Sensory perception during sleep in human

Although M50 and M100 showed no definite change during the two stages of sleep, M150 and M200 were significantly enhanced in both sleep stages ðp , 0 :05Þ: Peak latency of M150 and M200

Sensory perception during sleep in humans:

a magnetoencephalograhic study Ryusuke Kakigia,b,*, Daisuke Nakac, Tomohiro Okusaa, Xiohong Wanga,b, Koji Inuia, Yunhai Qiua,b, Tuan Diep Trana, Kensaku Mikia,b, Yohei Tamuraa,b, Thi Binh Nguyena,b,

Shoko Watanabea, Minoru Hoshiyamaa,d

a

Department of Integrative Physiology, National Institute for Physiological Sciences, Myodaiji, Okazaki 444-8585, Japan

b Department of Physiological Sciences, School of Life Sciences, The Graduate University for Advanced Studies, Okazaki 444-8585, Japan

c

Department of Neurological Surgery, Wakayama Medical University, Wakayama 641-8510, Japan

d

Department of Health Sciences, Faculty of Medicine, Nagoya University, Nagoya 461-8673, Japan Received 19 March 2003; received in revised form 12 May 2003; accepted 14 May 2003

Abstract

We reported the changes of brain responses during sleep following auditory, visual, somatosensory and painful somatosensory stimulation

by using magnetoencephalography (MEG) Surprisingly, very large changes were found under all conditions, although the changes in each were not the same However, there are some common findings Short-latency components, reflecting the primary cortical activities generated

in the primary sensory cortex for each stimulus kind, show no significant change, or are slightly prolonged in latency and decreased in amplitude These findings indicate that the neuronal activities in the primary sensory cortex are not affected or are only slightly inhibited during sleep By contrast, middle- and long-latency components, probably reflecting secondary activities, are much affected during sleep Since the dipole location is changed (auditory stimulation), unchanged (somatosensory stimulation) or vague (visual stimulation) between the state of being awake and asleep, different regions responsible for such changes of activity may be one explanation, although the activated regions are very close to each other The enhancement of activities probably indicates two possibilities, an increase in the activity of excitatory systems during sleep, or a decrease in the activity of some inhibitory systems, which are active in the awake state We have no evidence to support either, but we prefer the latter, since it is difficult to consider why neuronal activities would be increased during sleep

q2003 Elsevier B.V All rights reserved

Keywords: Magnetoencephalography; Sleep; Auditory; Visual; Somatosensory; Pain; Electroencephalography

1 Introduction

One is, of course, not aware of auditory, visual,

somatosensory and pain stimuli while asleep, but such

signals should reach the brain nonetheless It is important to

know what happens in the brain when stimulation is given

during sleep and how brain responses following auditory,

visual and somatosensory stimulation differ from the

waking state to that of sleep To investigate the brain

responses to a stimulus, several non-invasive methods are

used Recent advances in neuroimaging now enable such

responses to be monitored, even during sleep To examine

the changes in cortical activity during the early processing

of stimulus perception, within 100 or 200 ms after stimulus onset, electroencephalography (EEG) and magnetoence-phalography (MEG) are the most appropriate methods; their temporal resolution is very high, in the order of milli-seconds, as compared with other neuroimaging techniques such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) whose temporal resolution is over 1000 ms

MEG has theoretical advantages over EEG for detecting source localization because there is less effect from current conductivity caused by cerebrospinal fluid, skull and skin; the spatial resolution for MEG is only a few millimeters, but that for EEG is much larger Therefore, we studied the effects

of sleep on brain responses by analyzing evoked magnetic fields following auditory (AEF) [1], visual (VEF) [2], 1389-9457/$ - see front matter q 2003 Elsevier B.V All rights reserved.

doi:10.1016/S1389-9457(03)00169-2

www.elsevier.com/locate/sleep

* Corresponding author Address: Department of Integrative Physiology

National Institute for Physiological Sciences, Myodaiji, Okazaki 444-8585,

Japan Tel.: þ 81-564-55-7765x7779; fax: þ 81-564-52-7913.

E-mail address: kakigi@nips.ac.jp (R Kakigi).

somatosensory (SEF)[3]and painful somatosensory

(pain-related SEF)[4,5]stimulation The objective of this article is

to review our results, most of which are the first to report the

effects of sleep on evoked magnetic fields following each

kind of sensory stimulation Comparison of MEG with EEG

results (auditory, visual, somatosensory, and painful

soma-tosensory evoked potentials, AEP, VEP, SEP and

pain-related SEP, respectively) was often very useful

2 MEG system

Unlike EEG, MEG is not well known The magnetic

fields recorded from the human brain are very small,

approximately ten thousand to a million times smaller than

the earth’s steady magnetic field and environmental fields

(caused by a train, for example) A superconducting

quantum interference device (SQUID) is necessary to detect

these weak fields We use dual 37-channel axial-type

first-order biomagnetometers (Magnes, Biomagnetic

Technol-ogies Inc (BTi), San Diego, CA) The waveforms from 74

channels are thus simultaneously recorded The detection

coils of the biomagnetometers are arranged in a uniformly

distributed array in concentric circles over a spherically

concave surface All of the sensor coils are thus equally

sensitive to the brain’s weak magnetic signals Each device

is 144 mm in diameter with a radius of 122 mm The outer

coils are 72.58 apart, and each coil is connected to a SQUID

The spacing between the centers of the coils is 22 mm The

coils are 20 mm in diameter and have a 50 mm baseline

Recordings were performed in a magnetically shielded

room (MSR, Vacuumschmelze GmbH)

The placement of probes for our MEG was based on the

international 10/20 EEG system Dual probes centered at

positions C3 and C4 were used to cover the left and right

hemisphere for recording AEF and SEF, and at position Oz

to cover the occipital region for recording VEF A spherical

model[6] was fitted to the digitized shape of the head of

each subject, and the location (x; y and z location),

orientation and amplitude of the best-fitted equivalent

current dipole(s) (ECD) were estimated at each time point

The origin of the head-based coordinate system was the

midpoint between the preauricular points The x-axis

indicated the coronal plane with a positive value toward

the left preauricular point, and the z-axis lay on the

transverse plane perpendicular to the x – y line with a

positive value toward the upper side The correlation

between the recorded measurements and the values expected

from the ECD estimate was calculated as a measure of how

closely the measured values corresponded to the theoretical

field generated by the model and the observed field

Magnetic resonance imaging (MRI) was performed using

a Magnex 150XT 1.5T system (Shimadzu, Kyoto, Japan)

T1-weighted coronal, axial and sagittal images with a

contiguous 1.5 mm slice thickness were used for overlays

with ECD sources detected by MEG The same anatomical

landmarks used to create the head-based three-dimensional (3D) coordinate system (the nasion and bilateral preauri-cular points) were visualized in the MR images by affixing

to these points high-contrast cod liver oil capsules (3 mm diameter) The common MEG and MRI anatomical land-marks allowed easy transformation of the head-based 3D coordinate system (nasion and entrance of the auditory meatus of the left and right ear) used for the MEG source analysis to the MRI

3 Issues common to all studies All 22 subjects who participated in this study were researchers and technicians in our department (six females and 16 males; mean age 29.7 years, range 27 – 45 years), who were very familiar with MEG studies None of the subjects had any history of otological, ophthalmological, or neurological abnormalities or sleep disorders Some number

of the 22 subjects participated in each study All subjects were right-handed Informed consent was obtained from all participants prior to the study, which was first approved by the Ethical Committee at our institute Experiments were performed in the daytime, usually after lunch No medication was given to the subjects before the experiment

A somnogram was obtained for all subjects by recording EEG with the use of gold disk electrodes (1 cm diameter) which were placed at the Cz, Pz and Oz positions Three channels were used for the bipolar recording of EEG between Cz – Pz, Cz – Oz and Pz – Oz The sleep stage was determined according to the guidelines of Rechtschaffen and Kales[7]using the continuously recorded somnogram data We carefully observed the subject’s movements through a TV monitor during the experiments, and confirmed the position of the head using the head location sensor When a movement was large, we discarded the data Since most subjects did not reach stage 3, stage 4 or rapid eye movement (REM) sleep and body movements during stage 3 and stage 4 were sometimes very large, we analyzed results in the awake state (AW), stage 1 sleep and stage 2 sleep in each study All experiments were done at least twice

to confirm the reproducibility of results

For obtaining clear responses, 100 – 200 responses were averaged Trials in which the MEG deflection exceeded 3 pT were excluded from the averaging We measured the latency, amplitude and ECD location for each component The amplitude of MEG was calculated using the maximum amplitude of the outgoing and ingoing magnetic fields of each component The latency and ECD location were measured at the time of maximum amplitude in each component To determine the difference in effect between

AW and the sleep stages (stages 1 and 2) more clearly, we assigned a value of 100% to the amplitudes of the waveforms

in AW in some studies For example, if the amplitudes of the waveforms in AW and stage 1 sleep were 100 and 80 fT, respectively, the amplitude ratio was 80%

Two criteria were used in the application of the ECD

model to a magnetic response First, during the period of the

response, the location of the dipole determined by the ECD

model must remain stationary (within 5 mm of each

coordinate) Second, the correlation between the recorded

magnetic fields and values expected from the dipole

estimate must have a value larger than 0.95

The differences of latency, amplitude (or amplitude ratio) and ECD locations among AW and the sleep stages (stages 1 and 2) were statistically analyzed using a one-way factorial analysis of variance (ANOVA) with Bonferroni – Dunn’s correction or Fisher’s protected least significant difference (PLSD) test for multiple comparison Significance was accepted at the 0.05 level in all statistical analyses

Fig 1 AEF during the awake state and sleep stages 1 and 2 A pure tone of 250 Hz was delivered to the right ear and the magnetometer was placed on the left hemisphere (position C3) Thirty-seven superimposed waveforms at each stage for subject 1 are shown Four main components (M50, M100, M150 and M200) were identified in each stage Although M50 and M100 showed no definite change during the two stages of sleep, M150 and M200 were significantly enhanced

in both sleep stages ðp , 0 :05Þ: Peak latency of M150 and M200 to 250 Hz tone during sleep did not show a significant change as shown in Table 1 , and they appeared shortened during sleep in this subject However, M50, M100 and M200 were significantly prolonged to 1000 or 4000 Hz ( Table 1 ) Adopted from Ref [1]

4 Auditory evoked magnetic field (AEF)

4.1 Methods

Ten normal volunteers were studied A pure tone of 70 dB

sound pressure level (SPL) 70 ms in duration (including rise

and fall time, 10 ms each) was delivered to the right ear at a rate

of 1 Hz through a soft plastic tube Three different frequencies

of pure tone, i.e 250, 1000 and 4000 Hz, were presented in

the awake state and in each sleep stage (sleep stages 1 and 2)

White noise of 50 dB SPL was delivered to the opposite (left)

ear through the same soft plastic tube during the recordings

The recordings were performed during naturally

occur-ring daytime sleep in a magnetically shielded room with the

subject lying on a bed Responses were filtered with a 0.1 –

200 Hz bandpass filter and digitized at a sampling rate of

1041.7 Hz The analysis time was 100 ms before and

1000 ms after the application of the pure tone stimuli We

set the amplitude ratio of each component in AW as 100%

4.2 Results

We focused on the results recorded from the left

hemisphere, contralateral to the stimulated ear Four

components were identified with latencies of less than

250 ms at each frequency tone (Fig 1) The latency of each

component in AW, S1 and S2 following 250, 1000 and

4000 Hz tone stimuli was approximately 50 – 60, 95 – 110,

150 – 165 and 200 – 215 ms, and we termed them M50,

M100, M150 and M200, respectively

The sleep stage affected the waveforms of each of the

four components Therefore, to avoid confusion, we use the

following abbreviations to show each component in each

condition: AW for awake, S1 for stage 1 and S2 for stage 2,

for example, ‘M50-1000 Hz-AW’ means the M50

com-ponent of AEF following the 1000 Hz pure tone stimulation

in the awake state

M50-1000 Hz-S2 ðP , 0:05Þ; M100-1000 Hz-S2 ðP ,

0:01Þ; M200-4000 Hz-S1 ðP , 0:05Þ and M200-4000 Hz-S2

ðP , 0:05Þ showed a significant prolongation from those

components in AW (Table 1) Some other components

showed a similar tendency, but the changes were not

significant, probably due to a large inter-individual

differ-ence For example, M150 and M200 to 250 Hz in subject 1

appeared shortened during sleep (Fig 1)

To more clearly identify the effects of sleep on

amplitude, we used the amplitude ratio (Table 1) In

general, with an increase of sleep stage, there was a

tendency for amplitude reduction in the early-latency

components (M50 and M100) and amplitude enhancement

in the late-latency components (M150 and M200)

M150-250 Hz-S1, M150-250 Hz-S1, M150-M150-250 Hz-S2,

M200-250 Hz-S2, M200-1000 Hz-S2, M200-4000 Hz-S1,

M150-4000 Hz-S2 and M200-4000 Hz-S2 showed

signifi-cant enhancement compared with the same components in

AW (Table 1, Fig 2) Although M150-1000 Hz-S1,

M200-1000 Hz-S1, M150-1000 Hz-S2 and

M150-4000 Hz-S1 showed a tendency for enhancement, their changes were not significant

The ECD location of every component was estimated for each stimulus frequency and each sleep stage Every ECD in all conditions was estimated to be around the superior temporal cortex (the primary and secondary auditory cortex) (Fig 3)

Concerning the effect of sleep stage on ECD location, in the early-latency components (M50 and M100), the M50 component had a tendency to be located in a more anterior, lateral and superior region in the left primary auditory cortex when the sleep stage was deeper As for the M100 component, the ECDs tended to be located in a more anterior, medial and superior region in accord with the depth

of sleep However, the only components to show a significant change were M100-1000 Hz-S1 ðP , 0:05Þ and M100-1000 Hz-S2 ðP , 0:01Þ; which were more superior along the z-axis than M100-1000 Hz-AW (Table 1) In the late-latency components (M150 and M200), there was no consistent tendency for change in ECD location between each sleep stage

Concerning the effect of stimulus frequency on ECD location, the y coordinate (medio-lateral direction) showed smaller values when the delivered frequency was increased, that is, the ECDs of high-frequency tone stimuli were located more medially than those of low-frequency tone stimuli in all sleep stages (Fig 3)

Table 1 Peak latencies, amplitudes and ECD locations (x; y; z values) of identifiable components in the left hemisphere following a pure tone stimulus of three different frequencies (250, 1000 and 4000 Hz) applied to the right ear in the awake state (AW) and sleep stages 1 (S1) and 2 (S2)

Peak latency (ms)

Peak amplitude ratio (%) based on results in AW as 100%

ECD location (cm): z-axis

Only the results that showed a significant difference are given Values are shown as the mean (standard deviation) Statistical analysis was done using an one-way factorial analysis of variance (ANOVA) with Bonferroni – Dunn’s correction for multiple comparison *P , 0 :05;

**P , 0 :01; compared with AW.

4.3 Discussion

This is, to our knowledge, the first report of AEF changes

during sleep indicating a physiological change of auditory

perception The most important and interesting findings of

the present study are the changes of waveforms and ECD

location of each component during sleep

There was a reduction in the amplitude of M50 and M100

and an enhancement of M150 and M200 In a study of cats,

Chen and Buchwald[8]reported a ‘wave A’ recorded from

the vertex 20 – 22 ms in latency It decreased in amplitude or

disappeared during slow wave sleep (SWS), but dramatically

increased in amplitude or reappeared during the REM stage

The amplitude during the REM stage was equal to that while

awake This positive potential in the cat showed the same characteristics as M50 in the present study Furthermore, Steriade et al.[9] described that the discharge rates of the ascending reticular formation neurons in cats were increased during the awake state and REM sleep, but decreased during SWS From the results of these animal experiments and the AEF amplitude changes in the present study, we speculate that the amplitude reduction of the early-latency components (M50 and M100) during sleep is caused by a decrease in the sensitivity of the primary auditory processing system which may be controlled by the ascending reticular formation

As for the long-latency components, Wasensten and Badia[10]reported that the amplitude and latency of N200, probably corresponding to M200 in the present study,

Fig 2 Changes of amplitude ratio (% index) of each component recorded from the left hemisphere in response to a 250 Hz tone delivered to the right ear in both sleep stages (sleep stages 1 and 2) compared with a corresponding component in the awake state (100%) The statistical analysis comparing the awake state with both sleep stages was performed with a one-way factorial analysis of variance (ANOVA) with Bonferroni – Dunn’s correction for multiple comparison The M150 and M200 components of both stages of sleep were significantly enhanced compared with those of the awake state (probability,

*P , 0 :05) Adopted from Ref [1]

increased during sleep, and the amplitude was highest in

sleep stages 3 and 4 These observations suggest that

the generator systems may be different from those of the

early-latency components functionally and anatomically

The enhancement of amplitude for the M150 and M200

components during sleep in the present study might be due

to a decrease of inhibitory activity of the secondary auditory

processing system, for example the cognitive processing

system at the cortical level

The location of the ECD of each component moved in

the medial direction with an increase of tone frequency

during both the awake state and sleep This result was

compatible with those in previous studies of the spatial

tonotopic organization in monkeys [11] and humans [12]

during the awake state The ECD location of the

early-latency components, M50 and M100, were located at more

anterior and superior sites during sleep than AW In

contrast, the ECD location of the late-latency components,

M150 and M200, during sleep did not show such consistent

differences compared with that during AW Therefore, the

mechanisms generating the late-latency components may

be different from those behind the early-latency

components

5 Visual evoked magnetic field(s) (VEF)

5.1 Methods

Nine normal subjects were studied The center of the

array was set around the Oz position covering the visual

cortex of the subjects The head was fixed to the

biomagnetometer with adhesive tape to prevent movement

The subjects wore a bandage on their right eye and were

instructed to close their eyes throughout the experiment,

even when awake To mask the noise caused by the strobe

light, rubber earplugs were provided and white noise of

50 dB was delivered from a loud speaker during

the experiment Responses were filtered with a 0.1 –

200 Hz bandpass filter and digitized at a sampling rate of 1041.7 Hz The analysis time was 100 ms before and

300 ms after the application of the flash stimuli The flash stimulus was delivered with a light stimulator (SLS4100, Nihonkohden, Japan) every 2 s throughout the session The light was placed at a distance of 2.5 m from the eyes of the subjects The direction of light was about 408 below the line

of fixation of the subjects While the subjects slept for 40 –

60 min, 5 – 11 trials were recorded for various sleep stages After the subjects awoke, the awake VEF were recorded During the recording of the awake VEF, the subjects were required to keep their eyes closed

To examine the similarity between the magnetic components of awake and asleep, contour maps of the isomagnetic field were used The contour maps show a topographical distribution of the magnetic flux density on the surface of the scalp We calculated the correlation coefficient across 37 channels between contour maps of sleep and awake VEF, and judged the similarity to be high when the correlation value was greater than 0.52 (P ¼ 0:001; degree of freedom ¼ 35) Even if the peak latency and amplitude differed, similar contour maps indicate common generating mechanisms

5.2 Results

Fig 4shows the VEF waveforms for AW, sleep stage 1 and sleep stage 2 in four subjects, showing an inter-individual difference.Fig 5shows the nomenclature of each identifiable component and their isocontour maps in two subjects The waveforms for the awake condition showed multiple peaks from 40 to 200 ms, that is, A1 (41.4 ^ 4.5 ms, mean ^ standard deviation (SD)), A2 (54.2 ^ 4.3 ms), A3 (63.5 ^ 5.0 ms), A4 (76.2 ^ 6.4 ms), A5 (93.7 ^ 8.2 ms), A6 (112.1 ^ 6.5 ms), A7 (140.0 ^ 8.8 ms) and A8 (186.0 ^ 9.6 ms) A1 to A3 composed the first deflection of the VEF waveform, and

Fig 3 Location of ECDs of the M100 component observed in response to three different frequencies (250, 1000 and 4000 Hz) in subject 1 The magnetometer was placed on the left hemisphere (C3 position), and ECD overlapped on MRI (coronal slices) in each sleep stage All ECDs were located in the superior temporal cortex (the auditory cortex) However, ECDs were located deeper in response to the higher frequency tones than lower frequency tones Adopted from Ref [1]

A4 to A6 the second A7 and A8 were late and slow

deflections These components were considered to generally

correspond to the flash visual evoked potential (VEP)

components reported by Ciganek [13] In contrast, the

waveforms during sleep had fewer peaks, which were late in

latency and large in amplitude We identified three peaks,

S1 (64.2 ^ 3.5 ms in stage 1 and 65.5 ^ 3.5 ms in stage 2),

S2 (90.2 ^ 7.3 ms in stage 1 and 100.5 ^ 20.3 ms in

stage 2) and S3 (113.7 ^ 12.8 ms in stage 1 and

115.5 ^ 14.5 ms in stage 2) in the sleep VEF from the

root mean square (RMS) time course S1 was not identified

in some subjects (i.e subject 2) S2 overlapped S3, and in

some subjects S2 seemed a notch peak on the larger S3

component The longer latency components (later than

130 ms) for the sleep conditions were much weaker than S2

and S3, and not clearly identifiable

Because waveforms of the awake and sleep VEF differed

markedly (e.g subject 2), we used maps of the isomagnetic

field (contour map) to find correspondence between

the eight components for the awake VEF and the three for the sleep VEF The correlation coefficient across 37 chan-nels between each pair of contour maps was used to measure the similarity between the maps Fig 5shows the contour maps at each peak of A1 – A8 and S1 – S3 in two subjects Based on the correlation coefficiency, A2, A5 and A6 were considered to correspond to S1, S2 and S3, respectively Other components in the awake VEF were not significantly correlated with the components in the sleep VEF

After matching the components, we compared their peak latency and peak amplitude values The latencies of S1, S2 and S3 increased for A2, A5 and A6, respectively The difference was significant between S1 and A2, and between S3 and A6 (P , 0:05; by the paired t-test with Bonferroni– Dunn’s correction) The latencies were longer in stage 2 than stage 1, though not significantly The amplitudes of these components also showed a tendency to increase with sleep stage The difference was significant between A5 and S2, and between A6 and S3 ðP , 0:05Þ:

Fig 4 VEF waveforms for the awake state, sleep stage 1, and sleep stage 2 The data from two subjects are shown A flash was delivered at 0 ms on the horizontal axis A sharp large deflection at around 0 ms due to stimulus artifacts was excluded from the figure The inter-individual difference was large in the awake VEF, but the sleep VEF was rather similar among subjects, with one large peak at about 80 – 100 ms The VEFs were similar between stages 1 and 2, though the peak amplitude and latency increased for stage 2 Adopted from Ref [2]

The signal source for the VEF was estimated using an

equivalent current dipole (ECD) model The ECDs were

located in the primary visual cortex (V1) for the early

components in the awake VEF (A1 – A6), and three

components in the sleep VEF (S1 – S3), but the estimation

was unsuccessful for the later components This was

probably due to the complicated activities generated in

multiple regions in the visual cortex

5.3 Discussion

To our knowledge, this is the first study on the changes in

VEF during sleep, although there have been several studies

on the changes in VEP during sleep in neonates In fullterm

neonates, a positive component with a latency of about

200 ms was clearly identified while the subjects were

awake, but it disappeared during sleep [14] In preterm

neonates, a negative component at about 300 ms was

prominent in the awake state, but its amplitude was

markedly reduced during sleep[15] These studies suggest

that the change in VEF due to sleep occurs in the early

stages of development However, the VEP waveforms in the

neonates were so different from those in adults that it is

difficult to compare them with the present results

In the present study, the early component (A2) showed a

slight prolongation of latency, and the middle latency

components (A5 and A6) showed an enhancement of

amplitude As in the AEF study [1], the decrease in the

amplitude of the early components might be due to

a decrease in the sensitivity of the primary processing system, and on the other hand, the increase in the late components might be due to the decrease in inhibitory activity Although some components of the awake VEF were identified in the sleep VEF, others were not Thus several components (e.g A3, A7 or A8) of the awake VEF were considered to have disappeared, or to overlap with the other components, with the latency shift during sleep The reason for the change is not clear, but both the change in the neuronal activity of the lateral geniculate nuclei (LGN) and the change in the activity of the cortex are considered to cause the VEF change Neuronal recordings in animals have revealed that the mode of activity of neurons in the LGN differed between those asleep and those awake [16 – 18] The change in the relaying neurons is thought to affect the VEF during sleep The multiple components of the awake VEF are considered to be partially due to the construction of the visual cortex, which is composed of a network of more than 30 sub-cortical areas[19] V1 is reported to be the main signal source of VEP and VEF[20,21]and receives visual information not only via the LGN but also from the higher extrastriate cortical areas[19] The disappearance of the late components (A7 and A8) during sleep might reflect the change in the activity of the higher area Lamme et al

[21] reported that when a monkey was anesthetized the modulatory activity of the V1 neurons (later than

100 ms) decreased, which might be related to feedback from the extrastriate cortex The activity pattern is considered to change during sleep, resulting in the change in the VEF

Fig 5 Contour maps for identified components of the awake and sleep VEF are shown in two subjects (subjects 1 and 2) Vertical lines superimposed on the waveforms indicate the latencies at which the contour maps were drawn In the sleep VEF, not only maps for S1 to S3, but also those for 45, 75 ms and so on were drawn for comparison In each contour map, areas of the same magnetic flux density are drawn in the same color (see the inset) Red and pink indicate the flux going into the head, while blue indicates that coming out from the head The corresponding components are indicated with arrows, with the correlation coefficiency value between the contour maps Adopted from Ref [2]

6 Somatosensory evoked magnetic fields following

non-painful (SEF) and painful stimulation

(pain-related SEF)

6.1 Methods

Eight normal subjects were studied The electrical

stimulus was a constant current square wave pulse delivered

transcutaneously to the left index finger using ring electrodes

The stimulus duration was 1.0 ms Two levels of intensity

were adopted, ‘painful’ and ‘non-painful’ The degree of pain

was about 80% of intolerable pain, depending on the

subjective feeling of each subject, and its intensity ranged

from 14 to 16 mA The level of non-painful stimulation was

three times the sensory threshold, approximately 3 – 4 mA

The inter-stimulus interval was random, between 1500 and

5000 ms Basic methods were based on previous studies

[23– 26] Two measurement matrices were centered around

the C3 and C4 of the international 10 – 20 system in each

subject The primary and secondary somatosensory cortices

(SI and SII, respectively) in the bilateral hemispheres were

mostly covered by these positions

SEF responses were filtered with a 0.1 – 100 Hz bandpass

filter and digitized at a sampling rate of 1042 Hz The

analysis time was 100 ms before and 500 ms after the

application of each stimulus

A single dipole spherical model was estimated for each

point in time [6] Since the measured field in several

subjects was considered to contain two or more temporally

overlapping sources (as will be described in the results) the

single ECD model was inappropriate in such cases

Therefore, for the spatio-temporal multi-dipole model, we

used brain electric source analysis (BESA 99, NeuroScan,

Inc., McLean, VA) for the computation of theoretical source

generators in a three-layer spherical head model BESA was

modified for our 37-channel biomagnetometer[27 – 29] The

residual variance (%RV) indicated the percentage of data

that could not be explained by the model The

goodness-of-fit (GOF) was expressed as a percentage (100 2 %RV) The

GOF indicated the percentage of the data that can be

explained by the model, which is different from the

correlation value calculated in a single ECD model We

considered the adaptation of the dipoles to be significant

when the GOF was larger than 90%

6.2 Results

During the awake state, five consistent components

were identified from the hemisphere contralateral to the

stimulated finger in both non-painful and painful sessions

We termed them 1M, 2M, 3M, 4M, and 5M (Figs 6 and 7)

The peak latency and amplitude of each component are

shown inTable 2 The peak latency of each component did

not differ between the two stimulus conditions Although

1M, 2M and 3M, which peaked around 20, 35 and 45 ms,

respectively, did not differ in amplitude between painful

and non-painful stimulation, the amplitudes of 4M and 5M, which peaked around 80 and 160 ms, respectively, were significantly larger in the painful sessions ðP , 0:02Þ: Using

a single dipole model, ECDs for 1M, 2M and 3M were located in the posterior bank of the central sulcus, probably the SI, following both non-painful and painful stimulation The isocontour map of 4M was much different from the maps of 1M – 3M, indicating that different mechanisms are involved The ECD for 4M was localized in the upper bank

of the Sylvian fissure, probably in the SII, following both non-painful and painful stimulation (Figs 6 and 7) Since isocontour maps of 5M showed a complicated dipole pattern

in both non-painful and painful sessions and the ECD could not be reliably estimated by a single dipole model, 5M was considered to contain two or more sources

In the hemisphere ipsilateral to the stimulation, two components, 4M(I) and 5M(I) whose latencies were longer than those of 4M and 5M by approximately 10 – 15 ms, were identified in both stimulus conditions (Fig 7) Amplitudes

of these components were significantly larger for painful stimulation ðP , 0:02Þ (Table 2) The correlation value for 4M(I) was over 97% following both painful and non-painful stimulation The ECD for 4M(I) was localized to SII The correlation value for 5M(I) was relatively low following both painful and non-painful stimulation, but the ECD of 5M(I) was estimated in SII in all subjects

During sleep, the amplitude of 1M and 2M was slightly increased and decreased, respectively, but these changes did not reach the significant level ðP 0:05Þ: By contrast, the amplitude of 3M ðP , 0:05Þ; 4M ðP , 0:005Þ and 5M ðP ,

0:005Þ were significantly decreased during sleep (Table 2) The two components detected in the ipsilateral hemisphere, 4M(I) and 5M(I), were significantly decreased in amplitude

ðP , 0:005Þ during the sleep stages in a similar manner to 4M and 5M (Table 2)

For the analysis using the multi-dipole model, BESA, we found that a two-dipole model placing source 1 and 2 at SI and SII, respectively, was the most appropriate In the time range up to 50 ms, it was considered that only S1 was activated from the results of the single dipole model, and the period from the onset of the 4M component to the offset of the 5M component was used as the analysis period for BESA The GOF value was not reliably high (65.5 ^ 25.6%) with the SII source alone, but it markedly and reliably increased when the SI source was added (91.4 ^ 1.0%) We could not make a better model than this one Two components were identified in the SII source, an early and late component, and the source orientation of the former and latter was reversed The strength of both the early and late components of the SII source was significantly decreased during sleep ðP , 0:001Þ (Table 3,

Fig 8) The source strength of both the early and late components was reduced in stage 2 sleep compared to stage

1 sleep, but the difference was not significant Peak latencies

of early and late components were prolonged along with

Fig 6 SEF following painful and non-painful electrical stimulation of the index finger in the awake state, and stage 1 and 2 sleep Waveforms recorded from 37 channels in the hemisphere contralateral to the stimulation are superimposed During sleep, 1M and 2M showed no significant change, 3M was reduced in amplitude, and 4M and 5M were much reduced Isocontour maps of 2M and 4M at the peak latency following painful stimulation while awake Thin lines show fields directed out of the head, dotted lines those into the head, and thick lines the zero fields Isocontour lines are separated by 20 fT The ECD of 2M and 4M was estimated to lie in the SI and SII, respectively Adopted from Ref [4]