Modification of neuromagnetic cortical signals by thalamic infarctions

J.P. Makelaa,b,*, R. Salmelina, M. Kotilac, O. Salonend, R. Laaksonenc,L. Hokkanenc, R. Haria,c

aBrain Research Unit, Low Temperature Laboratory, Helsinki University of Technology, PO Box 2200, 02015 HUT, Espoo, FinlandbCentral Military Hospital, PO Box 50, 00301 Helsinki, Finland

cDepartment of Clinical Neurosciences, Helsinki University Central Hospital, 00290 Helsinki, FinlanddDepartment of Radiology, Helsinki University Central Hospital, 00290 Helsinki, Finland

Accepted for publication: 17 December 1997

Abstract

Auditory evoked responses and spontaneous cortical activity were recorded with a whole-scalp 122-channel neuromagnetometer from 7patients, who had small thalamic infarctions in the region of the left anterior tuberothalamic artery and associated memory defects. Incontrast to healthy control subjects, with dominant rhythmic activity at 10.6± 0.6 Hz in the parieto-occipital region, the spectral maximumin the patients was at 8.9± 0.4 Hz. Abnormal acceleration of rhythmic activity was also observed bilaterally in rolandic areas. Our findingsimply that lesions of non-specific thalamic nuclei may disturb human brain rhythms in widespread cortical areas. ‘Mismatch responses’ todeviant tones (1.1 kHz) among standards (1.0 kHz), suggested to reflect sensory auditory memory in healthy subjects, were absent in 2patients, markedly decreased in 3, and normal in 2, implying that pathways passing through the anteromedial thalamus contribute tomodulation of these responses. We conclude that local unilateral lesions in the anteromedial thalamus may cause extensive, bilateralalterations in the brain’s electric activity. 1998 Elsevier Science Ireland Ltd.

Keywords:Thalamic stroke; Memory; Alpha rhythm; Mismatch field (MMF); Magnetoencephalography (MEG); Man

1. Introduction

Thalamic infarctions in the region of the anterior tubero-thalamic artery, i.e. in the anteromedial thalamus, oftencause memory disorders, probably by severing the mamil-lothalamic tract (Kritchewsky et al., 1987) and the pathwaysfrom the hippocampus and the medial temporal lobe to thefrontal cortex (Kew et al., 1993). The more detailedmechanisms of the memory disruption are, however,unknown. Although it is quite difficult to explore humanthalamic function non-invasively, indirect information canbe obtained by studying alterations of cortical activity afterthalamic disorders.

The interplay between the thalamus and the cortexappears to form the basis of generation of some spontaneousbrain rhythms (Steriade et al., 1990). Medial thalamic nucleimediate the activating effects of the brain-stem reticularformation on the cortical activity, and their lesion couldthus disrupt normal brain rhythms. The spontaneous oscil-

latory brain activity has been suggested to bind single ele-ments of the cerebral circuitry into global functional statesnecessary for cognitive functions (Llinas, 1988). In linewith this hypothesis, high alpha frequency in EEG record-ings has been suggested to be correlated with good memoryperformance (Klimesch et al., 1993). Furthermore, theextent of suppression of spontaneous activity, predomi-nantly in the parietal areas, depends on the type and com-plexity of the memory task (Dujardin et al., 1995). Thus,cognitive defects following thalamic infarctions could beassociated with alterations in the patients’ spontaneous cor-tical activity.

Thalamocortical connections are indispensable for pro-cessing of sensory stimuli, and their disruption could changeneural representations of external stimuli. In healthy sub-jects, a comparison process between a new stimulus andtraces left by the previous stimuli has been suggested toproduce ‘mismatch responses’, which reflect proper func-tioning of auditory sensory memory. These responses areelicited by infrequent deviants among frequent standardtones (Na¨atanen, 1992). Frontal cortical lesions decreasethe electric mismatch response ipsilateral to the lesioned

Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

0013-4694/98/$19.00 1998 Elsevier Science Ireland Ltd. All rights reservedPII S0013-4694(98)00005-4 EEG 97631

* Corresponding author. Tel.: +358 9 4512954; fax: +358 9 4512969.

side (Alho et al., 1994). Anterior thalamic lesions, disrupt-ing connections between frontal cortex, thalamus, andhippocampal formation could have a similar effect. Themismatch responses are essentially normal in patients withmemory defects due to Alzheimer’s disease (Verleger etal., 1992; Pekkonen et al., 1994); we are not aware of simi-lar studies in patients with other types of memory disrup-tion.

To elucidate the effect of thalamic infarctions on brainelectrophysiology, we analyzed spontaneous brain activityand mismatch responses in 7 patients with lesions in theanteromedial thalamus, and compared them with age-matched control subjects. We also compared the observedalterations with neuropsychological findings in this patientgroup. The measurements were performed with a whole-scalp 122-channel neuromagnetometer (for a review ofthe method, see Ha¨malainen et al., 1993). The data havebeen previously presented in abstract form (Ma¨kela et al.,1994).

2. Patients and methods

2.1. Patients

Seven patients (4 men, 3 women, mean age 48 years,range 24–63 years) with left-sided thalamic infarction inthe anterior tuberothalamic artery area were studied. Thelesions were evident in the patients’ magnetic resonanceimages (MRIs) (see Fig. 1); obtained with a 1-T SiemensMagnetomy instrument. Patients had well-delineatedlesions in the left thalamus, including the dorsomedialnucleus and anterior thalamic nuclei. The lesions usuallyextended into the mamillothalamic tract, the periventricularnuclei, the anterior part of intralaminar nuclei, and the cen-trum medianum (see Table 1).

Patient 1 used oral contraceptives, Patient 7 was on hor-mone therapy for menopausal symptoms, Patient 4 hadhypertension, and Patient 5 had migraine and used antide-pressive medication. Aortocervical angiographies were nor-mal in 6 out of 7 patients; in Patient 4, left the vertebralartery was not visualized, probably due to hypoplasia.

Table 2 shows the neurological and behavioral symptomsin the acute stage. All patients had memory disturbances.Four patients were disoriented in time and/or place, 4 weresomnolent and 3 confused. Five patients had mild dysphasiaand 2 a mild hemiparesis. Two patients had a vertical gazeparesis. Two patients were psychotic for a short period.Inactivity and depression occurred as well. Five patientswere unaware of their memory disturbances at the acutestage.

Neuropsychological examinations were done both in theacute stage and during follow-up. In the acute stage, afteramelioration of the acute confusion, all patients had dysno-mia, problems in word retrieval, and they used circumloca-tions in their speech. Five patients had dysphasia. They also

had defective learning which appeared both in associativelearning of new verbal material as well as in immediaterecall of short passages of prose. Two patients had a ten-dency to confabulation. A specific feature of the memoryimpairment was a severe and persistent difficulty in recal-ling names of familiar persons and learning new propernames, without comparable difficulty in recognizing orassociating faces. Memory span was mostly intact. Theseneuropsychological findings in the acute stage have beendescribed previously in detail (Kotila et al., 1994). Fourpatients tried to return to their previous profession, butonly two, a cosmetologist (Patient 1) and a semi-profes-sional salesperson (Patient 2), succeeded. At the time ofthe MEG measurements, Patient 5 used 25 mg Thioridazineand 7.5 mg Zopiclone daily; other patients did not use psy-choactive drugs.

The neuropsychological follow-up of the patients wasperformed on average 27 months (range 2 months to 8years) after the injury. It included Wechsler memory scale(WMS) (Wechsler, 1945) and the Wechsler adult intelli-gence scale (WAIS) (Wechsler, 1955), the Benton visualretention test (Benton, 1974) and a 10-word list learningtask with 5 trials (Christensen, 1976). MEG measurementswere done in the stable phase, 13 months to 11 years afterthe injury. The MEG data were compared with those ofseven healthy controls (mean age 47 years, range 32–72years). MRIs of all control subjects were classified as nor-mal. The experiment was approved by the Ethics Commit-tee of the Department of Neurology, Helsinki UniversityCentral Hospital; both patients and controls gave theirinformed consent to the study.

2.2. Experimental set-up

Binaural 70-dB hearing level (HL) tones were delivered tothe subjects through plastic tubes and earpieces and wereheard at the head midline. The subjects were instructed toignore the stimuli. In experiment 1, 1-kHz 50-ms tones weredelivered once every 2 s. Four patients also received the tonesat interstimulus intervals (ISIs) of 0.5, 1 and 4 s. In experi-ment 2, tones were delivered in an oddball paradigm with anISI of 0.6 s; standard tones (85%) were 1 kHz in frequencyand deviants (15%, randomly interspersed among the stan-dards) were 1.1 kHz. In experiment 3, spontaneous corticalactivity was measured for 4 min when the subject was restingwith the eyes open and for 4 min when the eyes were closed.

2.3. Recording and data analysis

Auditory evoked fields (AEFs) and magnetic spontaneousactivity were recorded in a magnetically shielded room witha whole-scalp SQUID (superconducting quantum interfer-ence device) magnetometer (Neuromag 122y, Ahonen etal., 1993). The 122 sensor units of the device are arranged ina helmet-shaped array and measure the two orthogonal tan-gential derivatives ofBz, the magnetic field component nor-

434 J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

mal to the helmet surface, at each of the 61 measurementsites. Planar gradiometers couple most strongly to currentsjust below the sensor.

During the recordings, the subject was seated under thehelmet-shaped neuromagnetometer with the head leaningagainst its bottom. The exact location of the subject’shead with respect to the sensors was determined by measur-ing magnetic signals produced by small currents deliveredto 3 coils attached to the scalp. The locations of the coils

with respect to the nasion and preauricular points weredetermined with a 3-D digitizer; this information wasused to align the MEG data with the coordinates of theMRIs.

The recording passband was 0.03–100 Hz (3 dB points,high-pass roll-off 35 dB/decade, lowpass over 80 dB/dec-ade). The sampling rate was 0.4 kHz and the analysis period400 ms (including a 50-ms pre-stimulus baseline). The ver-tical electro-oculogram (EOG) was recorded simulta-

Fig. 1. MRIs of the 7 patients with left-sided infarction in tuberothalamic artery region. The site of the lesion is indicated by a circle. The insets show thelesions in enlarged form.

Table 1

Affected thalamic nuclei

Patient Sex/age Anterior Medial Dorso-medial

Anteriorintralaminar

Centrummedianum

Lateralposterior

Ventralposterior

1 F/24 + + + + + − −2 F/43 − + − − + − −3 M/44 + + + + − − −4 M/63 + + + + − − −5 M/59 + − + + ? − −6 M/54 + + + + − − −7 F/51 + + + + + ? ?

?, possible lesion.

435J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

neously and epochs in which the EOG activity exceeded±150 mV were rejected from the on-line averaging.Responses were averaged over 100 stimulus presentationsin experiment 1 and over 500 presentations in experiment 2.

The averaged responses were digitally low-pass filtered at40 Hz. The head was modeled as a spherical volume con-ductor, matched to the local curvature of the inner skull overthe auditory areas of the temporal lobe. Sources of N100mdeflections, peaking about 100 ms after the tone onset, weremodeled as equivalent current dipoles (ECDs) found by aleast-squares fit to signals recorded by a subset of 18–24channels over the auditory areas of each hemisphere. OnlyECDs accounting for at least 85% of the field variance wereaccepted. Thereafter, the analysis period was extended tothe entire response and the source strengths as a function oftime were computed assuming a two-dipole model with onedipole in each hemisphere. Noise levels were defined duringthe pre-stimulus period and were used to calculate confi-dence intervals for the ECD locations. The data of Patient6, who had no N100m and no mismatch field (MMF) wereomitted from the analysis.

The peak amplitudes and mean latencies of the MMFsover each hemisphere were examined from differencewaveforms obtained by subtracting responses to standardsfrom those to deviants. The sensor pair showing the largestsubtraction response was selected from the vector sums ofthe two orthogonal tangential derivatives. MMF was con-sidered to be present if its amplitude at 130–200 msexceeded the noise level, defined as two standard deviationsof the pre-stimulus signal variation. When no MMF wasobserved, the maximum amplitude in this time windowwas measured from the channel showing maximumN100m responses; only amplitudes, not latencies, wereincluded in the further analysis. MMFs in 6 hemispheresof the patients were not adequately modeled by a singleequivalent current dipole; therefore, source strengths werenot compared between patients and controls.

Amplitude spectra of spontaneous cortical activity werecalculated from the whole 4-min periods with eyes open andclosed by advancing a 3.4-s time window in 1.7-s steps, and

averaging the subsequent spectra. To facilitate comparisonbetween subjects, vector sums of the spectra in each sensorpair were calculated and averaged over 7–10 pairs overright and left frontal, rolandic, temporal and occipitalregions. The peak spectral amplitudes were defined fromthese areal mean spectra. Subject 6 and Patient 7 had nodefinite spectral peaks in the parieto-occipital region andwere excluded from the statistics. The mean spectral energyin the 6.5–9.5 and 9.5–12.5 Hz bands was obtained from theaveraged spectra by dividing the total power in each band bybandwidth in Hz. The ratio of the mean energy in the lowerversus upper band was calculated for each individual.

Source locations of individual deflections of rhythmicactivity over the parieto-occipital areas and over the regionof the central sulci and sylvian fissures were identified from2-min recordings during eyes open and closed periods. Thesignals were digitally filtered through 4–5 individuallyselected frequency bands between 4 and 32 Hz. Thereafter,using a subset of 36–40 sensors over the parieto-occipitalcortex and over the remaining left and right hemispheres,ECDs were sought every 10 ms. Sources were accepted ifthey accounted for at least 90% of the measured field var-iance within the selected subregion, and had confidencevolumes below 125 mm3. The number and average strengthof accepted ECDs were calculated within the individualsub-bands over the parieto-occipital area and over the leftand right temporal/rolandic regions.

If not stated otherwise, statistical comparisons betweenthe groups were done using unpaired, 2-tailedt tests; forcomparisons within the groups, paired, 2-tailedt tests wereapplied. The repeated measures of analysis of variance wereperformed with the Systat 5.0 program.

3. Results

3.1. Neuropsychological evaluation

Table 3 summarizes the performance of each patient inthe neurological follow-up tests. The memory had improved

Table 2

Neurological and behavioral symptoms after left thalamic infarction

Patient 1 2 3 4 5 6 7

Disoriented in time/place + − + − − + +Somnolent + + − + − + −Confused + − − − + + −Memory disturbance + + + + + + +Dysphasia + + − + + − +Right hemiparesis + − − + − − −Vertical gaze paresis + + − − − − −Depressive reactions − − + − + + −Anxious − − − − + + −Psychotic with paranoid ideas − − − − + + −Inactive − + + − − + −Unaware of memory disturbances − + + + + + −

436 J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

from the acute stage in all patients. In Patient 1, the onlynotable finding was the semantic memory deficit in theWAIS information subtest. In all other patients, the verballearning was defective. Patients 2 and 7 had impairments inboth verbal and visual memory as well as in attention (mea-sured with the mental control subtest of WAIS). Patients 3and 5 had retrieval deficits, more prominent in the verbalmaterial. All patients performed close to the average level intests of digit span.

3.2. Auditory evoked responses

Tones delivered once every 2 s elicited similar N100mresponses in the patients and controls. In the patients, thepeak latencies were 103± 13 and 96± 10 ms and the dipolestrengths were 46± 13 and 41± 17 nAm over the left andright hemispheres, respectively. The corresponding valuesin the controls were 95± 6 and 92± 6 ms and 47± 32 and38 ± 16 nAm; the differences between the groups were notsignificant. The source locations did not differ significantlybetween the patients and controls; they were 4 mm moreanterior and 10 mm more lateral in the right than left hemi-sphere in the patients and 5 mm more anterior and 8 mmmore lateral in the controls.

In the 4 patients tested, the increase in ISI from 1 to 4 sincreased the N100m amplitude, on average, by 16% in theright and by 25% in the left hemisphere; in healthy subjects,the corresponding increase is about 200% (Ma¨kela et al.,1993).

Fig. 2 depicts AEFs of Subject 1 and Patient 2. AEFs tostandard tones were of similar magnitude in both. Theresponses to deviants and standards differed clearly in thecontrol subject whereas no corresponding difference wasdetected in the patient.

Fig. 3 shows, for all subjects, the difference waveformsobtained by subtracting responses to standards from thoseto deviants. A brisk MMF 130–200 ms after the tone onsetis evident in 12 out of 14 control hemispheres. In thepatients, the signal in this time range exceeds the noiselevel in 6 out of 12 hemispheres, but only for a short periodin 2 of them.

The maximum MMF amplitudes over the left and righthemispheres were 35± 28 and 37± 24 fT/cm, respectively,in the controls and 14± 10 and 15± 17 fT/cm in thepatients. When pooled over both hemispheres, MMF wassignificantly smaller in the patients than in the controls(P , 0.02, Mann-WhitneyU-test); the difference alsoreached significance in the left hemisphere alone (P ,0.05, Mann-WhitneyU-test). In the controls, MMF peakedat 170± 17 ms over the left and at 156± 17 ms over theright hemisphere; the corresponding values in the patientswere 166± 28 and 179± 16 ms.

A single dipole explained over 80% of the MMF patternin 11 out of 14 hemispheres in the controls, and in 6 out of12 in the patients. In two patients, the patterns were dipolarover both hemispheres; however, the MMF source locationsdid not differ from those of N100m, contrary to that usuallyobserved in healthy subjects (Leva¨nen et al., 1996).

Table 3

Performance of each patient in the neurological follow-up tests, in Wechsler’s adult intelligence scale (WAIS), in Wechsler’s memory scale (WMS), inBenton visual retention test (Benton), and in 10-word list learning task.

Patient 1 2 3 4 5 6 7 Max.

Education (years) 14 9 14 20 17 20 10WAIS information 7 11 12 14 12 15 10 19Similarities 13 12 12 17 12 12 12 19Vocabulary 12 n.d. 10 13 15 11 14 19VIQ 103 105 104 130 120 118 113Digit symbol 12 10 13 10 10 – 8 19Picture completion 9 10 15 10 11 – 10 19Block design n.d. 9 19 9 14 – 10 19PIQ 101 99 134 112 125 n.d. 110WMS mental control 6 5 8 7 6 9 3 9Stories 14.5 8 7 12 8.5 12.5 10 23Stories 1 h delayed 14.5 5.5 7 10 n.d. 5 n.d. 23Span 6+ 5 6 + 5 6 + 5 6 + 4 7 + 4 6 + 5 6 + 5Figures 14 7 14 13 13 6 10 14Word pairs 21 8.5 9 13 5.5 14.5 5 21MQ 123 90 105 126 105 112 92Benton 7 9 7 n.d. 8 7 5 10List learning up to 5 trials 7-9-10 5-6-6-7-7 4-6-6-5-6 6-7-6-7-5 6-5-8-5-6 7-8-5-6-10 4-5-5-6-4 10Neuropsychological follow-up time 2 months 13 months 14 months 7 months 7 years

6 months4 years 8 months

MEG after stroke 1 year2 months

2 years7 months

3 years5 months

5 years6 months

11 years1 months

9 years7 months

2 years6 months

VIQ, verbal intelligence quotient; POQ, performance intelligence quotient; MQ, memory quotient; n.d., non-definable values.

437J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

3.3. Spontaneous activity

Fig. 4 illustrates the frequency spectra of spontaneousactivity during the eyes-closed period in Subject 1 and inPatient 1. Subject 1 has clearly delineated spectral peaks atabout 11 and 22 Hz over the occipital region. Patient 1 has abroad spectral maximum at 6–12 Hz, with sharp additionalpeaks overriding it at 5.8, 8.8 and 11.3 Hz. This activity issuppressed by opening the eyes, in line with the suppressionseen in normal subjects.

Fig. 5 illustrates the averaged parieto-occipital spectra ofcontrols and patients. The peak frequency was 10.6± 0.6Hz in the controls and 8.9± 0.4 Hz in the patients(P , 0.001, Mann-WhitneyU-test).

Fig. 6 demonstrates the mean (±SEM) spectral amplitudeper frequency unit in the subregions in the 6.5–9.5 and 9.5–12.5 Hz bands in the patients and controls. In analysis ofvariance with repeated measures, patients and controls dif-fered very significantly in the spectral content (F = 8.53, df1,P , 0.005); furthermore, interaction with the two subjectgroups and the spectral amplitude in different bands wassignificant (F = 4.0, df 1,P , 0.05). Within subject groups,region (F = 43.3, df 4,P , 0.001) and eye opening versusclosing (F = 7.2, df 4,P , 0.001) affected the spectral con-tent significantly. In addition, spectral amplitude in differentbands interacted significantly with region and subjectgroups (F = 2.67, df 4,P , 0.03).

In the controls, the integrated spectral energy in theparieto-occipital region was 12.1± 7.5 fT/cmÎHz in the6.5–9.5 Hz range and 22.8± 14.0 fT/cmÎHz in the 9.5–12.5 Hz range (P , 0.02); all control subjects had larger

Fig. 2. AEFs to standard (dashed line) and deviant (solid line) tones in Control 1 and in Patient 2. The head is viewed from above and the nose pointsupwards. The sensor array is flattened onto a plane. Responses from the sensors showing maximum amplitudes are enlarged in the insets.

Fig. 3. Difference waveforms (responses to standards subtracted fromthose to deviants) in controls (C1–7) and in patients (P1–5, P7). Patient6 did not produce reliable N100m or MMFs. Dashed lines show the± 2SD values of the pre-stimulus level; the deflections exceeding these valuesdiffer from noise with 95% probability.

438 J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

values in the upper frequency range. The correspondingvalues in the patients were 25.1± 16.6 and 17.7± 6.4 fT/cmÎHz (n.s.). Four patients had larger values in the lowerthan higher range. Patients had more energy in the lowerband and in all other subregions as well; the difference wassignificant in the rolandic regions, most prominently overthe left rolandic area. In addition, patients had more spectralenergy than the controls in the higher band in the rightrolandic region during the eyes open period. Frontal regions,

not displayed in the Fig. 6, did not show differences betweenthe groups.

The amplitude ratio of lower versus higher band in theeyes closed period over the parieto-occipital region was1.36± 0.54 in the patients and 0.61± 0.26 in the controls(P , 0.01). In the left temporal region the correspondingvalues were 1.46± 0.55 in the patients and 0.80± 0.22 inthe controls (P , 0.01), and in the right temporal region1.35± 0.51 and 0.83± 0.27 (P , 0.03). The differences

Fig. 4. Spectra of spontaneous cortical activity measured with the 122 sensors in Control 1 and in Patient 1 when they were resting with eyes closed. Thespectra from sensors showing maximum activation are displayed enlarged in the insets.

Fig. 5. The areal mean amplitude spectra of parieto-occipital cortical activity (range 2–35 Hz) in all patients (P) and controls (C). The vertical lines indicatethe 9.5-Hz limit. Solid line indicates eyes closed period and dotted line eyes open period.

439J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

did not reach significance in the eyes open period or in theother brain regions. The spectral amplitude of the parieto-occipital 6.5–12.5 Hz band after eye opening was 62± 14%of that in the eyes closed period in the patients and49 ± 14% in the controls (n.s.).

At frequencies above 15 Hz, patients and controls did notdiffer in the average spectral amplitudes or in the numberand average strength of accepted spontaneous activityECDs. However, at lower frequencies, a larger number ofECDs exceeded the acceptance criteria in the patients thanin the controls (P , 0.03 in the parieto-occipital region,eyes closed). The mean source strengths were 36–64 nAmin the eyes open and 46–77 nAm in the eyes closed period,with no systematic differences between the patients andcontrols.

In the eyes open period, the source locations of 3–15 Hzspontaneous activity in patients were clustered along thecentral sulcus, extending to temporo-parieto-occipital junc-tion, and around the parieto-occipital sulcus in the headmidline, as shown in Fig. 7. There were extensive individualdifferences in the number of localized sources. The sameregions appeared to produce spontaneous activity in the

controls as well; however, the patients had more sourceswithin each of the above mentioned regions than the con-trols, and localizable low-frequency sources were moreabundant in them. These differences in source numberreflect the abundance of individual sources of the rhythmicactivity. Closing the eyes increased the amount of accepteddipolar sources particularly in the parieto-occipital region,both in the patients and controls.

4. Discussion

The blood flow to thalamus stems from the four mainvessels (tuberothalamic, paramedian, inferolateral and pos-terior choroideal arteries), with considerable individual var-iation. Infarctions in the anterior tuberothalamic andparamedian region are known to produce confusion andmemory dysfunction. Dysphasia suggests affision of theleft anterior thalamus whereas hemineglect implies aright-sided lesion; vertical gaze paresis and muscular weak-ness suggest paramedian infarction (Bogousslavsky et al.,1988). The semiology of our patients thus implies lesion inthe anteromedial thalamus; no traces of hemianopia, imply-ing posterior choroideal stroke, or somatosensory dysfunc-tion, implying inferolateral infarction, were observed.

The infarctions, depicted in MRIs, extended to thedorsomedial and anteromedial thalamic nuclei, usuallyincluding the mamillothalamic tract and the intralaminarand periventricular nuclei. The dorsomedial nucleus ispredominantly connected to the granular prefrontal cortexbut it also has dense connections with the intralaminarand periventricular nuclei. Anteromedial and anteroventralthalamic nuclei send projections to the cingulate cortex,hippocampus, and also to the temporo-parieto-occipital cor-tex (Armstrong, 1990). Periventricular and intralaminarnuclei are integrated into the reticular system. The efferentsfrom these nuclei project sparsely and diffusely to the entireneocortex and have a significant role in mediating theeffects of reticular formation on cortical activity. In thecat, intralaminar nuclei have robust reciprocal projectionsto frontal eye fields, parietal areas 5 and 7, and to auditory,visual, and insular cortices (Kaufman and Rosenquist,1985).

The rich thalamocortical and intrathalamic connectivitysuggests that a small, circumscribed thalamic lesion mayhave widespread effects on both thalamic and cortical func-tions. Accordingly, our results demonstrate that a local uni-lateral lesion in anteromedial thalamus may disrupt theelectric brain activity extensively and bilaterally. Slowingof EEG has been observed over the side of a lesion inpatients with memory disorders due to stroke in the para-median artery region (Ghidoni et al., 1989). MEG record-ings, having better spatial accuracy than EEG, thus confirmthat widespread electric alteration. This is also in line with ahuman PET study showing bilateral cortical hypometabo-lism after unilateral thalamic lesions (Baron et al., 1986).

Fig. 6. The mean spectral energy of patients (P) and controls (C) in theeyes closed (EC; black bars) and eyes open (EO; white bars) periods overthe left and the right rolandic, the left and the right temporal, and over theparieto-occipital areas. Significant differences at theP , 0.05 (x) andP , 0.03 (xx) levels are indicated.

440 J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

Similarly, unilateral lesion in the rat ventromedial thalamicnucleus has been shown to decrease local glucose utilizationbilaterally in cingulate, sensorimotor and visual cortices, aswell as in the dorsomedial and reticular thalamic nuclei.These metabolic alterations have been explained by thespread of abnormal activity through callosal connections(Girault et al., 1985).

Both the behavioral effects and electrophysiologicalalterations generated by the thalamic infarcts showed con-siderable variation from subject to subject, possibly becauseof individual variability in the affected thalamic nucleiand in their connectivity. Memory has been previouslydescribed to remain intact in some patients after damageto the dorsomedial thalamus, probably because of thepreserved mamillothalamic pathway (Kritchewsky et al.,

1987). In cats, experimental lesions in the midline thalamusproduce EEG slowing in an extremely variable manner interms of amplitude, irregularity and focality, and thalamiclesions of similar size and location produce greatly differenteffects between animals (Gloor et al., 1977).

A crucial role has been suggested for thalamocorticalconnections in generating spontaneous brain oscillations(Steriade et al., 1990). Most prominent spectral peaks ofspontaneous MEG activity are seen in the vicinity of theparieto-occipital sulcus, and over somatomotor areas (Hariand Salmelin, 1997). Our data show that these rhythms areaffected by anteromedial thalamic strokes, with no directdisruption of sensory pathways. Thalamic lesions in ourpatients did not extend to the medial and lateral geniculateor ventroposterolateral nuclei, thus saving the primary audi-

Fig. 7. Clusters of equivalent current dipoles for spontaneous activity in the 3–15 Hz band in two patients and two control subjects, overlaid on the subjects’individual brain surface renderings. The illustrated sources explained more than 90% of the regional field variance. Note that an increase in spectral energy ina certain frequency range does not necessarily lead to an increase in localizable sources.

441J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

tory, visual, and somatosensory pathways. Normal function-ing of the auditory pathways is also supported by normalAEFs in our patients; most probably, visual pathways, pas-sing close to auditory ones in the thalamus, have alsoremained unaffected by the lesion. Thus, it seems evidentthat lesions in non-specific thalamic nuclei can affect cor-tical rhythms. Two mechanisms may mediate this modifica-tion. First, dysfunction in, e.g. connections between parietalassociation areas and intralaminar nuclei could disturb theparieto-occipital 10–12 Hz activity. Second, the modulationcould occur at thalamic level, e.g. through the reticular tha-lamic nucleus. The present data do not allow distinctionbetween these two possibilities.

The prominent neuromagnetic 10-Hz rhythm in the par-ieto-occipital region (Salmelin and Hari, 1994) was stronglyaffected in our patients. It appears that this rhythm is notonly linked to processing of visual input. For example,spontaneous magnetic 8–12 Hz activity is suppressed overparieto-occipital region areas when subjects perform tasksrequiring recall of letter shapes for visual imagery (Saleniuset al., 1995). Activation of the precuneus in the parietalmidline, close to the generation site of this rhythm, hasbeen observed in a PET experiment requiring recall ofpaired associates in tasks related to episodic memory(Fletcher et al., 1995). Occipital EEG slows down in somepatients having memory dysfunction due to Alzheimer’sdisease; however, similar slowing may occur without directrelationship to memory disorders, like in non-dementingparkinsonism, and during normal aging (for a review seeBrenner, 1993). Although alteration of the parieto-occipitalspontaneous activity was clear in our group data, it did notpredict any applied measure of neuropsychological assess-ment on an individual level.

Four out of 6 patients had altered MMFs to deviant sti-muli. These alterations were not related to memory disor-ders in any simple, straightforward way; two patients withgross memory disorder had normal MMFs, and one patientwith no MMF had mild memory disorders. Mismatchresponses, mainly generated in the supratemporal auditorycortex (Hari et al., 1984), have been attributed to compar-ison of afferent input elicited by deviant stimuli with aneural sensory memory trace generated by standard stimuli(Naatanen, 1992). The sensory memory implicated by mis-match responses differs from immediate-recall type shortterm memory, which can be evaluated by digit span exam-ination, and which was shown to be normal in our patients.In this phonological loop memory, information is dividedinto segments, and maintained by repetitive vocal or sub-vocal rehearsal. By contrast, the auditory sensory memorythat underlies mismatch response generation reflects acous-tic attributes of stimuli; the memory trace left by the pre-vious stimuli is maintained without rehearsal. Thus thedifferences between the two markers reflecting short termmemory are not surprising.

The 4 tested patients had an abnormally short recoverycycle of the N100m response, suggesting shortened duration

of sensory memory (Picton et al., 1978; Lu¨ et al., 1992;Sams et al., 1993). At the neural level, N100m suppressionat short ISIs has been related to active inhibition in thecortex (Loveless et al., 1989), which thus may haveincreased by lesions in the medial thalamus.

In summary, infarctions in the left anteromedial thalamusimpaired the patients’ memory, disrupted the magneticspontaneous activity in extensive cortical areas, and dimin-ished magnetic responses to deviant tones. The results showthat diencephalic strokes may modify cortical function faraway from the lesioned structure.

Acknowledgements

The work was supported by the Academy of Finland. Wethank Prof. S. Sarna for comments on statistical analysis ofthe data.

References

Ahonen, A., Hamalainen, M., Kajola, M., Knuutila, J., Laine, P.,Lounasmaa, O.V., Parkkonen, L., Simola, J. and Tesche, C. 122-channelSQUID instrument for investigating the magnetic signals from thehuman brain. Phys. Scr., 1993, T49: 198–205.

Alho, K., Woods, D., Algazi, A., Knight, R. and Na¨atanen, R. Lesions offrontal cortex diminish the auditory mismatch negativity. Electroen-ceph. clin. Neurophysiol., 1994, 91: 353–362.

Armstrong, E. Limbic thalamus: anterior and mediodorsal nuclei. In: G.Paxinos (Ed.), The Human Nervous System. Academic Press, SanDiego, CA, 1990, pp. 469–481.

Baron, J., D’Antona, R., Pantano, P., Serdaru, M., Samson, Y. andBousser, M. Effects of thalamic stroke on energy metabolism of thecerebral cortex. Brain, 1986, 109: 1243–1259.

Benton, A. The Revised Visual Retention Test. The Psychological Cor-poration, New York, 1974.

Bogousslavsky, J., Regli, F. and Uske, A. Thalamic infarcts: clinical syn-dromes, etiology and prognosis. Neurol., 1988, 38: 837–848.

Brenner, R. EEG and dementia. In: E. Niedermeyer, F. Lopes Da Silva(Eds.), Electroencephalography: Basic Principles, Clinical Applicationsand Related Fields. Williams and Wilkins, Baltimore, MD, 1993, pp.339–349.

Christensen, A. Luria’s Neuropsychological Investigation. Munksgaard,Copenhagen, 1976.

Dujardin, K., Bourriez, J. and Guieu, J. Event-related desynchronization(ERD) patterns during memory processes: effects of aging and taskdifficulty. Electroenceph. clin. Neurophysiol., 1995, 96: 169–182.

Fletcher, P., Frith, C., Baker, S., Shallice, T., Frackowiak, R. and Dolan, R.The mind’s eye-precuneus activation in memory-related imagery. Neu-roimage, 1995, 2: 195–200.

Ghidoni, E., Pattacini, F., Galimberti, D. and Aguzzoli, L. Lacunar thala-mic infarctions and amnesia. Eur. Neurol., 1989, 29 (Suppl. 2): 13–15.

Girault, J.-A., Savaki, H., Desban, M., Glowinski, J. and Besson, M.J.Bilateral cerebral metabolic alterations following lesion of the ventro-medial thalamic nucleus: mapping by the 14-deoxyglucose method inconscious rats. J. Comp. Neurol., 1985, 231: 137–149.

Gloor, P., Ball, G. and Schaul, N. Brain lesions that produce delta waves inthe EEG. Neurol., 1977, 27: 326–333.

Hari, R., Hamalainen, M., Ilmoniemi, R., Kaukoranta, E., Reinikainen, K.,Salminen, J., Alho, K., Na¨atanen, R. and Sams, M. Responses of theprimary auditory cortex to pitch changes in a sequence of tone pips:neuromagnetic recordings in man. Neurosci. Lett., 1984, 50: 127–132.

442 J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443

Hari, R. and Salmelin, R. Human cortical oscillations: a neuromagneticview through the skull. Trends Neurosci., 1997, 20: 44–49.

Hamalainen, M., Hari, R., Ilmoniemi, R., Knuutila, J. and Lounasmaa, O.Magnetoencephalography: theory, instrumentation and applications tonon-invasive studies of the working human brain. Rev. Mod. Phys.,1993, 65: 413–497.

Kaufman, E. and Rosenquist, A. Efferent projections of the thalamic intra-laminar nuclei in the cat. Brain Res., 1985, 335: 257–279.

Kew, J., Goldstein, L., Leigh, P., Abrahams, S., Cosgrave, N., Passingham,R., Frackowiak, R. and Brooks, D. The relationship between abnormal-ities of cognitive function and cerebral activation in amyotrophic lateralsclerosis. Brain, 1993, 116: 1399–1423.

Klimesch, W., Schimke, H. and Pfurtscheller, G. Alpha frequency, cogni-tive load and memory performance. Brain Topogr., 1993, 5: 241–251.

Kotila, M., Hokkanen, L., Laaksonen, R. and Valanne, L. Long-term prog-nosis after left tuberothalamic infarction: a study of 7 cases. Cerebro-vasc. Dis., 1994, 4: 44–50.

Kritchewsky, M., Graff-Radford, N. and Damasio, A. Normal memoryafter damage to medial thalamus. Arch. Neurol., 1987, 44: 959–962.

Levanen, S., Ahonen, A., Hari, R., McEvoy, L. and Sams, M. Deviantauditory stimuli activate human left and right auditory cortexdifferently. Cereb. Cortex, 1996, 6: 288–296.

Llinas, R. The intrinsic electrophysiological properties of mammalian neu-rons: insights into central nervous system function. Science, 1988, 242:1654–1664.

Loveless, N., Hari, R., Ha¨malainen, M. and Tiihonen, J. Evoked responsesof human auditory cortex may be enhanced by preceding stimuli. Elec-troenceph. clin. Neurophysiol., 1989, 74: 217–227.

Lu, Z.-L., Williamson, S. and Kaufman, L. Behavioral lifetime of humanauditory sensory memory predicted by physiological measures. Science,1992, 258: 1668–1670.

Makela, J.P., Ahonen, A., Ha¨malainen, M., Hari, R., Ilmoniemi, R.,Kajola, M., Knuutila, J., Lounasmaa, O.V., McEvoy, L., Salmelin, R.,

Salonen, O., Sams, M., Simola, J., Tesche, C. and Vasama, J.-P. Func-tional differences between auditory cortices of the two hemispheresrevealed by whole-head neuromagnetic recordings. Hum. Brain Map-ping, 1993, 1: 48–56.

Makela, J.P., Salmelin, R., Kotila, M. and Hari, R. Neuromagnetic corre-lates of memory disturbances caused by infarction in anterior thalamus.Soc. Neurosci. Abstr., 1994, 20 (1): 810.

Naatanen, R. Attention and Brain Function. Lawrence Erlbaum, Hillsdale,NJ, 1992.

Pekkonen, E., Jousma¨ki, V., Kononen, M., Reinikainen, K. and Partanen,J. Auditory sensory memory impairment in Alzheimer’s disease: anevent-related potential study. NeuroReport, 1994, 5: 2537–2540.

Picton, T., Campbell, K., Baribeau-Braun, J. and Proulx, G. The neuro-physiology of human attention. In: J. Requin (Ed.), Attention and per-formance. Lawrence Erlbaum, Hillsdale, NJ, 1978, pp. 429–467.

Salenius, S., Kajola, M., Thompson, W., Kosslyn, S. and Hari, R. Reac-tivity of magnetic parieto-occipital alpha rhythm during visual imagery.Electroenceph. clin. Neurophysiol., 1995, 95: 453–462.

Salmelin, R. and Hari, R. Characterization of spontaneous MEG rhythmsin healthy adults. Electroenceph. clin. Neurophysiol., 1994, 91: 237–248.

Sams, M., Hari, R., Rif, J. and Knuutila, J. The human auditory sensorymemory trace persists about 10s: neuromagnetic evidence. J. Cogn.Neurosci., 1993, 5: 363–370.

Steriade, M., Gloor, P., Llina´s, R., Lopes Da Silva, F. and Mesulam, M.-M.Basic mechanism of cerebral rhythmic activities. Electroenceph. clin.Neurophysiol., 1990, 76: 481–508.

Verleger, R., Ko¨mpf, D. and Neuka¨ter, W. Event-related EEG potentials inmild dementia of the Alzheimer type. Electroenceph. clin.Neurophysiol., 1992, 84: 332–343.

Wechsler, D. A standardized memory scale for clinical use. J. Psychol.,1945, 19: 87–95.

Wechsler, D. Wechsler Adult Intelligence Scale. Psychological Corpora-tion, New York, 1955.

443J.P. Makela et al. / Electroencephalography and clinical Neurophysiology 106 (1998) 433–443