169
Neurophysiol Clin (1990) 20, 169-187 © Elsevier, Paris
Original article
Magnetoencephalography in the study of epilepsy* R Paetau, M Kajola, R Hari L o w Temperature Laboratory, Helsinki University o f Technology, 02150 Espoo, Finland (Received 15 March 1990; accepted 20 May 1990)
Summary - A brief review is given about the basic principles of magnetoeucephalography (MEG), a noninvasive brain research method in which weak magnetic fields are detected outside the h u m a n head with SQUID (Superconducting Q u a n t u m Interference Device) magnetometers. The active brain areas, producing the signal, are modelled by current dipoles, which are assumed to be situated in a spherically symmetric volume conductor. The locations of these "equivalent dipoles" can be found, in the optimal case, with a precision of a few millimeters. The new multichannel magnetometers allow measurements of spontaneous brain activity without EEG-triggered averaging. The 3-dimensional locations of superficial epileptogenic foci can be determined with respect to external landmarks on the skull and to known generator areas of evoked responses in the brain. Examples are given about M E G recordings of epileptic patients. magnetoencephalography / epilepsy / source models / man
R~sum~ - La magn~toenc~phalographie dans I'~tude de l'~pilepsie. Les auteurspr~sentent une brbve revue sur les principes de la magn~todlectroenc6phalographie qui d~tecte de faibles champs magndtiques ~ l'extdrieur du cr~ne au moyen de magndtombtre S Q U I D (superconducting quatum interference device). Les zones cdr~brales aetives, g~n6ratrices du signal sont mod~lis~es sous la f o r m e de courants induits par des dip6les situ~s clans un volumes de conduction sph~rique el symdtrique. La localisation de ces dquivalents dip6les peut ~tre d~finie avec une prdcision de l'ordre de quelques millimdtres. Les nouveaux magndtombtres multicanaux permettent des mesures de l' activitd E E G spontande sans moyennage pr6alable. La localisation en 3 dimensions des f o y e r s ~pileptogbnes peat ~tre prdcisde p a r rapport d des repbres osseux el aux gdndraleurs connus de rdponses ~voqudes : quelques exemples de magndtoencdphalographie de patients dpileptiques sont prdsentds. magn~toenc~phalographie / dpilepsie / moddisation 1 homme
* Based on a talk given by R Hari at the R6union de la Soci6t6 de Neurophysiologie, Clinique, Paris, December 12-13, 1989.
170
R P a e t a u et al
Introduction In western countries 0 . 5 - 1 % of the population suffers from epilepsy. In spite advances in antiepJleptic medication, the seizures o f 20=30% of these patients c~ not be controlled adequately, and surgical treatment is considered. In skilled han~ epilepsy surgery gives good results in 6 0 - 8 0 % of carefully assessed patients (Rz mussen, 1983; Spencer, 1988 ; Silfvenius et al, 1989). The necessary conditions 1 successful outcome are (i) the existence of only 1 trigger area, (ii) accurate locali: tion of it ; and (iii) its situation in a brain region which can be removed without t much harm to the subject. The pre-operative evaluation of epileptic patients is based on a combination neuroimaging techniques, intensive long-term electrophysiological monitoril detailed seizure symptomatology and neuropsychological methods. Although a str tural lesion usually underlies focal-onset epilepsy, its site may remain undetected means of neuroradiological imaging, even in cases of frequent seizures. Further, site of an epileptogenic focus may not always be identical to that of a visible str tural lesion. For accurate localization of the focus, invasive techniques, such as sl dural or depth electrode recordings, are often needed. The non-invas magnetoencephalographic (MEG) recordings, which will be discussed in this bl review, alsoprovide additional information in the selection and pre-operative ev~ ation of candidates for epilepsy surgery. We will first describe the basic principles of the MEG technique and the phy~, logical basis of the observed activity. Then the data analysis methods of epilet MEG phenomena will be discussed. Some results of MEG studies on epileptic tients obtained so far are then described. Finally, some examples of MEG recc ings carried out in our own laboratory are presented. The interested reader is refer to review articles and conference proceedings for more detailed discussions of technical and theoretical basis of neuromagnetism (Williamson and Kaufman, 19 Weinberg et al, 1985" Hari and Ilmoniemi, t,986" Atsumi et al, 1988; Hari Lounasmaa, 1989).
The neuromagnetic technique History The first experimental data about cerebral magnetic fields were obtained in 1968 w D Cohen detected, with a two-million turn induction-coil magnetometer, the Sl: taneous magnetic alpha activity of the human brain (Cohen, 1968). The simult~ ously recorded electric alpha-rhythm signal was used as a trigger for signal averag Introduction of SQUID (Superconducting QUantum Interference Device) magnet, eters significantly improved the sensitivity of the measurements and allowed sg taneous activity to be monitored in real time. The first magnetic evoked resp~
Magnetoencephalography and epilepsy
171
recordings were carried out a few years later (Brenner et al, 1975 ; Teyler et al, 1975) and the field of neuromagnetism started to emerge. Since then S Q U I D magnetometers have been used to detect evoked activity of sensory projection cortices and, more recently, of association areas. A few laboratories have been intensively studying spontaneous brain activity in epileptic patients.
General features o f the M E G method In magnetoencephalography weak cerebral magnetic fields are detected outside the head with sensitive S Q U I D magnetometers (for details of the instrumentation, see Ryh/inen et al, 1989). Evoked cerebral magnetic fields are typically 100-1 000 fT (femtotesla = 10-15 Tesla) in amplitude, only one part in 109 or 10 s of the earth's geomagnetic field. Signals associated with spontaneous activity and epileptic discharges can be considerably stronger. To avoid external magnetic artefacts, the recordings are preferably carried out inside a magnetically shielded room, usually made of mu-metal and aluminium (Kelh~i et al, 1982). The main purpose o f M E G recordings is (i) to detect temporospatial changes in cerebral activity and (ii) to locate source areas with respect to known landmarks on the skull, or preferably in the brain. It is important to know the exact sites and orientations of the magnetometer channels with respect to the head. We obtain this information by placing a small plate containing 3 wire loops to a known site on the skull and by measuring the field pattern produced by currents led through the loops (Knuutila et al, 1987). In determining the source locations one is faced with the "inverse p r o b l e m " , ie, how to deduce f r o m the magnetic field measured outside the head the underlying electric activity within the brain. Unfortunately this problem does not have a unique solution since an infinite number of different current distributions in the brain could produce exactly the same field pattern. Consequently, one has to use source models to describe active brain areas and a volume conductor model to represent the head in which the currents are flowing. A current dipole model is c o m m o n l y used to describe the movement of electric charges in the brain over short distances. An equivalent current dipole represents the net current in an active brain area. The electrical circuit is completed by volume currents from the head of the dipole to its tail in the surrounding medium, so that there is no build-up of charge. A typical strength of the equivalent dipole encountered in evoked response measurements is of the order 10 n A . m (nanoAmpere.meter). For a current dipole within a sphere the distribution of the magnetic field normal to the surface of the sphere has 2 extrema, as illustrated in figure 1. I f a similar field pattern is measured outside the head, the most probable active brain area is halfway between the extrema, with the depth determined by the distance between the field m a x i m u m and the minimum. Usually the source is found by applying a least-squares fit to the measured data. The accuracy o f this solution strongly depends on the signalto-noise ratio, and can be a few millimeters for superficial sources (Hari et al, 1988).
172
R Paetau et al
The head as a volume conductor can be successfully modeled by a sphere. In t/c approximation, only currents which are tangential to the surface (or tangential co~ ponents of tilted dipoles) produce a magnetic field outside, while radial dipoles a externally silent. Because of the orientation of the pyramidal cells and their apic dendrites, the main direction of current flow in the cortex is perpendicular to its st face. Therefore, M E G measures activity mainly from the sulci and fissures of t cortex, which often simplifies the interpretation of the data. Fortunately, the prima auditory, somatosensory, and visual projection areas are situated in the fissural cc tex. When a large brain area of any orientation is active during an epileptic dischar~ the source current usually has a tangential component which can be detected wi MEG. This fact is further emphasized by the typically 10 to 100-fold larger curre strengths associated with epileptic spikes than evoked responses. Physiologically it is reasonable to divide the neural currents associated with po synaptic potentials to (i) transmembrane currents at the active synaptic area, (ii) i tracellular currents within the neuron, and (iii) extracellular volume currents. Ir sphere, the external magnetic field is produced by the intracellular currents aloi as has been shown both theoretically (Hari and Ilmoniemi, 1986) and by compari the polarities of electric and magnetic evoked responses (Hari et al, 1980). In ott geometries, volume currents also become important. A dipole within a sphere c be characterized by means of 5 parameters, 3 for its 3-dimensional location, 1 t its orientation (only the plane parallel to the surface of the sphere is relevant), a 1 for its strength. Instrumentation
The SQUID is a superconducting ring, interrupted by 1 or 2 weak links, called Joset son junctions. It is a parametric amplifier that is very sensitive to magnetic fl, penetrating the ring, and there suitable for detection of the weak cerebral magne fields. The flux transformer (fig 2) is an important part of the magnetometer a can be used to detect special source configurations or to cancel artefacts. A tra: former containing just 1 loop is called a magnetometer; it is very sensitive to sign but also to artefacts. Therefore, a compensation coil is usually added : the first o: er gradiometer cancels artefacts which arise from distant sources and thus prodl similar signals in both loops. With decreasing baseline of the gradiometer, ie, 1 distance between the pickup and the compensation coils, the device becomes less v nerable to distant artefacts but also less sensitive to deep cerebral sources. A plal flux-transformer, also shown in figure 2, can be used to detect field gradients Figure 3 illustrates a typical experimental arrangement for MEG recordings w our 7-channel SQUID gradiometer (Knuutila et al, 1987). The flux transformer s tem consists of 7 first-order axial gradiometer coils in a hexagonal array. The ch~ nels are separated by 3.65 cm and the diameter of the active measurement are~ 9.3 cm. The sensitivity of each channel is 5 - 6 fT x/Hz. A new 24-channel instrument, show in figure 4, was recently completed in
173
Magnetoencephalography and epilepsy
laboratory (Kajola et al, 1990). This device uses a planar flux transformer configuration: with 2 orthogonal loops at each location, the 2 off-diagonal field gradients 8 B,./8 x and 8 Br/8 y are obtained at 12 sites simultaneously. While the axial magnetometer detects the field extrema on both sides of the dipole the 24-SQUID planar gradiometer records the largest signal just above the dipolar source. The sensitivity of the device is 3 - 6 fT/(cm.~H-~). The covered active measurement area, 12.5 cm in diameter, often allows the whole field pattern to be determined with a single measurement. Main differences between E E G and M E G MEG is closely related to electroencephalography, E E G ; both electric and magnetic signals are generated by neural currents. The temporal resolution of both methods is better than a millisecond. To understand the differences between E E G and MEG, let us consider a current dipole in a spherical volume conductor consisting of 4 layers of different conductivities, simulating the brain, the cerebrospinal fluid, the skull
current
dipole
h,,
,\
}.,/
\
:
•
i °
,
.
,'
t
l • s 7 f ' s" . J ." , s ° e ° °°.° f.° .~
s"
."
,r"
i"
Z! \
.-
°.° /
MAGNETIC
FIELD
ELECTRIC
POTENTIAL
Fig 1. Magnetic field pattern and electric potential distribution produced by a current dipole (arrow) in a 4-layer sphere model simulating the head. The continuous lines refer to magnetic flux oriented out of the sphere or to positive potential.
174
R Paetau et al
a n d t h e s c a l p (fig 1). T h e r e s u l t i n g electric p o t e n t i a l d i s t r i b u t i o n a n d m a g n e t i c fi, pattern are both "dipolar",
ie, d i s p l a y 2 e x t r e m a o f o p p o s i t e p o l a r i t i e s , b u t a r e
t a t e d b y 90 d e g r e e s w i t h r e s p e c t to e a c h o t h e r . T h e i s o c o n t o u r lines are r e l a t i v e l y t i g h t e r in t h e m a g n e t i c t h a n in t h e electric p t e r n . T h i s is b e c a u s e c o n c e n t r i c electric i n h o m o g e n e i t i e s s m e a r t h e e l e c t r i c p o t e n t i
Axial
Planar
Fig 2. Schematic illustration of flux transformer configurations : an axial first order gradiometer ( and a planar gradiometer (right). Magnetic flux directed along the thick arrows induces a current in coil (small arrows). With the axial gradiometer, 2 field maxima are detected, 1 on each side of the cur dipole. The figure-of-eight planar flux transformer, being sensitive to the field gradient, detects the rn mum signal when the difference between the fluxes threading the 2 loops is largest. For a current diF this occurs just above the source. (Figure kindly provided by Dr Jukka Knuutila).
Magnetoencephalography and epilepsy
175
preamplifiers
channel configuration
SQUIDs within niobium shields
liquid helium (-269°C)
gradiometer coils
(°",*J vacuum cast
Fig 3. Typical experimental arrangement for MEG recordings. The subject lies on a bed, with his head supported by a vacuum cast to prevent movements. The concave b o t t o m of the dewar is placed as close to the head as possible. Within the dewar, the SQUID detectors and flux transformers are immersed in liquid helium at T = - 2 6 9 °C. Adapted from Hari and Lounasrnaa (1989).
176
R Paetau et al
Fig 4. Photograph of the measuring head of our 24-channe! magnetometer (Kajola et al, 1990). Exam of 2 figure-of-eight loops have been drawn on the figure.
Magnetoencephalography and epilepsy
177
whereas the magnetic pattern remains intact. Further, while the measured electric potential distribution is affected by the location of the reference electrode, no corresponding problem exists in the magnetic recordings. Consequently, MEG has good spatial resolution for superficial cortical sources. On the other hand, EEG is sensitive to both radial and tangential sources and also reflects activity of the deepest parts of the brain.
Artefacts Since the magnetic brain signals are very weak, rejection of external disturbances is of extreme importance. Magnetic noise is caused, for example, by cars and elevators, by radio, and by power transmission lines. For the best results, MEG measurements should be performed in a magnetically shielded room. Biological tissues also generate fields which can disturb MEG signals. For example, the heart's magnetic field over the chestis 2 - 3 orders of magnitude larger than the brain signals outside the head. This type of disturbance can be decreased by using a gradiometric flux transformer. Eye movements and blinks also cause artefacts which are large in comparison with the cerebral signals (Antervo et al, 1985). We routinely apply electro-oculogram to reject contaminated signals. In rare cases muscular artefacts may also cause problems. The subject should not wear clothes containing magnetized metallic material since these generate artefacts in the rhythm of respiration or of cardiac activity. A modern source of artefacts, observed in our studies of epileptic children, is produced by intraoral metallic devices for orthodontics. Post-operative M E G evaluation may be impossible if magnetized material is used in clips, sutures, etc.
Methods to analyze MEG data of epileptic patients Advanced data analysis methods have been developed for interpreting neuromagnetic evoked fields. Localization of the generators of spontaneous magnetic brain activity, measured sequentially at several locations, is much more difficult since the activity does not necessarily repeat itself. Different analysis methods are available to determine the source configuration.
Signal averaging In locating the generation site of MEG discharges it is common to apply averaging : the simultaneously recorded EEG signal is used as a trigger for MEG waveforms, which are recorded sequentially from several locations (Barth et al, 1982). Unfortunately, the choice of the trigger channel largely determines the observed field pattern, and the method only works if well discernable spikes or sharp waves are present.
178
R Paetau et al
Relative covariance The Rome group (Chapman et al, 1984) introduced a method where relative covai ances are calculated between 1 electric channel and the magnetic recordings. The da are normalized by the amplitude of the electric signal since the activity can vary co~ siderably as a function of time, This metho d allows studies of irregular signals. Agai: the drawback is that only activity reflected in the electric recordings will be detect¢ and that the field patterns will be dissimilar when different EEG derivations are chose as the reference.
Spectral maps Useful information about source locations can be obtained by calculating frequen. spectra of the signals and by mapping the abundance of different frequencies at va ious locations. When multi-channel recordings are available, phase-spectra betwe~ different channels suggest which signals arise from the same source.
Instantaneous determination of the field pattern With the most advanced multi-channel M E G instruments it is already possible locate the source with a single measurement which greatly increases the accuracy the method (Stefan et al, 1990; Tiihonen et al, 1990). Using devices which cover wi areas of the head, it is possible to find the equivalent source at any instance al also to employ signal averaging, if needed, by first choosing a template with whi the activity will be compared. It will then be possible to automatically classify t abnormal waveforms into different classes and to find out their source locatio separately.
Qualitative recording Accurate source localization is not always required: one might be interested to s the signal w a v e f o r m and the temporal changes in the qualitative features of the fie pattern. These types of measurements, which resemble the classical use of EEG reco~ ings can be useful in screening candidates for surgical treatment: clear changes the field pattern from 1 spike to another discourage the assumption of a local on~ for the discharge, although they do not rule out a c o m m o n trigger.
Reference sites for source localization It is c o m m o n to determine the source locations with respect to external landmal of the skull. Localization in relation to known sources of some evoked respons like the area 3b in the posterior wall of the central sulcus or the auditory area arou the gyrus of Heschl at the superior bank of the temporal lobe, might in the futt
Magnetoencephalography and epilepsy
179
become more important since it can help the surgeon to avoid areas which are functionally irretrievable. For example, some neurosurgeons stimulate the cortex and consider the representation area of the tongue the most posterior border for temporal lobe resection. By recoding magnetic responses to short electric pulses applied to the tongue (Karhu et al, 1990) it has become possible to pre-operatively localize the sensory tongue area with respect to the epileptic focus.
MEG studies of epilepsy In the first MEG recording of an epileptic patient (Cohen, 1972), slow EEG and MEG activity increased during hyperventilation; the patient had psychomotor epilepsy. Later, more subjects were studied in several laboratories. Modena et al (1982) reported, in patients with focal epilepsy, a sharp localization of paroxysmal magnetic activity. In patients with Rolandic foci the MEG findings were positive even in the absence of EEG abnormalities. This difference was probably due to the tangential orientation of the source and also due to the fact that a more dense measurement grid was used in the MEG than the EEG recordings. The group at the University of California in Los Angeles localized both single sources and multiple epileptogenic areas (Barth et al, 1982, 1984). One interesting finding was a systematic delay of 20 ms between left and right temporal spikes with respect to an electric signal. Such time differences can be used to discriminate between the primary and secondary (mirror) foci. In another interesting study both scalp and sphenoidal electrode spikes were used as triggers for averaging the magnetic spikes (Sutherling and Barth, 1989). A consistent depth difference was observed in the equivalent source areas of these 2 groups of spikes. The finding was suggested to indicate propagation of the discharge from deeper to more superficial brain areas. Recently, magnetic slow shifts occurring during epileptic seizures have also been recorded (Vieth et al, 1990).
Own examples Temporal epilepsy An adult female patient with medically intractable epilepsy was studied (for detailed data, see Tiihonen et al, 1990). She had suffered from complex partial seizures for over 2 decades. In spite of frequent seizures, the MRI and CT scans were negative. The magnetic field was measured at several locations over the right hemisphere with our 7-channel large-area magnetometer. The spontaneous magnetic activity showed abundant spikes which reversed their polarity between the anterior and posterior measurement locations over the right temporal area. Sometimes the magnetic spikes reversed polarity even in one 7-channel measurement, as shown in figure 5.
180
R P a e t a u et al
By constructing topographic field maps it was possible to follow the temporal e~ lution of the activity during the spike : the location of the equivalent source travel] about 20 m m during 23 ms. Activation of the auditory cortex, monitored by m e uring evoked magnetic fields, was used as a landmark for the source location: ( source of the epileptogenic activity was about 2 cm anterior to the auditory cort, It was encouraging to be able to locate the source already with a single shot 7-chan~ recording. Similar results have been reported recently with another multi-chan~ device (Stefan et al, 1990). Intra-operative electric recordings from the cortical surface and f r o m the del; of the temporal lobe revealed a focus, the 3-dimensional location of which was good agreement with the M E G results. After resection of the right temporal lc (including amygdalohippocampectomy), the subject has now been seizure-free 1 over 1 year.
Epilepsia partial& continua A 14-year-old boy with Rasmussen's encephalitis had repeated periods of twitchi in the right thumb since the age of 3 years. Occasionally the twitching spread to l whole upper limb, sometimes associated with clouding of consciousness. Over | years he gradually developed a mild mental retardation and a mild right-sided hemil resis with loss of fine useful finger movements. M R I and CT scans were negati' Ictal and interictal E E G showed spikes and slow waves over the left central regk At the time of our measurements the right thumb was continuously twitchit Figure 6 shows somatosensory evoked fields (SEFs) to ulnar nerve stimulation a single epileptic spikes recorded from 3 measurement locations over the left somatom sory cortex with the 7-channel device. Both the 20-ms and the 30-ms deflections
[6 pT
MEG I 2 H_
4 5 6
..lJ
. .]
J.A
......
~. . . . . . . . . . . . .
......
J J _ J . = J . . , = = J = -
~ . . , . . * . ~
, . . . . . .
I ................... hJJ_.=~
kl
~
,
I.,
l . . . . .
J. . . . . .
[I.,.L,.~ ........
J" , . , . . , . I
J..I.*lJiJ,.,~.Ll~a
~1
,,
i. . . . .
......
LL
I
...... LI;.a~.
, I.
, ~.=
j
•
. .
•
7
[200 l~V EOG [400 I~V EEG12 "J']"r'+'I" "+ "I " "~-r v-~,........ I"'"I'" ' I " " " +". . . . . . . . . . . . . . i v - "
Fig
'I
......
"
"
. . . . . .
t
" l ' , ' , - -
"-,-
"
,
,"
,+
I
5 cm
1
5. S p o n t a n e o u s b r a i n a c t i v i t y o f a p a t i e n t w i t h t e m p o r a l lobe epilepsy. T h e r e c o r d i n g is f r o m
r i g h t a n t e r o t e m p o r a l a r e a , f r o m l o c a t i o n s s h o w n a t left. Spikes with similar t o p o g r a p h i c d i s t r i b u t i are indicated by arrows: they are directed downwards on Channels 1 and 2 and upwards on Chain 3, 4, a n d 5, i n d i c a t i n g t h a t the i n s t r u m e n t h a s b e e n j u s t o v e r the s o u r c e . T h e p a s s b a u d is 2 - 4 0 H z . r a r r o w s o n t h e s c h e m a t i c h e a d i l l u s t r a t e the l o c a t i o n s o f t h e e q u i v a l e n t dipoles f o r the a u d i t o r y N1C d e f l e c t i o n ( s u p r a t e m p o r a l a u d i t o r y c o r t e x ) a n d f o r t h e s p o n t a n e o u s spikes.
Magnetoencephalography and epilepsy
EVOKEDFIELDS ~ 3. ~~
s
~~/~
Neurophysiol Clin (1990) 20, 169-187 © Elsevier, Paris
Original article
Magnetoencephalography in the study of epilepsy* R Paetau, M Kajola, R Hari L o w Temperature Laboratory, Helsinki University o f Technology, 02150 Espoo, Finland (Received 15 March 1990; accepted 20 May 1990)
Summary - A brief review is given about the basic principles of magnetoeucephalography (MEG), a noninvasive brain research method in which weak magnetic fields are detected outside the h u m a n head with SQUID (Superconducting Q u a n t u m Interference Device) magnetometers. The active brain areas, producing the signal, are modelled by current dipoles, which are assumed to be situated in a spherically symmetric volume conductor. The locations of these "equivalent dipoles" can be found, in the optimal case, with a precision of a few millimeters. The new multichannel magnetometers allow measurements of spontaneous brain activity without EEG-triggered averaging. The 3-dimensional locations of superficial epileptogenic foci can be determined with respect to external landmarks on the skull and to known generator areas of evoked responses in the brain. Examples are given about M E G recordings of epileptic patients. magnetoencephalography / epilepsy / source models / man
R~sum~ - La magn~toenc~phalographie dans I'~tude de l'~pilepsie. Les auteurspr~sentent une brbve revue sur les principes de la magn~todlectroenc6phalographie qui d~tecte de faibles champs magndtiques ~ l'extdrieur du cr~ne au moyen de magndtombtre S Q U I D (superconducting quatum interference device). Les zones cdr~brales aetives, g~n6ratrices du signal sont mod~lis~es sous la f o r m e de courants induits par des dip6les situ~s clans un volumes de conduction sph~rique el symdtrique. La localisation de ces dquivalents dip6les peut ~tre d~finie avec une prdcision de l'ordre de quelques millimdtres. Les nouveaux magndtombtres multicanaux permettent des mesures de l' activitd E E G spontande sans moyennage pr6alable. La localisation en 3 dimensions des f o y e r s ~pileptogbnes peat ~tre prdcisde p a r rapport d des repbres osseux el aux gdndraleurs connus de rdponses ~voqudes : quelques exemples de magndtoencdphalographie de patients dpileptiques sont prdsentds. magn~toenc~phalographie / dpilepsie / moddisation 1 homme
* Based on a talk given by R Hari at the R6union de la Soci6t6 de Neurophysiologie, Clinique, Paris, December 12-13, 1989.
170
R P a e t a u et al
Introduction In western countries 0 . 5 - 1 % of the population suffers from epilepsy. In spite advances in antiepJleptic medication, the seizures o f 20=30% of these patients c~ not be controlled adequately, and surgical treatment is considered. In skilled han~ epilepsy surgery gives good results in 6 0 - 8 0 % of carefully assessed patients (Rz mussen, 1983; Spencer, 1988 ; Silfvenius et al, 1989). The necessary conditions 1 successful outcome are (i) the existence of only 1 trigger area, (ii) accurate locali: tion of it ; and (iii) its situation in a brain region which can be removed without t much harm to the subject. The pre-operative evaluation of epileptic patients is based on a combination neuroimaging techniques, intensive long-term electrophysiological monitoril detailed seizure symptomatology and neuropsychological methods. Although a str tural lesion usually underlies focal-onset epilepsy, its site may remain undetected means of neuroradiological imaging, even in cases of frequent seizures. Further, site of an epileptogenic focus may not always be identical to that of a visible str tural lesion. For accurate localization of the focus, invasive techniques, such as sl dural or depth electrode recordings, are often needed. The non-invas magnetoencephalographic (MEG) recordings, which will be discussed in this bl review, alsoprovide additional information in the selection and pre-operative ev~ ation of candidates for epilepsy surgery. We will first describe the basic principles of the MEG technique and the phy~, logical basis of the observed activity. Then the data analysis methods of epilet MEG phenomena will be discussed. Some results of MEG studies on epileptic tients obtained so far are then described. Finally, some examples of MEG recc ings carried out in our own laboratory are presented. The interested reader is refer to review articles and conference proceedings for more detailed discussions of technical and theoretical basis of neuromagnetism (Williamson and Kaufman, 19 Weinberg et al, 1985" Hari and Ilmoniemi, t,986" Atsumi et al, 1988; Hari Lounasmaa, 1989).
The neuromagnetic technique History The first experimental data about cerebral magnetic fields were obtained in 1968 w D Cohen detected, with a two-million turn induction-coil magnetometer, the Sl: taneous magnetic alpha activity of the human brain (Cohen, 1968). The simult~ ously recorded electric alpha-rhythm signal was used as a trigger for signal averag Introduction of SQUID (Superconducting QUantum Interference Device) magnet, eters significantly improved the sensitivity of the measurements and allowed sg taneous activity to be monitored in real time. The first magnetic evoked resp~
Magnetoencephalography and epilepsy
171
recordings were carried out a few years later (Brenner et al, 1975 ; Teyler et al, 1975) and the field of neuromagnetism started to emerge. Since then S Q U I D magnetometers have been used to detect evoked activity of sensory projection cortices and, more recently, of association areas. A few laboratories have been intensively studying spontaneous brain activity in epileptic patients.
General features o f the M E G method In magnetoencephalography weak cerebral magnetic fields are detected outside the head with sensitive S Q U I D magnetometers (for details of the instrumentation, see Ryh/inen et al, 1989). Evoked cerebral magnetic fields are typically 100-1 000 fT (femtotesla = 10-15 Tesla) in amplitude, only one part in 109 or 10 s of the earth's geomagnetic field. Signals associated with spontaneous activity and epileptic discharges can be considerably stronger. To avoid external magnetic artefacts, the recordings are preferably carried out inside a magnetically shielded room, usually made of mu-metal and aluminium (Kelh~i et al, 1982). The main purpose o f M E G recordings is (i) to detect temporospatial changes in cerebral activity and (ii) to locate source areas with respect to known landmarks on the skull, or preferably in the brain. It is important to know the exact sites and orientations of the magnetometer channels with respect to the head. We obtain this information by placing a small plate containing 3 wire loops to a known site on the skull and by measuring the field pattern produced by currents led through the loops (Knuutila et al, 1987). In determining the source locations one is faced with the "inverse p r o b l e m " , ie, how to deduce f r o m the magnetic field measured outside the head the underlying electric activity within the brain. Unfortunately this problem does not have a unique solution since an infinite number of different current distributions in the brain could produce exactly the same field pattern. Consequently, one has to use source models to describe active brain areas and a volume conductor model to represent the head in which the currents are flowing. A current dipole model is c o m m o n l y used to describe the movement of electric charges in the brain over short distances. An equivalent current dipole represents the net current in an active brain area. The electrical circuit is completed by volume currents from the head of the dipole to its tail in the surrounding medium, so that there is no build-up of charge. A typical strength of the equivalent dipole encountered in evoked response measurements is of the order 10 n A . m (nanoAmpere.meter). For a current dipole within a sphere the distribution of the magnetic field normal to the surface of the sphere has 2 extrema, as illustrated in figure 1. I f a similar field pattern is measured outside the head, the most probable active brain area is halfway between the extrema, with the depth determined by the distance between the field m a x i m u m and the minimum. Usually the source is found by applying a least-squares fit to the measured data. The accuracy o f this solution strongly depends on the signalto-noise ratio, and can be a few millimeters for superficial sources (Hari et al, 1988).
172
R Paetau et al
The head as a volume conductor can be successfully modeled by a sphere. In t/c approximation, only currents which are tangential to the surface (or tangential co~ ponents of tilted dipoles) produce a magnetic field outside, while radial dipoles a externally silent. Because of the orientation of the pyramidal cells and their apic dendrites, the main direction of current flow in the cortex is perpendicular to its st face. Therefore, M E G measures activity mainly from the sulci and fissures of t cortex, which often simplifies the interpretation of the data. Fortunately, the prima auditory, somatosensory, and visual projection areas are situated in the fissural cc tex. When a large brain area of any orientation is active during an epileptic dischar~ the source current usually has a tangential component which can be detected wi MEG. This fact is further emphasized by the typically 10 to 100-fold larger curre strengths associated with epileptic spikes than evoked responses. Physiologically it is reasonable to divide the neural currents associated with po synaptic potentials to (i) transmembrane currents at the active synaptic area, (ii) i tracellular currents within the neuron, and (iii) extracellular volume currents. Ir sphere, the external magnetic field is produced by the intracellular currents aloi as has been shown both theoretically (Hari and Ilmoniemi, 1986) and by compari the polarities of electric and magnetic evoked responses (Hari et al, 1980). In ott geometries, volume currents also become important. A dipole within a sphere c be characterized by means of 5 parameters, 3 for its 3-dimensional location, 1 t its orientation (only the plane parallel to the surface of the sphere is relevant), a 1 for its strength. Instrumentation
The SQUID is a superconducting ring, interrupted by 1 or 2 weak links, called Joset son junctions. It is a parametric amplifier that is very sensitive to magnetic fl, penetrating the ring, and there suitable for detection of the weak cerebral magne fields. The flux transformer (fig 2) is an important part of the magnetometer a can be used to detect special source configurations or to cancel artefacts. A tra: former containing just 1 loop is called a magnetometer; it is very sensitive to sign but also to artefacts. Therefore, a compensation coil is usually added : the first o: er gradiometer cancels artefacts which arise from distant sources and thus prodl similar signals in both loops. With decreasing baseline of the gradiometer, ie, 1 distance between the pickup and the compensation coils, the device becomes less v nerable to distant artefacts but also less sensitive to deep cerebral sources. A plal flux-transformer, also shown in figure 2, can be used to detect field gradients Figure 3 illustrates a typical experimental arrangement for MEG recordings w our 7-channel SQUID gradiometer (Knuutila et al, 1987). The flux transformer s tem consists of 7 first-order axial gradiometer coils in a hexagonal array. The ch~ nels are separated by 3.65 cm and the diameter of the active measurement are~ 9.3 cm. The sensitivity of each channel is 5 - 6 fT x/Hz. A new 24-channel instrument, show in figure 4, was recently completed in
173
Magnetoencephalography and epilepsy
laboratory (Kajola et al, 1990). This device uses a planar flux transformer configuration: with 2 orthogonal loops at each location, the 2 off-diagonal field gradients 8 B,./8 x and 8 Br/8 y are obtained at 12 sites simultaneously. While the axial magnetometer detects the field extrema on both sides of the dipole the 24-SQUID planar gradiometer records the largest signal just above the dipolar source. The sensitivity of the device is 3 - 6 fT/(cm.~H-~). The covered active measurement area, 12.5 cm in diameter, often allows the whole field pattern to be determined with a single measurement. Main differences between E E G and M E G MEG is closely related to electroencephalography, E E G ; both electric and magnetic signals are generated by neural currents. The temporal resolution of both methods is better than a millisecond. To understand the differences between E E G and MEG, let us consider a current dipole in a spherical volume conductor consisting of 4 layers of different conductivities, simulating the brain, the cerebrospinal fluid, the skull
current
dipole
h,,
,\
}.,/
\
:
•
i °
,
.
,'
t
l • s 7 f ' s" . J ." , s ° e ° °°.° f.° .~
s"
."
,r"
i"
Z! \
.-
°.° /
MAGNETIC
FIELD
ELECTRIC
POTENTIAL
Fig 1. Magnetic field pattern and electric potential distribution produced by a current dipole (arrow) in a 4-layer sphere model simulating the head. The continuous lines refer to magnetic flux oriented out of the sphere or to positive potential.
174
R Paetau et al
a n d t h e s c a l p (fig 1). T h e r e s u l t i n g electric p o t e n t i a l d i s t r i b u t i o n a n d m a g n e t i c fi, pattern are both "dipolar",
ie, d i s p l a y 2 e x t r e m a o f o p p o s i t e p o l a r i t i e s , b u t a r e
t a t e d b y 90 d e g r e e s w i t h r e s p e c t to e a c h o t h e r . T h e i s o c o n t o u r lines are r e l a t i v e l y t i g h t e r in t h e m a g n e t i c t h a n in t h e electric p t e r n . T h i s is b e c a u s e c o n c e n t r i c electric i n h o m o g e n e i t i e s s m e a r t h e e l e c t r i c p o t e n t i
Axial
Planar
Fig 2. Schematic illustration of flux transformer configurations : an axial first order gradiometer ( and a planar gradiometer (right). Magnetic flux directed along the thick arrows induces a current in coil (small arrows). With the axial gradiometer, 2 field maxima are detected, 1 on each side of the cur dipole. The figure-of-eight planar flux transformer, being sensitive to the field gradient, detects the rn mum signal when the difference between the fluxes threading the 2 loops is largest. For a current diF this occurs just above the source. (Figure kindly provided by Dr Jukka Knuutila).
Magnetoencephalography and epilepsy
175
preamplifiers
channel configuration
SQUIDs within niobium shields
liquid helium (-269°C)
gradiometer coils
(°",*J vacuum cast
Fig 3. Typical experimental arrangement for MEG recordings. The subject lies on a bed, with his head supported by a vacuum cast to prevent movements. The concave b o t t o m of the dewar is placed as close to the head as possible. Within the dewar, the SQUID detectors and flux transformers are immersed in liquid helium at T = - 2 6 9 °C. Adapted from Hari and Lounasrnaa (1989).
176
R Paetau et al
Fig 4. Photograph of the measuring head of our 24-channe! magnetometer (Kajola et al, 1990). Exam of 2 figure-of-eight loops have been drawn on the figure.
Magnetoencephalography and epilepsy
177
whereas the magnetic pattern remains intact. Further, while the measured electric potential distribution is affected by the location of the reference electrode, no corresponding problem exists in the magnetic recordings. Consequently, MEG has good spatial resolution for superficial cortical sources. On the other hand, EEG is sensitive to both radial and tangential sources and also reflects activity of the deepest parts of the brain.
Artefacts Since the magnetic brain signals are very weak, rejection of external disturbances is of extreme importance. Magnetic noise is caused, for example, by cars and elevators, by radio, and by power transmission lines. For the best results, MEG measurements should be performed in a magnetically shielded room. Biological tissues also generate fields which can disturb MEG signals. For example, the heart's magnetic field over the chestis 2 - 3 orders of magnitude larger than the brain signals outside the head. This type of disturbance can be decreased by using a gradiometric flux transformer. Eye movements and blinks also cause artefacts which are large in comparison with the cerebral signals (Antervo et al, 1985). We routinely apply electro-oculogram to reject contaminated signals. In rare cases muscular artefacts may also cause problems. The subject should not wear clothes containing magnetized metallic material since these generate artefacts in the rhythm of respiration or of cardiac activity. A modern source of artefacts, observed in our studies of epileptic children, is produced by intraoral metallic devices for orthodontics. Post-operative M E G evaluation may be impossible if magnetized material is used in clips, sutures, etc.
Methods to analyze MEG data of epileptic patients Advanced data analysis methods have been developed for interpreting neuromagnetic evoked fields. Localization of the generators of spontaneous magnetic brain activity, measured sequentially at several locations, is much more difficult since the activity does not necessarily repeat itself. Different analysis methods are available to determine the source configuration.
Signal averaging In locating the generation site of MEG discharges it is common to apply averaging : the simultaneously recorded EEG signal is used as a trigger for MEG waveforms, which are recorded sequentially from several locations (Barth et al, 1982). Unfortunately, the choice of the trigger channel largely determines the observed field pattern, and the method only works if well discernable spikes or sharp waves are present.
178
R Paetau et al
Relative covariance The Rome group (Chapman et al, 1984) introduced a method where relative covai ances are calculated between 1 electric channel and the magnetic recordings. The da are normalized by the amplitude of the electric signal since the activity can vary co~ siderably as a function of time, This metho d allows studies of irregular signals. Agai: the drawback is that only activity reflected in the electric recordings will be detect¢ and that the field patterns will be dissimilar when different EEG derivations are chose as the reference.
Spectral maps Useful information about source locations can be obtained by calculating frequen. spectra of the signals and by mapping the abundance of different frequencies at va ious locations. When multi-channel recordings are available, phase-spectra betwe~ different channels suggest which signals arise from the same source.
Instantaneous determination of the field pattern With the most advanced multi-channel M E G instruments it is already possible locate the source with a single measurement which greatly increases the accuracy the method (Stefan et al, 1990; Tiihonen et al, 1990). Using devices which cover wi areas of the head, it is possible to find the equivalent source at any instance al also to employ signal averaging, if needed, by first choosing a template with whi the activity will be compared. It will then be possible to automatically classify t abnormal waveforms into different classes and to find out their source locatio separately.
Qualitative recording Accurate source localization is not always required: one might be interested to s the signal w a v e f o r m and the temporal changes in the qualitative features of the fie pattern. These types of measurements, which resemble the classical use of EEG reco~ ings can be useful in screening candidates for surgical treatment: clear changes the field pattern from 1 spike to another discourage the assumption of a local on~ for the discharge, although they do not rule out a c o m m o n trigger.
Reference sites for source localization It is c o m m o n to determine the source locations with respect to external landmal of the skull. Localization in relation to known sources of some evoked respons like the area 3b in the posterior wall of the central sulcus or the auditory area arou the gyrus of Heschl at the superior bank of the temporal lobe, might in the futt
Magnetoencephalography and epilepsy
179
become more important since it can help the surgeon to avoid areas which are functionally irretrievable. For example, some neurosurgeons stimulate the cortex and consider the representation area of the tongue the most posterior border for temporal lobe resection. By recoding magnetic responses to short electric pulses applied to the tongue (Karhu et al, 1990) it has become possible to pre-operatively localize the sensory tongue area with respect to the epileptic focus.
MEG studies of epilepsy In the first MEG recording of an epileptic patient (Cohen, 1972), slow EEG and MEG activity increased during hyperventilation; the patient had psychomotor epilepsy. Later, more subjects were studied in several laboratories. Modena et al (1982) reported, in patients with focal epilepsy, a sharp localization of paroxysmal magnetic activity. In patients with Rolandic foci the MEG findings were positive even in the absence of EEG abnormalities. This difference was probably due to the tangential orientation of the source and also due to the fact that a more dense measurement grid was used in the MEG than the EEG recordings. The group at the University of California in Los Angeles localized both single sources and multiple epileptogenic areas (Barth et al, 1982, 1984). One interesting finding was a systematic delay of 20 ms between left and right temporal spikes with respect to an electric signal. Such time differences can be used to discriminate between the primary and secondary (mirror) foci. In another interesting study both scalp and sphenoidal electrode spikes were used as triggers for averaging the magnetic spikes (Sutherling and Barth, 1989). A consistent depth difference was observed in the equivalent source areas of these 2 groups of spikes. The finding was suggested to indicate propagation of the discharge from deeper to more superficial brain areas. Recently, magnetic slow shifts occurring during epileptic seizures have also been recorded (Vieth et al, 1990).
Own examples Temporal epilepsy An adult female patient with medically intractable epilepsy was studied (for detailed data, see Tiihonen et al, 1990). She had suffered from complex partial seizures for over 2 decades. In spite of frequent seizures, the MRI and CT scans were negative. The magnetic field was measured at several locations over the right hemisphere with our 7-channel large-area magnetometer. The spontaneous magnetic activity showed abundant spikes which reversed their polarity between the anterior and posterior measurement locations over the right temporal area. Sometimes the magnetic spikes reversed polarity even in one 7-channel measurement, as shown in figure 5.
180
R P a e t a u et al
By constructing topographic field maps it was possible to follow the temporal e~ lution of the activity during the spike : the location of the equivalent source travel] about 20 m m during 23 ms. Activation of the auditory cortex, monitored by m e uring evoked magnetic fields, was used as a landmark for the source location: ( source of the epileptogenic activity was about 2 cm anterior to the auditory cort, It was encouraging to be able to locate the source already with a single shot 7-chan~ recording. Similar results have been reported recently with another multi-chan~ device (Stefan et al, 1990). Intra-operative electric recordings from the cortical surface and f r o m the del; of the temporal lobe revealed a focus, the 3-dimensional location of which was good agreement with the M E G results. After resection of the right temporal lc (including amygdalohippocampectomy), the subject has now been seizure-free 1 over 1 year.
Epilepsia partial& continua A 14-year-old boy with Rasmussen's encephalitis had repeated periods of twitchi in the right thumb since the age of 3 years. Occasionally the twitching spread to l whole upper limb, sometimes associated with clouding of consciousness. Over | years he gradually developed a mild mental retardation and a mild right-sided hemil resis with loss of fine useful finger movements. M R I and CT scans were negati' Ictal and interictal E E G showed spikes and slow waves over the left central regk At the time of our measurements the right thumb was continuously twitchit Figure 6 shows somatosensory evoked fields (SEFs) to ulnar nerve stimulation a single epileptic spikes recorded from 3 measurement locations over the left somatom sory cortex with the 7-channel device. Both the 20-ms and the 30-ms deflections
[6 pT
MEG I 2 H_
4 5 6
..lJ
. .]
J.A
......
~. . . . . . . . . . . . .
......
J J _ J . = J . . , = = J = -
~ . . , . . * . ~
, . . . . . .
I ................... hJJ_.=~
kl
~
,
I.,
l . . . . .
J. . . . . .
[I.,.L,.~ ........
J" , . , . . , . I
J..I.*lJiJ,.,~.Ll~a
~1
,,
i. . . . .
......
LL
I
...... LI;.a~.
, I.
, ~.=
j
•
. .
•
7
[200 l~V EOG [400 I~V EEG12 "J']"r'+'I" "+ "I " "~-r v-~,........ I"'"I'" ' I " " " +". . . . . . . . . . . . . . i v - "
Fig
'I
......
"
"
. . . . . .
t
" l ' , ' , - -
"-,-
"
,
,"
,+
I
5 cm
1
5. S p o n t a n e o u s b r a i n a c t i v i t y o f a p a t i e n t w i t h t e m p o r a l lobe epilepsy. T h e r e c o r d i n g is f r o m
r i g h t a n t e r o t e m p o r a l a r e a , f r o m l o c a t i o n s s h o w n a t left. Spikes with similar t o p o g r a p h i c d i s t r i b u t i are indicated by arrows: they are directed downwards on Channels 1 and 2 and upwards on Chain 3, 4, a n d 5, i n d i c a t i n g t h a t the i n s t r u m e n t h a s b e e n j u s t o v e r the s o u r c e . T h e p a s s b a u d is 2 - 4 0 H z . r a r r o w s o n t h e s c h e m a t i c h e a d i l l u s t r a t e the l o c a t i o n s o f t h e e q u i v a l e n t dipoles f o r the a u d i t o r y N1C d e f l e c t i o n ( s u p r a t e m p o r a l a u d i t o r y c o r t e x ) a n d f o r t h e s p o n t a n e o u s spikes.
Magnetoencephalography and epilepsy
EVOKEDFIELDS ~ 3. ~~
s
~~/~
