Electroencephalography and clinical Neurophysiology , 83 (1992) 146-152 © 1992 Elsevier Scientific Publishers Ireland, Ltd. 0013-4649/92/$05.00
146
E E G 91094
Late magnetic fields and positive evoked potentials following infrequent and unpredictable omissions of visual stimuli Robert L. Rogers, Andrew C. Papanicolaou, Stephen B. Baumann and Howard M. Eisenberg Magnetoencephalography Laboratory, Division of Neurosurgery, Unicersity of Texas Medical Branch, and Transitional Learning Community, Galceston, TX (U.S.A.) (Accepted for publication: 30 March 1992)
Summary Randomized and infrequent omissions during presentation of a steady train of visual stimulation produced distinctive wave forms of both the magnetic fields and electrical potentials. Electrical potentials at Pz showed a positive peak in response to the omitted stimuli which occurred on the average 445 msec after the time when a stimulus was anticipated. Analyses of the magnetic wave forms indicated that at least two separate sources appear to be active coincident with the electrical positive peak. O n e source localized in the occipital lobes in the vicinity of the visual cortex while the other source was located in the medial aspects of the temporal lobe or even deeper in the lateral thalamus. Judging from the calculated direction of current flow it appeared that the deep source would contribute greater potentials in the frontal areas of the scalp while the source in the occipital area would contribute to more posterior placement of electrodes, especially at Pz. Key words: Visual stimulation; Magnetic field; Evoked potential; Electrical potential; Pz
Infrequent deviations or omissions of sensory stimuli in the visual, auditory or somatosensory modalities result in an average wave form which differs from evoked responses obtained from a steady train of stimulus presentations. The most prominent difference is the large positive peak which occurs after 300 msec (Squires et al. 1975). These late positive evoked responses have received much attention since their discovery because of their relative sensitivity to cognitive factors (e.g., attention, task relevance, etc.; see Donchin et al. 1978; Hillyard and Picton 1987) as opposed to stimulus parameters (e.g., intensity, duration, etc.). However, only recently have the underlying sources, responsible for generation of these late positive potentials, been systematically investigated. In order to help delineate the character of the source or sources responsible for the late positive responses, several divergent methods are being employed including: (1) depth electrode studies in humans (Stapleton and Halgren 1987; McCarthy et al. 1989) and animals (Halgren et al. 1980); (2) mapping of surface electrical potentials (Johnson 1989); (3) mapping of external magnetic fields (Okada et al. 1983; Gordon et
Correspondence to: Robert L. Rogers, Ph.D., Division of Neurosurgery E-17, University of Texas Medical Branch, Galveston, T X 77550 (U.S.A.). Tel.: (409) 772-3965.
al. 1987; Joutsiniemi and Hari 1989; Lewine et al. 1989; Rogers et ai. 1991); and (4) studying the effects of lesions on these evoked responses (Johnson 1989; Knight et al. 1989). The present study was undertaken in order to investigate the relationships between sources extracted from external magnetic fields and the electrical potential at Pz which is typically the area where the maximal late positive response is recorded.
Methods
Seven normal right-handed subjects were tested using an oddball paradigm in which visual stimulation was delivered every 600 msec and lasted for 50 msec. Visual stimulation consisted of 4 light emitting diodes forming a square arrangement of 6 cm on each side. Subjects were instructed to fixate on the center of the square to insure binocular stimulation. On 10% of the trials the stimulation was omitted on a randomized basis in which no fewer than 3 and no greater than 15 consecutive visual stimuli were presented between the omissions. Subjects were instructed to count the number of missing flashes. The on-line computer collected evoked potential and magnetic field (MEF) epochs 1800 msec long extending 600 msec prior and 1200 after the omitted stimuli. One hundred such epochs were averaged. Subjects were actually given 100-110
RELATIONSHIPS BETWEEN SOURCES FROM MEFs AND EPs
omitted stimulus trials so that they could not guess the number of stimulus omissions A 7-channel N e u r o m a g n e t o m e t e r (Biomagnetic Technologies, Inc., Model 607), using second-order gradiometers with a 4 cm baseline and 1.8 cm coil diameters, was used to measure MEFs. Seven to 9 placements (49-63 sensor recordings) were taken on each of the subjects (4 over the occipital area and 3 - 5 over the right hemisphere) each constituting a run. The order of magnetic probe placements was randomized across subjects. Additionally, electrical potentials were obtained at Pz, Oz and Fz and both horizontal and vertical eye movements were monitored. Both electric and magnetic signals were analog bandpassed between 0.3 and 50 Hz at a digitization rate of 200 Hz. Further off-line digital filtering bandpassed the averaged wave forms between 0.3 and 20 Hz. In order to reduce fatigue effects measurements were taken over 4 sessions of approximately 1 h each. Subjects were never given more than 3 runs in a single session and they were given at least 15 min between each run. In order to insure the integrity of the average magnetic wave forms, individual wave forms were inspected for: (1) eye movement artifacts, (2) variability about the peaks, and (3) distributions of variance. Individual runs which had noticeable eye movement intrusions into the magnetic wave forms were removed before averaging. This never exceeded 5% of the total number of wave forms. Variances of the magnetic wave forms were comparable to their electrical counterparts. Frequency histograms of the individual scores were normally distributed about the means. The location of the magnetic sensors in relation to the subject's head was determined using 3 lowfrequency magnetic field sensors placed on the head with a transmitter on the cryogenic container that housed the M E G sensors. Error tolerance for the location of the sensors was less than 3 mm. Source estimations were made on the basis of a model for a single dipolar source within a homogeneous sphere. Dipole p a r a m e t e r s (location in cartesian coordinates, strength, and orientation) were estimated using a finite difference version of the LevenbergMarquardt algorithm that c o m p a r e d the measured values of the M E F with an estimated M E F produced by a single dipole. The size and origin of that sphere were determined using a least-squares algorithm based on the digitized outer surface of the head of each subject, limited to the area in which probe placements were taken. Thus, for dipole fits over the occipital area only that area was digitized and fitted to a sphere and for fits over the right hemisphere that portion of the head was fitted to a sphere. Since the side of the head is much flatter than other areas, the fitted sphere would be much larger and would not necessarily be confined to the size of the subject's head. The average radii of
147
the model spheres was 8.8 cm for the occipital area and 9.4 cm for the right hemisphere. A single dipolar source produces a distinctive contour map with two extrema (maximum and minimum) around which it diminishes in strength. A measurement of goodness-of-fit of the estimated dipole parameters is obtained by correlating the field predicted by those parameters with the measured field. Acceptance of a good dipolar fit was set a priori at a level that accounted for at least 81% of the total variance (r = 0.90). The inverse solution was applied at every 5 msec interval in accordance with the maximum temporal
S #1
S #2
~., ~ / v
¥ -.~..~.
I
l
[
I
I
I
0
300
0
300
0
300
MILLISECONDS PRECEDING STIMULUS
OMITTED STIMULUS
FOLLOWING STIMULUS
Fig. 1. Electrical evoked responses recorded at Pz for all 7 subjects showing three 600 msec epochs, the epoch preceding the omitted stimulus, the omitted stimulus and the following frequent stimulus epoch. Positive potentials for electrical wave forms are represented as upward deflections.
148
R.L. R O G E R S ET AL.
resolution provided by the 200 Hz digitization rate for both the frequent and omitted stimulus intervals. This accounted for the sources of the fields throughout the duration of the recorded epoch that included the response to the omitted and the immediately preceding stimulus. Magnetic resonance images (MRIs) were obtained for each subject. The procedures and details of the MRI images and for transferring the M E G coordinates into MRI coordinates is described elsewhere (Rogers et al. 1990).
uli are shown in the top portion of Fig. 2. Each subject also had easily recognizable negative and positive peaks to the frequent stimulus both preceding and following the omitted stimulus as seen in Fig. 1. For the preceding stimulus, mean latency for the negative peak was 133 + 11 msec and for the positive peak was 222 + 13 msec. In order to determine if there were any degradation of the late positive response to the omitted stimulus across trials, a repeated measures analysis of variance was performed on the mean amplitudes and tatencies. No statistical trend was discernible.
Results
Magnetic fields MEFs to the frequent stimulation produced visible peaks with reversals corresponding to the electric negative and positive peaks in all 7 subjects. These early peaks were limited to the area of the head over the occipital lobes and no peaks above noise levels were seen over the right hemisphere. Fig. 3 shows typical isocontour maps for these distributions corresponding to the electrically positive peaks. As can be seen from the figure the fields produced during both these peaks to the frequent stimuli demonstrate a more quadrupolar pattern (2 maxima and 2 minima). As would be anticipated, single dipole modelling of these distributions did not produce acceptable fits. However, judging
Electric potent&& Electric potentials were measured at Pz (referenced to the right mastoid with the common ground on the forehead) and a grand average over all presentations (approximately 700-900 trials) was calculated for each subject. As can be seen in Fig. 1 all 7 subjects showed a distinctive positive wave following the infrequent and unpredictable omissions of visual stimulation. The mean latencies of the peaks of the late positive responses to the omitted stimuli were 445 + 28.7 msec and the mean amplitudes were 8.3 + 2.3/.LV. The grand average wave forms for the omitted and frequent stim-
GRAND AVERAGE (7 Ss) Frequent
SUBJECT #1
Omitted
Frequent
Omitted
u~
.J
_< I.Z
UJ
Pz
IO
O.
_o I..(3 LU
..I
W
"~
"
'
RIGHT
HEMISPHERE
~ Ea V ,o
I ~
"
,v
n
,,,=E I--
.E_
,~
g ,,
t',
OCCIPITAL
o
soo
o
MILLISECONDS
3o0
6o0
o
3o0
o
300
eoo
MILLISECONDS
Fig. 2. The left-hand side shows the average electrical potential at Pz, the m a x i m u m and m i n i m u m magnetic evoked responses over both the right hemisphere and occipital lobes. The right side of the figure shows the same wave forms for a single subject.
RELATIONSHIPS BETWEEN SOURCES F R O M MEFs AND EPs
by the location (centered over the midline above the occipital lobes), and the fact that these extrema were all close together, it appears as though the sources would be in the general area of the primary visual cortex. MEFs to the omitted stimulus produced very consistent pattens in all 7 subjects. That is, two distinctive sets of peaks with reversals were found on the isocontour maps in the range of 300-500 msec: one pair over the occipital area and another over the right hemisphere. Fig. 4 illustrates these two patterns. Although more than one combination of maxima and minima could be possible, the two extrema over the right hemisphere and the two over the occipital lobe appeared to be the most likely pairing based on the following considerations: (1) Morphology of the MEFs - - the waves over the right hemisphere had deflections that lasted for 250-375 msec, while those over the occipital lobes resembled slow waves which started at the point in time at which the stimulus would have occurred and lasted the entire 600 msec epoch (see Fig. 2). (2) The matching of the extrema - - pairing the extrema according to amplitude (e.g., highest maximum with highest minimum) paired the right hemisphere extrema and the occipital extrema with each other in 6 of the 7 subjects. (3) The two distributions did not overlap. Fig. 4 compares the isocontour magnetic field maps for both the actual and predicted data. Tan et al. (1990) used a spatial filtering model to estimate the ability of M E G to resolve two separate dipoles and concluded that the dipoles could be resolved as long as the distance between the two dipoles exceeded the distance between the source and the magnetic sensor coils. In the present study, the mean depth of the subcortical dipole was 4.9 + 0.5 cm and for the occipital source was 3.5 + 0.6 cm. Adding approximately 2 cm for the distance from the surface of the head to the neuromagnetometer coils would mean
FREQUENT STIMULATION NEGATIVEWAVE (N100)
POSITIVEWAVE (P200)
Fig. 3. Isofield maps of the magnetic field distributions for the negative and positive peaks following frequent stimulation. Note the multiple minima and maxima present.
149
3O ~j
--I
9o - 120
~/
MIN
I-
OCCIPITAL
-+
uJ I0
~
60 90
12o
(MIN
m
14J n" a.
Fig. 4. Isofield maps for the two sets of extrema for the late positive peak following infrequent omission of the stimulus. Both show a typical dipolar pattern and the two distributions are spatially distinct. The top and bottom of the figure contrast the actual and predicted fields. Note the similarity between the two.
that there was about 7 and 5.5 cm between these sources and the sensors. The distance between the mean locations of the two sources was about 8.4 cm, thus allowing good resolution of these two concurrent but separate sources. In order to further determine the possible influence that one dipole might have on localizing the other dipole, a single probe position was chosen where the two fields might overlap, and the predicted values for this location were calculated for each source based on the calculated dipole parameters. The mean values for these predicted values were 8.2 fT for the right hemisphere and - 1 3 . 7 f r for the contribution from the occipital source. These values are within the background noise levels of the average wave forms. Although multiple dipole algorithms for magnetic fields are currently under development they are yet to be experimentally verified. Thus, in order to fit single dipolar sources to the late positive MEFs, the MEF distribution over the right hemisphere and those over the occipital area were analyzed separately. This is a
150
R.L. R O G E R S ET AL.
valid procedure if the dipoles are spatially separated enough to have little effect on each other. This was considered valid since on inspection of the isocontour maps the two set of fields did not overlap, that is, the strength of the MEFs had diminished to within noise levels before the two M E F distributions came together (see Fig. 4). The results of the dipole fits are illustrated in Fig. 5. Instead of projecting individual sources on each subject's M R I a single representative section was chosen at the appropriate level as dictated by the mean value in the transverse, coronal or sagittal plane. The mean and 95% confidence intervals were then projected onto M R I sections, thus producing a general area in which the dipoles were located based on the distribution of coordinate values across subjects. As can be seen from Fig. 5, the dipole sources calculated from the right hemisphere and the occipital probes appear to result from two spatially distinct sources. The calculated dipole over the occipital area is a more superficial source located in the vicinity of the primary and secondary visual cortex. On the other hand, the other source which is suggested by wave forms distributed over the right hemisphere localizes to a deeper source
in the temporal lobe or medial thalamus. A multivariate analysis of variance on the cartesian coordinates of the two sources indicated a highly significant difference ( P < 0.001). Univariate tests (within the multivariate procedure) indicated that this difference was in the lateral-to-medial plane ( P < 0.00001) and the anteriorto-posterior plane ( P < 0.0001), but no difference was evident in the superior-to-inferior axis. There was no significant difference in the strength of the two sources. The mean strength of the occipital source was 5.2 nA-m and for the source in the medial temporal lobe was 6.7 nA-m. The orientation of the dipole vector is defined on a plane perpendicular to the line between the dipole and the origin of the coordinate system that is on a plane tangential to the surface of the head. Fig. 6 shows the mean angle (_+ S.E.) for the two calculated sources and the grand mean wave forms taken at Oz and Fz. The mean angle of the source in the occipital area was - 2 5 _+ 25.8 ° and pointed in the general direction of the Pz electrode indicating that a positive difference in potential would be anticipated at the Pz electrode referenced to the mastoid. The other dipole, located in the medial temporal area or the lateral thalamus (mean
POSITIVE WAVE TO OMITTED STIMULATION RIGHT HEMISPHERE
!
OCCIPITAL LOBE
Ji Fig. 5. M R I sections showing the m e a n and 95% confidence interval (2 S.D.s above and below the m e a n s based on between-subject values) for both the occipital and the deep generator sources of the late positive component in the transverse, coronal and sagittal planes.
RELATIONSHIPS BETWEEN SOURCES F R O M MEFs AND EPs
(A) >
Oz
/~
Fz
l 0
I 600
0
600
milliseconds
(B)
DIPOLE O R I E N T A T I O N
Fig. 6. A: grand mean electrical wave forms at Oz and Fz. B: the average orientation of the cortical (left) and deep source (right) of the late positive component.
a n g l e = - 6 6 + 10.1°), points in the direction of the frontal lobes suggesting a positive potential in the anterior areas of the head.
Discussion
Magnetic mapping of the responses to frequent stimuli corresponding to the electrical negative (N100m) and positive (P200m) peaks agrees with earlier research indicating the presence of multiple dipoles in the vicinity of the occipital lobes following visual stimulation. Previous MEG studies of visual stimulation (Aine et al. 1989; George et al. 1989) demonstrated the presence of multiple extrema even when specific areas of the visual field were stimulated. Such a pattern is not surprising given the complex geometry of the visual cortex and the close proximity of the two hemispheres. Comparing the data from the present study with the previous investigations mentioned above suggests that binocular stimulation results in two sets of rather equivalent minima and maxima, while stimulation of a single visual quadrant produces peaks which differ in both magnitude and morphology. Thus, comparing the extrema during variations in presentation of stimuli (e.g.,, monocular stimulation of specific visual
151
fields) should help determine which positive and negative peaks belong together. Although early components to the frequent stimuli were associated with a complex field the P3 component to the missing stimuli resulted in a simple single dipolar distribution over the occipital lobe and a separate one over the right hemisphere. That does not mean that a single dipole source for the P3 exists in the visual cortex. Such a pattern could result from the combined activity of many cell units in both hemispheres. The same may be true for the second deep source of the P3 centered in the medial temporal lobe or thalamus which could also reflect the summed activity of more than one source. Recent research from our laboratory (Rogers et al. 1991) on P3 magnetic fields in an auditory oddball paradigm suggested that an early subcortical source was followed by activity in the primary auditory cortex on the superior plane of the temporal lobe. However, in that case, the field distributions resulting from the two sources were superimposed given that both sources appeared to be on the same medial-temporal plane. In this case, given the relative position of the visual cortex with respect to the thalamus or medial temporal lobe, the external distributions of the two sources were clearly distinct. Consequently, it appears that the late positive waves to infrequent deviations in stimulus presentation produce responses in the lateral thalamus or medial aspects of the temporal lobe in both visual and auditory modalities and also produce cortical activity in the vicinity of the corresponding primary a n d / o r secondary sensory cortices. Recent depth electrode studies in humans have demonstrated activity coincident with the scalp recorded electrical P3 in the hippocampus, thalamus, frontal lobes, temporal-parietal junction, and in the temporal lobes during auditory stimulation (Halgren et al. 1980; Alaine et al. 1989; Knight et al. 1989; McCarthy et al. 1989; Smith et al. 1990). It is not possible to discern whether the magnetically derived generator in our results represents a single deep source or the summed activity of multiple generators. More detailed mapping of both electric potentials and magnetic fields may help answer these questions. However, it is becoming more apparent that scalp recorded P3s reflect the summed activity of multiple generators, thus, the ability to analyze separate components would be of significant value. An ongoing discussion in the P3 literature is whether or not it reflects modality specific processing or more general cognitive levels of processing across the various sensory modalities (Johnson 1989). The data demonstrating activity in the primary auditory cortex on the superior plane of the temporal lobe during auditory stimulation (Rogers et al. 1991) and activity in the occipital lobes during visual stimulation would strongly
152
suggest that these components of the P3 are indeed modality specific. The specificity of deep sources remains less clear. In summary, there appear to be two distinct generator sites that produce external magnetic fields coincident with the late positive electrical potential which are spatially separate and nearly orthogonal to each other, thus allowing them to be investigated as separate entities. Further research into the effects of cognitive manipulations on neocortical versus deeper structures should significantly increase our understanding of the functioning of the human brain. This research was supported in part by grants from the Moody Foundation of Galveston, TX, No. 87-45 and the National Institute of Health, Washington, DC.
References Aine, C., George, J., Meduick, P., Supek, S., Flynn, E. and BodisWollner, I. Identification of multiple sources in transient visual evoked neuromagnetic responses. In: S.J. Williamson, M. Hoke, G. Stroink and M. Kotani (Eds.), Advances in Neuromagnetism. Plenum Press, New York, 1989: 193-196. Alaine, C., Richer, F., Achim, A. and Saint-Hilaire, J. Human intracerebral potentials associated with target, novel, and omitted auditory stimuli, Brain Topogr., 1989, 1: 237-245. Donchin, E., Ritter, W. and McCallum, W.C. Cognitive psychophysiology: the endogenous component of the ERP. In: E. Callaway, P. Tueting and S. Koslow (Eds.), Event-Related Brain Potentials in Man. Academic Press, New York, 1978: 349-432. George, J.S., Aine, C.J., Meduick, P.A. and Flynn, E.R. Spatial/temporal resolution of multiple sources: paths of activation in human visual cortex. In: S.J. Williamson, M. Hoke, G. Stroink and M. Kotani (Eds.), Advances in Biomagnetism. Plenum Press, New York, 1989: 197-200. Gordon, E., Sloggett, G., Harvey, I., Kraiuhin, C., Yiannikas, C. and Mears, R. Magnetoencephalography: locating the source of P300 via magnetic field recording. Clin. Exp. Neurol., 1987, 23: 101110. Halgren, E., Squires, N.K., Wilson, C.L., Rohrbaugh, J.W., Babb,
R.L. ROGERS ET AL. T.L. and Crandall, P.H. Endogenous potentials generated in the human hippocampal formation and amygdala by infrequent events. Science, 1980, 210: 803-805. Hillyard, S.A. and Picton, T. Electrophysiology of cognition. In: F. Plum (Ed.), Handbook of Psychophysiology. American Psychological Society, Baltimore, MD, 1987: 519-584. Johnson, R. Auditory and visual P300 in temporal lobectomy patients. Evidence for modality-dependent generators. Psychophysiology, 1989, 26: 633-650. Joutsiniemi, S.L. and Hari, R. Omissions of auditory stimuli may activate frontal cortex. Eur. J. Neurosci., 1989, 1: 524-528. Knight, R.T., Seabini, D., Woods, D.L. and Clayworth, C.C. Contributions of temporal parietal junction to the human auditory P3. Brain Res., 1989, 502: 109-116. Lewine, J.D., Roeder, M.T., Oakley, M.T., Arthur, D.L., Aine, C.J., George, J.S. and Flynn, E.R. A modality-specific neuromagnetic P3. In: S.J. Williamson, M. Hoke, G. Stroink and M. Kotani (Eds.), Advances in Biomagnetism. Plenum Press, New York, 1989: 197-200. McCarthy, G., Wood, G.C., Williamson, P.D. and Spencer, D.D. Task dependent field potentials in human hippocampal formation. J. Neurosci., 1989, 9: 4253-4268. Okada, Y.C., Kaufmann, L. and Williamson, S.J. The hippocampal formation as a source of the slow endogenous potentials. Electroenceph, clin. Neurophysiol., 1983, 55: 417-426. Rogers, R.L., Papanicolaou, A., Baumann, S.B., Saydjari, C. and Eisenberg, H.M. Neuromagnetic evidence of a dynamic excitation pattern generating the N100 auditory response. Electroenceph. clin. Neurophysiol., 1990, 77: 237-240. Rogers, R.L., Papanicolaou, A.C., Baumann, S.B., Bourbon, T.W., Alagarsamy, S. and Eisenberg, H.M. Localization of P3 sources using magnetoencephalopgraphy and magnetic resonance imaging. Electroenceph. clin. Neurophysiol., 1991, 79: 308-321. Smith, M.E., Halgren, E., Sokolik, M., Baudena, P., Musolino, A., Leigeois-Chauvel, C. and Chauvel, P. The intracranial topography of the P3 event-related potential elicited during auditory oddball. Electroenceph. clin. Neurophysiol., 1990, 76:235-248 Squires, N., Squires, K. and Hillyard, S. Two varieties of long-latency positive waves evoked by unpredictable auditory stimuli in man. Electroenceph. clin. Neurophysiol., 1975, 38: 387-401. Stapleton, J.M. and Halgren, E. Endogenous potentials evoked in simple cognitive tasks: depth components and tasks correlates. Electroenceph. clin. Neurophysiol., 1987, 67: 44-52. Tan, S., Roth, B.J. and Wikswo, J.P. The magnetic field of cortical surrent sources: the application of a spatial filtering model to the forward and inverse problems. Electroenceph. clin. Neurophysiol., 1990, 76: 73-85.
146
E E G 91094
Late magnetic fields and positive evoked potentials following infrequent and unpredictable omissions of visual stimuli Robert L. Rogers, Andrew C. Papanicolaou, Stephen B. Baumann and Howard M. Eisenberg Magnetoencephalography Laboratory, Division of Neurosurgery, Unicersity of Texas Medical Branch, and Transitional Learning Community, Galceston, TX (U.S.A.) (Accepted for publication: 30 March 1992)
Summary Randomized and infrequent omissions during presentation of a steady train of visual stimulation produced distinctive wave forms of both the magnetic fields and electrical potentials. Electrical potentials at Pz showed a positive peak in response to the omitted stimuli which occurred on the average 445 msec after the time when a stimulus was anticipated. Analyses of the magnetic wave forms indicated that at least two separate sources appear to be active coincident with the electrical positive peak. O n e source localized in the occipital lobes in the vicinity of the visual cortex while the other source was located in the medial aspects of the temporal lobe or even deeper in the lateral thalamus. Judging from the calculated direction of current flow it appeared that the deep source would contribute greater potentials in the frontal areas of the scalp while the source in the occipital area would contribute to more posterior placement of electrodes, especially at Pz. Key words: Visual stimulation; Magnetic field; Evoked potential; Electrical potential; Pz
Infrequent deviations or omissions of sensory stimuli in the visual, auditory or somatosensory modalities result in an average wave form which differs from evoked responses obtained from a steady train of stimulus presentations. The most prominent difference is the large positive peak which occurs after 300 msec (Squires et al. 1975). These late positive evoked responses have received much attention since their discovery because of their relative sensitivity to cognitive factors (e.g., attention, task relevance, etc.; see Donchin et al. 1978; Hillyard and Picton 1987) as opposed to stimulus parameters (e.g., intensity, duration, etc.). However, only recently have the underlying sources, responsible for generation of these late positive potentials, been systematically investigated. In order to help delineate the character of the source or sources responsible for the late positive responses, several divergent methods are being employed including: (1) depth electrode studies in humans (Stapleton and Halgren 1987; McCarthy et al. 1989) and animals (Halgren et al. 1980); (2) mapping of surface electrical potentials (Johnson 1989); (3) mapping of external magnetic fields (Okada et al. 1983; Gordon et
Correspondence to: Robert L. Rogers, Ph.D., Division of Neurosurgery E-17, University of Texas Medical Branch, Galveston, T X 77550 (U.S.A.). Tel.: (409) 772-3965.
al. 1987; Joutsiniemi and Hari 1989; Lewine et al. 1989; Rogers et ai. 1991); and (4) studying the effects of lesions on these evoked responses (Johnson 1989; Knight et al. 1989). The present study was undertaken in order to investigate the relationships between sources extracted from external magnetic fields and the electrical potential at Pz which is typically the area where the maximal late positive response is recorded.
Methods
Seven normal right-handed subjects were tested using an oddball paradigm in which visual stimulation was delivered every 600 msec and lasted for 50 msec. Visual stimulation consisted of 4 light emitting diodes forming a square arrangement of 6 cm on each side. Subjects were instructed to fixate on the center of the square to insure binocular stimulation. On 10% of the trials the stimulation was omitted on a randomized basis in which no fewer than 3 and no greater than 15 consecutive visual stimuli were presented between the omissions. Subjects were instructed to count the number of missing flashes. The on-line computer collected evoked potential and magnetic field (MEF) epochs 1800 msec long extending 600 msec prior and 1200 after the omitted stimuli. One hundred such epochs were averaged. Subjects were actually given 100-110
RELATIONSHIPS BETWEEN SOURCES FROM MEFs AND EPs
omitted stimulus trials so that they could not guess the number of stimulus omissions A 7-channel N e u r o m a g n e t o m e t e r (Biomagnetic Technologies, Inc., Model 607), using second-order gradiometers with a 4 cm baseline and 1.8 cm coil diameters, was used to measure MEFs. Seven to 9 placements (49-63 sensor recordings) were taken on each of the subjects (4 over the occipital area and 3 - 5 over the right hemisphere) each constituting a run. The order of magnetic probe placements was randomized across subjects. Additionally, electrical potentials were obtained at Pz, Oz and Fz and both horizontal and vertical eye movements were monitored. Both electric and magnetic signals were analog bandpassed between 0.3 and 50 Hz at a digitization rate of 200 Hz. Further off-line digital filtering bandpassed the averaged wave forms between 0.3 and 20 Hz. In order to reduce fatigue effects measurements were taken over 4 sessions of approximately 1 h each. Subjects were never given more than 3 runs in a single session and they were given at least 15 min between each run. In order to insure the integrity of the average magnetic wave forms, individual wave forms were inspected for: (1) eye movement artifacts, (2) variability about the peaks, and (3) distributions of variance. Individual runs which had noticeable eye movement intrusions into the magnetic wave forms were removed before averaging. This never exceeded 5% of the total number of wave forms. Variances of the magnetic wave forms were comparable to their electrical counterparts. Frequency histograms of the individual scores were normally distributed about the means. The location of the magnetic sensors in relation to the subject's head was determined using 3 lowfrequency magnetic field sensors placed on the head with a transmitter on the cryogenic container that housed the M E G sensors. Error tolerance for the location of the sensors was less than 3 mm. Source estimations were made on the basis of a model for a single dipolar source within a homogeneous sphere. Dipole p a r a m e t e r s (location in cartesian coordinates, strength, and orientation) were estimated using a finite difference version of the LevenbergMarquardt algorithm that c o m p a r e d the measured values of the M E F with an estimated M E F produced by a single dipole. The size and origin of that sphere were determined using a least-squares algorithm based on the digitized outer surface of the head of each subject, limited to the area in which probe placements were taken. Thus, for dipole fits over the occipital area only that area was digitized and fitted to a sphere and for fits over the right hemisphere that portion of the head was fitted to a sphere. Since the side of the head is much flatter than other areas, the fitted sphere would be much larger and would not necessarily be confined to the size of the subject's head. The average radii of
147
the model spheres was 8.8 cm for the occipital area and 9.4 cm for the right hemisphere. A single dipolar source produces a distinctive contour map with two extrema (maximum and minimum) around which it diminishes in strength. A measurement of goodness-of-fit of the estimated dipole parameters is obtained by correlating the field predicted by those parameters with the measured field. Acceptance of a good dipolar fit was set a priori at a level that accounted for at least 81% of the total variance (r = 0.90). The inverse solution was applied at every 5 msec interval in accordance with the maximum temporal
S #1
S #2
~., ~ / v
¥ -.~..~.
I
l
[
I
I
I
0
300
0
300
0
300
MILLISECONDS PRECEDING STIMULUS
OMITTED STIMULUS
FOLLOWING STIMULUS
Fig. 1. Electrical evoked responses recorded at Pz for all 7 subjects showing three 600 msec epochs, the epoch preceding the omitted stimulus, the omitted stimulus and the following frequent stimulus epoch. Positive potentials for electrical wave forms are represented as upward deflections.
148
R.L. R O G E R S ET AL.
resolution provided by the 200 Hz digitization rate for both the frequent and omitted stimulus intervals. This accounted for the sources of the fields throughout the duration of the recorded epoch that included the response to the omitted and the immediately preceding stimulus. Magnetic resonance images (MRIs) were obtained for each subject. The procedures and details of the MRI images and for transferring the M E G coordinates into MRI coordinates is described elsewhere (Rogers et al. 1990).
uli are shown in the top portion of Fig. 2. Each subject also had easily recognizable negative and positive peaks to the frequent stimulus both preceding and following the omitted stimulus as seen in Fig. 1. For the preceding stimulus, mean latency for the negative peak was 133 + 11 msec and for the positive peak was 222 + 13 msec. In order to determine if there were any degradation of the late positive response to the omitted stimulus across trials, a repeated measures analysis of variance was performed on the mean amplitudes and tatencies. No statistical trend was discernible.
Results
Magnetic fields MEFs to the frequent stimulation produced visible peaks with reversals corresponding to the electric negative and positive peaks in all 7 subjects. These early peaks were limited to the area of the head over the occipital lobes and no peaks above noise levels were seen over the right hemisphere. Fig. 3 shows typical isocontour maps for these distributions corresponding to the electrically positive peaks. As can be seen from the figure the fields produced during both these peaks to the frequent stimuli demonstrate a more quadrupolar pattern (2 maxima and 2 minima). As would be anticipated, single dipole modelling of these distributions did not produce acceptable fits. However, judging
Electric potent&& Electric potentials were measured at Pz (referenced to the right mastoid with the common ground on the forehead) and a grand average over all presentations (approximately 700-900 trials) was calculated for each subject. As can be seen in Fig. 1 all 7 subjects showed a distinctive positive wave following the infrequent and unpredictable omissions of visual stimulation. The mean latencies of the peaks of the late positive responses to the omitted stimuli were 445 + 28.7 msec and the mean amplitudes were 8.3 + 2.3/.LV. The grand average wave forms for the omitted and frequent stim-
GRAND AVERAGE (7 Ss) Frequent
SUBJECT #1
Omitted
Frequent
Omitted
u~
.J
_< I.Z
UJ
Pz
IO
O.
_o I..(3 LU
..I
W
"~
"
'
RIGHT
HEMISPHERE
~ Ea V ,o
I ~
"
,v
n
,,,=E I--
.E_
,~
g ,,
t',
OCCIPITAL
o
soo
o
MILLISECONDS
3o0
6o0
o
3o0
o
300
eoo
MILLISECONDS
Fig. 2. The left-hand side shows the average electrical potential at Pz, the m a x i m u m and m i n i m u m magnetic evoked responses over both the right hemisphere and occipital lobes. The right side of the figure shows the same wave forms for a single subject.
RELATIONSHIPS BETWEEN SOURCES F R O M MEFs AND EPs
by the location (centered over the midline above the occipital lobes), and the fact that these extrema were all close together, it appears as though the sources would be in the general area of the primary visual cortex. MEFs to the omitted stimulus produced very consistent pattens in all 7 subjects. That is, two distinctive sets of peaks with reversals were found on the isocontour maps in the range of 300-500 msec: one pair over the occipital area and another over the right hemisphere. Fig. 4 illustrates these two patterns. Although more than one combination of maxima and minima could be possible, the two extrema over the right hemisphere and the two over the occipital lobe appeared to be the most likely pairing based on the following considerations: (1) Morphology of the MEFs - - the waves over the right hemisphere had deflections that lasted for 250-375 msec, while those over the occipital lobes resembled slow waves which started at the point in time at which the stimulus would have occurred and lasted the entire 600 msec epoch (see Fig. 2). (2) The matching of the extrema - - pairing the extrema according to amplitude (e.g., highest maximum with highest minimum) paired the right hemisphere extrema and the occipital extrema with each other in 6 of the 7 subjects. (3) The two distributions did not overlap. Fig. 4 compares the isocontour magnetic field maps for both the actual and predicted data. Tan et al. (1990) used a spatial filtering model to estimate the ability of M E G to resolve two separate dipoles and concluded that the dipoles could be resolved as long as the distance between the two dipoles exceeded the distance between the source and the magnetic sensor coils. In the present study, the mean depth of the subcortical dipole was 4.9 + 0.5 cm and for the occipital source was 3.5 + 0.6 cm. Adding approximately 2 cm for the distance from the surface of the head to the neuromagnetometer coils would mean
FREQUENT STIMULATION NEGATIVEWAVE (N100)
POSITIVEWAVE (P200)
Fig. 3. Isofield maps of the magnetic field distributions for the negative and positive peaks following frequent stimulation. Note the multiple minima and maxima present.
149
3O ~j
--I
9o - 120
~/
MIN
I-
OCCIPITAL
-+
uJ I0
~
60 90
12o
(MIN
m
14J n" a.
Fig. 4. Isofield maps for the two sets of extrema for the late positive peak following infrequent omission of the stimulus. Both show a typical dipolar pattern and the two distributions are spatially distinct. The top and bottom of the figure contrast the actual and predicted fields. Note the similarity between the two.
that there was about 7 and 5.5 cm between these sources and the sensors. The distance between the mean locations of the two sources was about 8.4 cm, thus allowing good resolution of these two concurrent but separate sources. In order to further determine the possible influence that one dipole might have on localizing the other dipole, a single probe position was chosen where the two fields might overlap, and the predicted values for this location were calculated for each source based on the calculated dipole parameters. The mean values for these predicted values were 8.2 fT for the right hemisphere and - 1 3 . 7 f r for the contribution from the occipital source. These values are within the background noise levels of the average wave forms. Although multiple dipole algorithms for magnetic fields are currently under development they are yet to be experimentally verified. Thus, in order to fit single dipolar sources to the late positive MEFs, the MEF distribution over the right hemisphere and those over the occipital area were analyzed separately. This is a
150
R.L. R O G E R S ET AL.
valid procedure if the dipoles are spatially separated enough to have little effect on each other. This was considered valid since on inspection of the isocontour maps the two set of fields did not overlap, that is, the strength of the MEFs had diminished to within noise levels before the two M E F distributions came together (see Fig. 4). The results of the dipole fits are illustrated in Fig. 5. Instead of projecting individual sources on each subject's M R I a single representative section was chosen at the appropriate level as dictated by the mean value in the transverse, coronal or sagittal plane. The mean and 95% confidence intervals were then projected onto M R I sections, thus producing a general area in which the dipoles were located based on the distribution of coordinate values across subjects. As can be seen from Fig. 5, the dipole sources calculated from the right hemisphere and the occipital probes appear to result from two spatially distinct sources. The calculated dipole over the occipital area is a more superficial source located in the vicinity of the primary and secondary visual cortex. On the other hand, the other source which is suggested by wave forms distributed over the right hemisphere localizes to a deeper source
in the temporal lobe or medial thalamus. A multivariate analysis of variance on the cartesian coordinates of the two sources indicated a highly significant difference ( P < 0.001). Univariate tests (within the multivariate procedure) indicated that this difference was in the lateral-to-medial plane ( P < 0.00001) and the anteriorto-posterior plane ( P < 0.0001), but no difference was evident in the superior-to-inferior axis. There was no significant difference in the strength of the two sources. The mean strength of the occipital source was 5.2 nA-m and for the source in the medial temporal lobe was 6.7 nA-m. The orientation of the dipole vector is defined on a plane perpendicular to the line between the dipole and the origin of the coordinate system that is on a plane tangential to the surface of the head. Fig. 6 shows the mean angle (_+ S.E.) for the two calculated sources and the grand mean wave forms taken at Oz and Fz. The mean angle of the source in the occipital area was - 2 5 _+ 25.8 ° and pointed in the general direction of the Pz electrode indicating that a positive difference in potential would be anticipated at the Pz electrode referenced to the mastoid. The other dipole, located in the medial temporal area or the lateral thalamus (mean
POSITIVE WAVE TO OMITTED STIMULATION RIGHT HEMISPHERE
!
OCCIPITAL LOBE
Ji Fig. 5. M R I sections showing the m e a n and 95% confidence interval (2 S.D.s above and below the m e a n s based on between-subject values) for both the occipital and the deep generator sources of the late positive component in the transverse, coronal and sagittal planes.
RELATIONSHIPS BETWEEN SOURCES F R O M MEFs AND EPs
(A) >
Oz
/~
Fz
l 0
I 600
0
600
milliseconds
(B)
DIPOLE O R I E N T A T I O N
Fig. 6. A: grand mean electrical wave forms at Oz and Fz. B: the average orientation of the cortical (left) and deep source (right) of the late positive component.
a n g l e = - 6 6 + 10.1°), points in the direction of the frontal lobes suggesting a positive potential in the anterior areas of the head.
Discussion
Magnetic mapping of the responses to frequent stimuli corresponding to the electrical negative (N100m) and positive (P200m) peaks agrees with earlier research indicating the presence of multiple dipoles in the vicinity of the occipital lobes following visual stimulation. Previous MEG studies of visual stimulation (Aine et al. 1989; George et al. 1989) demonstrated the presence of multiple extrema even when specific areas of the visual field were stimulated. Such a pattern is not surprising given the complex geometry of the visual cortex and the close proximity of the two hemispheres. Comparing the data from the present study with the previous investigations mentioned above suggests that binocular stimulation results in two sets of rather equivalent minima and maxima, while stimulation of a single visual quadrant produces peaks which differ in both magnitude and morphology. Thus, comparing the extrema during variations in presentation of stimuli (e.g.,, monocular stimulation of specific visual
151
fields) should help determine which positive and negative peaks belong together. Although early components to the frequent stimuli were associated with a complex field the P3 component to the missing stimuli resulted in a simple single dipolar distribution over the occipital lobe and a separate one over the right hemisphere. That does not mean that a single dipole source for the P3 exists in the visual cortex. Such a pattern could result from the combined activity of many cell units in both hemispheres. The same may be true for the second deep source of the P3 centered in the medial temporal lobe or thalamus which could also reflect the summed activity of more than one source. Recent research from our laboratory (Rogers et al. 1991) on P3 magnetic fields in an auditory oddball paradigm suggested that an early subcortical source was followed by activity in the primary auditory cortex on the superior plane of the temporal lobe. However, in that case, the field distributions resulting from the two sources were superimposed given that both sources appeared to be on the same medial-temporal plane. In this case, given the relative position of the visual cortex with respect to the thalamus or medial temporal lobe, the external distributions of the two sources were clearly distinct. Consequently, it appears that the late positive waves to infrequent deviations in stimulus presentation produce responses in the lateral thalamus or medial aspects of the temporal lobe in both visual and auditory modalities and also produce cortical activity in the vicinity of the corresponding primary a n d / o r secondary sensory cortices. Recent depth electrode studies in humans have demonstrated activity coincident with the scalp recorded electrical P3 in the hippocampus, thalamus, frontal lobes, temporal-parietal junction, and in the temporal lobes during auditory stimulation (Halgren et al. 1980; Alaine et al. 1989; Knight et al. 1989; McCarthy et al. 1989; Smith et al. 1990). It is not possible to discern whether the magnetically derived generator in our results represents a single deep source or the summed activity of multiple generators. More detailed mapping of both electric potentials and magnetic fields may help answer these questions. However, it is becoming more apparent that scalp recorded P3s reflect the summed activity of multiple generators, thus, the ability to analyze separate components would be of significant value. An ongoing discussion in the P3 literature is whether or not it reflects modality specific processing or more general cognitive levels of processing across the various sensory modalities (Johnson 1989). The data demonstrating activity in the primary auditory cortex on the superior plane of the temporal lobe during auditory stimulation (Rogers et al. 1991) and activity in the occipital lobes during visual stimulation would strongly
152
suggest that these components of the P3 are indeed modality specific. The specificity of deep sources remains less clear. In summary, there appear to be two distinct generator sites that produce external magnetic fields coincident with the late positive electrical potential which are spatially separate and nearly orthogonal to each other, thus allowing them to be investigated as separate entities. Further research into the effects of cognitive manipulations on neocortical versus deeper structures should significantly increase our understanding of the functioning of the human brain. This research was supported in part by grants from the Moody Foundation of Galveston, TX, No. 87-45 and the National Institute of Health, Washington, DC.
References Aine, C., George, J., Meduick, P., Supek, S., Flynn, E. and BodisWollner, I. Identification of multiple sources in transient visual evoked neuromagnetic responses. In: S.J. Williamson, M. Hoke, G. Stroink and M. Kotani (Eds.), Advances in Neuromagnetism. Plenum Press, New York, 1989: 193-196. Alaine, C., Richer, F., Achim, A. and Saint-Hilaire, J. Human intracerebral potentials associated with target, novel, and omitted auditory stimuli, Brain Topogr., 1989, 1: 237-245. Donchin, E., Ritter, W. and McCallum, W.C. Cognitive psychophysiology: the endogenous component of the ERP. In: E. Callaway, P. Tueting and S. Koslow (Eds.), Event-Related Brain Potentials in Man. Academic Press, New York, 1978: 349-432. George, J.S., Aine, C.J., Meduick, P.A. and Flynn, E.R. Spatial/temporal resolution of multiple sources: paths of activation in human visual cortex. In: S.J. Williamson, M. Hoke, G. Stroink and M. Kotani (Eds.), Advances in Biomagnetism. Plenum Press, New York, 1989: 197-200. Gordon, E., Sloggett, G., Harvey, I., Kraiuhin, C., Yiannikas, C. and Mears, R. Magnetoencephalography: locating the source of P300 via magnetic field recording. Clin. Exp. Neurol., 1987, 23: 101110. Halgren, E., Squires, N.K., Wilson, C.L., Rohrbaugh, J.W., Babb,
R.L. ROGERS ET AL. T.L. and Crandall, P.H. Endogenous potentials generated in the human hippocampal formation and amygdala by infrequent events. Science, 1980, 210: 803-805. Hillyard, S.A. and Picton, T. Electrophysiology of cognition. In: F. Plum (Ed.), Handbook of Psychophysiology. American Psychological Society, Baltimore, MD, 1987: 519-584. Johnson, R. Auditory and visual P300 in temporal lobectomy patients. Evidence for modality-dependent generators. Psychophysiology, 1989, 26: 633-650. Joutsiniemi, S.L. and Hari, R. Omissions of auditory stimuli may activate frontal cortex. Eur. J. Neurosci., 1989, 1: 524-528. Knight, R.T., Seabini, D., Woods, D.L. and Clayworth, C.C. Contributions of temporal parietal junction to the human auditory P3. Brain Res., 1989, 502: 109-116. Lewine, J.D., Roeder, M.T., Oakley, M.T., Arthur, D.L., Aine, C.J., George, J.S. and Flynn, E.R. A modality-specific neuromagnetic P3. In: S.J. Williamson, M. Hoke, G. Stroink and M. Kotani (Eds.), Advances in Biomagnetism. Plenum Press, New York, 1989: 197-200. McCarthy, G., Wood, G.C., Williamson, P.D. and Spencer, D.D. Task dependent field potentials in human hippocampal formation. J. Neurosci., 1989, 9: 4253-4268. Okada, Y.C., Kaufmann, L. and Williamson, S.J. The hippocampal formation as a source of the slow endogenous potentials. Electroenceph, clin. Neurophysiol., 1983, 55: 417-426. Rogers, R.L., Papanicolaou, A., Baumann, S.B., Saydjari, C. and Eisenberg, H.M. Neuromagnetic evidence of a dynamic excitation pattern generating the N100 auditory response. Electroenceph. clin. Neurophysiol., 1990, 77: 237-240. Rogers, R.L., Papanicolaou, A.C., Baumann, S.B., Bourbon, T.W., Alagarsamy, S. and Eisenberg, H.M. Localization of P3 sources using magnetoencephalopgraphy and magnetic resonance imaging. Electroenceph. clin. Neurophysiol., 1991, 79: 308-321. Smith, M.E., Halgren, E., Sokolik, M., Baudena, P., Musolino, A., Leigeois-Chauvel, C. and Chauvel, P. The intracranial topography of the P3 event-related potential elicited during auditory oddball. Electroenceph. clin. Neurophysiol., 1990, 76:235-248 Squires, N., Squires, K. and Hillyard, S. Two varieties of long-latency positive waves evoked by unpredictable auditory stimuli in man. Electroenceph. clin. Neurophysiol., 1975, 38: 387-401. Stapleton, J.M. and Halgren, E. Endogenous potentials evoked in simple cognitive tasks: depth components and tasks correlates. Electroenceph. clin. Neurophysiol., 1987, 67: 44-52. Tan, S., Roth, B.J. and Wikswo, J.P. The magnetic field of cortical surrent sources: the application of a spatial filtering model to the forward and inverse problems. Electroenceph. clin. Neurophysiol., 1990, 76: 73-85.
