position of the marker coil attached to the head was estimated before and after the SEF measurement.The displacement of each marker coil before and after the measurement was below 0.72 mm, suggesting that the change in the head position during the measurement was negligible.In conventional SQUID-based MEGs, the helmet-   shaped sensor array used to record MEG has a fixed size regardless of the subject, which may cause the head to move because of the gap between the head and sensor array.In contrast, the sensors in the sensor array used in this study maintained close contact with the head on both sides throughout the measurement, and the helmet-shaped sensor holder was fixed to a heavy aluminum frame, which provided excellent head position stability.Sensors that are inadequately fixed introduce artifacts caused by body motions when the head strikes the sensors.During our measurements, the sensor was firmly fixed to the sensor holder with screws to prevent it from being affected by body-motion artifacts.

C. M20 amplitude compared with that using SQUID sensors
In SEF measurements using SQUID sensors with similar filter conditions, the maximum M20 amplitude was approximately 300 fT [14], [16]- [18].The amplitude of 725 fT (519-1,134 fT) in this study was larger than that of the SEF measurements using SQUID sensors.This is because the SQUID sensor was located farther from the signal source owing to the thickness of the vacuum layer in the cryostat, whereas the MR sensor was located closer to the signal source [7], [19].

D. Comparison with OPM
Several reports indicated that the SEF was detected by the MEG measurement based on OPM, another type of highly sensitive magnetic sensors recently applied to biomagnetic measurements other than SQUIDs, at a higher signal-to-noise ratio than that obtained in this study.The MR sensor still has the following superiorities compared with the OPM.

1) Low-cost production
MR sensors can be manufactured relatively inexpensively using existing manufacturing processes similar to those used for hard disk drive heads with an already established mass-production process.Despite installing a flux concentrator, the structure remains simple, and the manufacturing cost can be reduced.
2) Usable in a simple magnetically shielded room While the OPM requires a zero-magnetic field environment or magnetic field compensation by a built-in coil inside the sensor [5], [20], [21], the MR sensor has a wide dynamic range and can be used in a −32 dB MSR in this study.The maximum environmental noise level was 32 nT even in the MSR as the experimental setting was located in an urban area.An MR sensor can measure the magnetic field of a living body even at considerably high noise levels without any tuning operation.
3) Low power consumption Typical commercially available OPM products have a power consumption of 5 W (sensor head = 0.7 W) [22].Low power consumption of the sensor head is an indispensable feature for building further multi-channel systems in the future.
As aforementioned, MR sensors are currently limited in their applications because of the restriction of magnetic field resolution; however, in certain cases, these sensors may be a promising option for realizing a reasonable MEG application device.For example, there are reports of neuromagnetic field measurements using SQUID sensors to measure the magnetic fields in the spinal cord and peripheral nerves [23], [24].MR sensors could be effective in measuring the SEF and such magnetoneurographic (MNG) signals by SQUIDs simultaneously.To obtain the SEF signals by MR sensors, several thousands of repetitive electrical a plural stimulations were necessary.However, in the case of simultaneous recording of the MEG and MNG signals especially from the spinal cord, it is acceptable as the signal detection from the spinal cord at the neck or back also requires averaging for several thousand times to achieve a sufficient signal-to-noise ratio, despite using SQUIDs, owing to the small signal intensity.

E. Comparison with preceding studies
There have already been reported cases wherein the M20 component of SEF has been measured using MR sensors [6].In the previous study, multisite measurements were performed using a one-channel sensor and by changing its position.However, in this study, the magnetic field distribution was demonstrated by simultaneously measuring the magnetic field using 30 sensor channels with the same performance.The advantages of simultaneous multisite measurements are that the measurement time is shortened to 1/(number of channels) and that the magnetic field source of spontaneous signals can be analyzed, and noise reduction can be applied using the multidimensionality of the signal space.
In a previous study [20] that measured the magnetic fields originating from the cerebrum using multiple sensors that could operate at room temperature, a whole-head mask tailored to a specific individual was produced using a 3D printer.However, the advantage of the array used in this study is that the sensor positions can be adjusted, thereby making it applicable to multiple subjects, and the position and orientation of each sensor were readily revealed by the measurement of the fiducial magnetic field from a spherical calibration coil after the SEF measurement.

V. CONCLUSION
Using a helmet-shaped 30-channel MR sensor array, the SEF of three healthy subjects was measured with a median nerve stimulation at the right wrist joint and averaging of 8000 epochs.In all cases, an M20 component originating from the primary somatosensory cortex was confirmed at approximately 20 ms in the magnetic field waveform, and a phase inversion was observed around C3 on the contour map of the magnetic field.The MR sensor is a relatively inexpensive and easy-touse magnetic sensor that does not require a zero-field environment and can be used as an "instant-on" device without any parameter adjustment before measurement.The MR sensor may be a promising option for realizing a reasonably priced MEG device.

Fig. 2 .
Fig. 2. Mechanism of sensor adjustment.Each sensor was moved in the radial direction and fixed with a screw.

Fig. 3 .
Fig. 3. Appearance and structure of the sensor array.Thirty MR sensors were installed at a position corresponding to the left hemisphere from the parietal of the cerebrum in a helmetshaped sensor holder attached to an aluminum frame.

Fig. 5 .
Fig. 5. (a), (b), and (c) Magnetic field waveforms and the contour map of the magnetic field obtained from subject 1, 2, and 3, respectively.Left side: Magnetic field waveforms.Right side: Isomagnetic field diagram and maximum amplitude waveforms.