ORIGINAL ARTICLES
Brain Electrical Activity Mapping (BEAM): A Method for Extending the Clinical Utility of EEG and Evoked Potential Data Frank H. Duffy, MD, James I. Burchfiel, PhD, and Cesare T. Lombroso, MD
The difficulties inherent in extracting clinically useful information by visual inspection alone from the massive amounts of data contained in multichannel polygraphic recordings have placed limits on the accuracy and range of utility of electroencephalography and evoked potentials. A method for condensing and summarizing the spatiotemporal information contained in recordings from multiple scalp electrodes is described. Data dimensionality is reduced and visibility increased by computer-controlled topographic mapping and display of data as color television images. Examples are given in which such brain electrical activity mapping (BEAM) (1) localizes tumors in patients with normal or nondiagnostic EEGs, ( 2 ) adds additional information to that visible on computerized axial tomography, and (3) demonstrates electrophysiological abnormalities in patients with functional lesions but normal CT scans. A sensitivity to the functional component of a neurological lesion suggests that BEAM may provide complementary information to the anatomical definition provided by the CT scan. Duffy FH, Burchfiel JL, Lombroso C T Brain electrical activity mapping (BEAM):a method for extending the clinical utility of EEG and evoked potential data. Ann Neurol 5:309-321,1979
Computerized axial tomography-the C T scan-has largely replaced cerebral angiography and pneumoencephalography for the diagnosis of anatomical abnormalities of the brain. The impact of the CT scan upon electroencephalography has been more variable. This stems from the use of EEG to diagnose not only anatomical lesions (e.g., tumor, abscess) but functional lesions as well (e.g., epilepsy). Whereas EEG may fail to diagnose 15% of brain tumors [3], it remains the best or only test for epilepsy and certain other functional disorders (e.g., Reye syndrome, brain death). The clinical utilization of evoked potentials (EP) to assist in the diagnosis of various sensory disorders [6, 9, 10, 341, brainstem lesions [351, and demyelinating diseases [22, 281 has increased. Nevertheless, the clinical value of EEG and EP in the diagnosis of certain forms of purely functional lesions is still limited. For instance, there are many reports of EEG and EP abnormalities in children with dyslexia 117, 19, 33, 36, 371; however, these findings are too nonspecific to form useful diagnostic constellations. We do not believe that EEG and EP suffer from an
~
inherent insensitivity to underlying brain dysfunction. O n the contrary, we propose that brain electrical activity presents not too little, but too much information to be easily grasped and assimilated by visual inspection alone. The approach we have taken t o enhance its utility in the diagnosis of both anatomical and functional lesions is based upon topographic mapping. In this technique, EEG and EP data recorded from multiple scalp electrodes are graphically displayed o n a computer-driven color video screen. Values between electrodes are obtained by interpolation. We call this technique brain electrical activity mapping, or BEAM. This report describes the methods and results of a feasibility study illustrating the ability of BEAM to define both anatomical and functional lesions in selected cases.
Method Data from 24 standard EEG scalp electrodes are amplified via a 20-channel polygraph (Grass Model 78) and recorded
on a 28-channel FM analogue tape recorder (Honeywell 5600E) for off-line computer processing (Digital Equip-
~~~~
From the Seizure Unit and Division of Neurophysiology, Department of Neurology, Children’s Hospital Medical Center and Harvard Medical School, Boston, MA.
Address reprint requests to Dr F. H. Duffy, Seizure Unit, Children’s Hospital Medical Center, 300 Longwood Ave, Bosron, MA 02115.
Accepted for publication Aug 21, 1978.
0 3 6 4 - ~ 1 3 4 / 7 ~ / 0 4 0 3 0 9 - 1 3 $ 0 1 .@ 2 5 1978 by Frank
H.Duffy 309
ment Corp PDP-12). Data analysis of three types is performed under the SIGSYS-12 biomedical software package (Agrippa Data Systems). First, spectral analysis from 0 to 32 Hz is performed o n background EEG using the fast Fourier transform (FFT) technique. To avoid artifacts from muscle and mains (60 Hz) activity, signals are tightly filtered at 24 d b per octave using active Butterworth filters (EEG Associates Mark 4 x 24). The amount of energy in each classic EEG band (e.g., delta, theta, alpha, and beta) is determined for each electrode and stored for future display. Second, visual EPs of 5 12 msec duration are formed from stimulation by randomly presented strobe flashes (Grass PS-2) set 20 cm from the subject's eyes at intensity level 2. Trials containing muscle or blink artifact are automatically excluded. The average EP voltage over 128 epochs each lasting 4 msec is determined for each electrode and stored for later display. Third, epochs of raw EEG signals containing epileptic spikes are digitized. Next, a 5to 20-msec epoch containing the feature of interest (spike) is averaged and stored for later display. By means of a hardware digital interface (Agrippa Data Systems 8ETV3) between the computer and an ordinary Sony color television set, results are displayed within a graphic outline of the head. Individual topographic displays of four sorts are commonly prepared: (1) spectral energy in any individual EEG band, (2) EP voltage at any 4-msec epoch after stimulus onset, ( 3 ) EEG amplitude during the peak of an epileptic spike, and ( 4 )results of statistical analysis on any of the preceding features (e.g., standard deviation). For each display, scalp areas around the original 24 electrode values are filled by linear three-dimensional interpolation based on the values at the three nearest electrodes (see Figure 1 for details of this process). Furthermore, the interface hardware is capable of displaying a new video image every 100 msec; accordingly, images may be cartooned at variable rates of up to 10 frames per second in an endless-loop manner. Thereby, for example, one can observe an EP in both space (a single topographic video image) and time (a cartooned series of images representing 512 msec in real time).
Subjects and Tests Twelve patients (7 with tumors, 1 with a suspected tumor, 3 with dyslexia, and 1 with a sociopathic personality) of the many referred for clinical EEGs and EPs were evaluated using the described techniques; the results of these evaluations are included in this report. These patients were carefully chosen to demonstrate the clinical utility of BEAM: they were selected to compare the ability of BEAM to define anatomical lesions (tumors) and functional lesions (dyslexia and behavioral disorders). Seven right-handed subjects with normal EEGs and normal neurological examinations were used as controls. The 12 patients were right-handed and had either normal or nondiagnostically abnormal EEGs containing no focal or lateralizing features. Seven ambulatory patients with unoperated supratentorial brain tumors diagnosed by C T scan were studied with spectral and visual EP topography. The tumors included gliomas of all four lobes of the right hemisphere and all but the frontal lobe of the left hemisphere. One patient, sus-
310 Annals of Neurology
Vol 5
No 4 April 1979
pected by C T scan of having a right frontal tumor, was similarly studied. This 13-year-old girl was admitted with a one and one-half year history of brief periods of abnormal C N S function. These periods included episodes of bizarre behavior, focal seizures of the right side of the face, transient left-sided hemianopia, and nominal aphasia. Due to a more recent history of headache and lethargy, a CT scan was performed which revealed areas of density in the right frontal lobe with shift of the midline structures from right to left (Fig 2). An angiogram confirmed a right frontal avascular mass. The patient was admitted for surgical exploration. Neuropsychological testing, however, revealed that in addition to nondominant frontal abnormalities, she showed bilateral parietal deficits. Three neurologically intact adolescents of normal intelligence, referred with the diagnosis of specific reading disability (dyslexia), were evaluated. In addition to resting spectral and visual EP studies, their spectral topographies were examined while they were listening to music or speech. O n e neurologically intact adolescent patient with a history of sociopathic behavior was studied with resting spectral and visual EP topography. His EEG was considered borderline due to excessive theta slowing in the anterior quadrants; however, his C T scan was within normal limits.
Results Controls Figure 3 demonstrates a BEAM spectral plot for a normal subject in the eyes-open and the eyes-closed state. For this subject and all the control subjects, spatial distributions were those one might predict using classic EEG analysis. For example, beta activity was maximal anteriorly (eyes open), and alpha activity tended to be maximal in the midline leads and posterior leads with a slight right-sided predominance (eyes closed). Figure 4 demonstrates five representative frames of a 128-frame visual evoked potential (VEP) examination in which each frame represents 4 msec. The symmetrical appearance of activity in the -occipital regions should be noted; the sequential appearance of symmetrical vertex waves is also illustrated. All normal subjects demonstrated spectral and VEP plots similar to those shown in Figures 3 and 4. Tumor Patients Figure 5 demonstrates the abnormal spectral plot for a patient with a left occipital glioma (his CT scan is shown in Figure 2 for comparison). Note the increased delta, decreased beta, and decreased alpha activity over the tumor site. Six of the 7 tumor patients showed the same abnormalities that this subject demonstrated, namely, increased slow activity and decreased fast activity overlying the tumor site. Although the seventh subject had a normal spectral plot, the spatial distribution of the standard deviation
P3 p4
F7 T4
73
4-
$
T5 01
A
B
C
D
F i g I . Example of the construction of a topographic map for E P data. Mean EPs are formed from each of 20 recordiq sites. Each EP is divided into 128 4 - m e c intervals, and the mean voltage value for each interval is calculated. In (A) the individual EPs are shown for the electrode locations indicated on the head diagram. In IB) the mean voltage values at these locations are shown for the interval beginning I 9 2 msec after the stimulus (the vertical line in A indicates this time on the EPs). Next the head region is treated as a 64 x 64 matrix; the resulting 4,096 spatial domains are illustrated in (C). Each domain is assigned a voltage value by linear interpolation from the three nearest known points. Finally, for display, the raw voltage values are fitted t o a discrete-level equal-interval intensity scale as shown in ( D ) .An illustration of topographic maps in two-color display for other selected intervals over the time course of the EP is shown in Figure 6.
Duffy et al: Brain Electrical Activity Mapping
311
Fig 2. Each photograph represents the resnlts of a computerized axial toniographic radiographic examination. ( A ) The primary lesion i n the patient u!ith the left occipital glionia shown in Figures 5 and G is marked with un X. Note the dilatation of the lefi lateral ventricle; this may explain the disruption of forzrjard spread of actitJitynoted in the VEP BEAM ( F i g 6 ) . ( B ) The lesion in the patient with lymphomatoid granuloniatosis is marked by an arrow. C T scan lesions were limited to the right frontal lobe. No abnormalities were seen i n the parietal regions to correlate with the biparietal abnormalities obserred by neurop.ryc-hologii-altesting and V E P BEAM (see F i g 8).
312
Annals of Neurology Vol 5
No 4
April 1979
F i g .3. Example of a normal BEAM spectral plot in the eyesopen (EO; A-D) and eyes-closed (EC; E-H) state, formed according to the method outlined in Figure 1 ; however, spectral fiinctions instead of EPs are used as input. Each spec~irmis dirfided u p into the classic E E G spectral bands (i.e.9 delta = 0 l o 4 H z , theta = 4 t o 8 H z , rrlphu = 8 t o 12 HL,und beta = 12 t o 20 Hz). Each plot is scaled i n d i d u a l I y . Note the ex-
pected asymmetry of alpha activity, which is greater over the right occipzrt for this right-handed subject. Note the general shift i n energy from the anterior to the posterior regionsfor all frequencies with EC, a frequentjnding in normal persons. Note the posterior beta asymmetry with EC; this i s seen when beta includes the 12 t o 16 H z range, adjacent t o alpha. I t is not seen fnr the 16 to 3 2 H z range (not shown here).
Duffy et
al:
Brain Electrical Activity Mapping 313
Fig 4
of his spectral energy (not illustrated) demonstrated localization to the site of his tumor. Figure 6 demonstrates a typical BEAM VEP plot for tumor subjects. All 7 tumor patients showed VEP plots demonstrating: (1) depression or absence of early electrical activity at the site of the tumor; (2) abnormally persistent late activity overlying the tumor; (3) asymmetrical spread of activity; and (4) diminished or absent vertex waves. In every patient, the topographically demonstrated abnormalities extended beyond the strict anatomical limits of the lesion on C T scan (for example, see the lesion size in Figure 6 as compared to the C T scan lesion in Figure 2). The ability of measurements taken from topographic plots to differentiate tumor patients from normal subjects is shown in Figure 7. The time spent in asymmetry at the locus of maximum asymmetry is plotted for all 7 controls and the 7 tumor patients. Note that a line drawn at 7 5 msec completely separates these two groups.
Patient with Suspected Tumor Figure 8 demonstrates the abnormal BEAM VEP plot for the 13-year-old girl with the suspected right frontal lobe tumor by CT scan and angiogram. This VEP BEAM showed repetitive midline parietal ab~
F i g 4. Example of a normal BEAM VEP plot formed according t o the method described in Figure 1. Five representative plots of the total of 128 plots spanning 512 msec are shown. Each plot represents a 4-msec averaged epoch. In this and all other VEP plots, red represents positive and blue represents negative activity with respect t o linked ear references. At 40 msec note the frontal electsoretinogram IERG). The primary EP wave begins with maximum amplitude overlying PZ, just forward of the occipital region 152 msec); this is a normal variant. This activity spreads to the “vertex”at CZ for the first vertex wave (1 08 msec). Note the slight vertex wave asymmetry, which is within normal limits. The next positive component,just beginning at 108 msec in the occipital region, sweeps up t o CZ t o form the next vertex wave (168 msec). Much kater, another vertex waveforms (440 msec). Normal VEPs consist of a series of vertex waves with positive or negative wavefronts converging on the vertex region. Asymmetries of early vertex waves are not uncommon. The ERG is a common contaminant (40 msec), as is some temporalis muscle activity (i.e., left ear at 52 msec). F i g 5. Example of a tumor (left occipital glioma) demonstrated & a BEAM spectral plot, formed as for Figure 3 in the EC state. Note the increased leji occipital delta, markedly decreased /eft occipital alpha, and slightly decreased lefi occipital beta activity. Theta shows only a slight increase on the left. Although alpha asymmetries are seen in normal subjects ( F i g 31, a marked depression of alpha activity overlying one occiput, as in thisjigure, is not. (See also Figure 6 for the VEP plot of this subject.)
normalities, possibly involving both medial parietal lobes, and bifrontal abnormalities. Thus, in contrast to the radiographic studies, BEAM demonstrated multifocal abnormalities that were more in line with the clinical history and with neuropsychological testing. At operation no tumor was found. The subsequent diagnosis by examination of biopsied material was lymphomatoid granulomatosis, a disease entity that may involve the CNS in a multifocal manner [20, 241. Dyslexics
None o f the dyslexic patients showed abnormalities on standard spectral plots. However, when they were studied under special circumstances, asymmetries were noted (not illustrated). When these subjects were asked to listen to music, alpha activity was suppressed over the right hemisphere as in many normal subjects [12, 21, 231. When they were asked to listen to speech, however, the left hemisphere alpha suppression often seen in normal subjects was not noted [12, 21, 231. The standard VEP plots of the dyslexics were normal. Unusual visual EP features in other special test situations will be the subject of a future report.
Patient with Behavioral Abnormality Although the spectral plot for the sociopathic patient was within normal limits, his visual EP topography showed consistent abnormalities consisting of recurring asymmetries of activity overlying the frontal lobes (Fig 9). Discussion Interpretation of EEG and EP records requires correlation of large volumes of data across both space and time. We propose that the difficulties involved in making spatiotemporal correlations by unaided visual inspection of multichannel polygraphic recordings constrain the clinical utility of EEG and EP. Accordingly, we have developed a technique for the topographic display of scalp-recorded signals on a color television monitor. Such topographic mapping reduces the dimensionality of data in a twofold manner. First, topographic mapping presents spatial information in a more concise and summarized form. For example, one must ordinarily examine many pages of EEG to appreciate subtle slow-wave asymmetries. With spectral BEAM plots, such asymmetries can be summarized in a single image (see Figs 3 , 5 ) . Second, linking together individual images, as in endless-loop cinematography (cartooning), enables the visualization of temporal patterns of brain activity that are not otherwise easily perceived. For example, EP analysis
Duffy et al: Brain Electrical Activity Mapping 315
Fig 6. BEAM VEP plot, showing 42 of the 128 frames generated from the 512-msec VEP of a patient with a kji occipital brain tumor, formed as in Figure 4. Note the initial negativity beginning at PZ (72 msec), which then sweeps forward primarily on the left. The first occipital positivity appears only on the right 188 to I12 msec). In a similar manner, the second occipital positivity appears asymmetrically on the right. Next, it sweeps simultaneously i n two directions: ( 1 ) forward on the right side exclusively, and (2) across the midline t o activate the left occipital region (I 52 to 21 6 msec). The kji occipital positivity persists despite disappearance of the right occipital positivity (I 60 t o 296 msec), reaching a maximum between 200 and 224 msec. The delayed activation of the &Jt occipital region overlying the tumor and the prolonged retention of activity i n this region, once activated, are typical of abnormalities associated with tumors. Note also the lack of forward spread of activity from the leji occipital region to the kji frontal region; we hypothesized that this might represent the electrical correlate of a functional disconnection between the k f t occipital and frontal lobes caused by the tumor (i.e., interrup-
tion of the left occipitofrontalpatbway). The late waves then
repeat the same sequence. Note the right occipital negativity developing at 344 msec, which then sweeps forward on the right only (424 to 472 msec); notice also that it grws in amplitude on the right (432 t o 448 msec) well before it spreads t o the kji occipital region 1456 to 504 msec). Again, the left side i s late and persistently activated (504 msec +). Thus the process offirst appearance in the right occipital region, spread forward on the right only, and late activation of the lefi occipital region from the right was repeated twice. It is of interest t o note that this patient had a completeb normal standard neurological examination, including visual field studies; the E E G was also normal. The chical impression was migraine headache. The significance of the peak of positivity appearing in the left frontal region ( I 5 2 to 224 msec) is not known (perhaps it represents the effect of deafferentationfrom the leji occipital region). Neuropsychological tests were not performed on this patient; therefore, one cannot comment about the possibility of a defect of frontal lobe function. (See also Figure 5 for the spectral plot of this subject.)
Duffy et al: Brain Electrical Activity Mapping
317
250
1
200
-
175
150
-
-
125
25 5o
1
0 NORMALS
TUMORS
F i g 7 . This graph demonstrates the possibility of differentiating patients with tumor from normal subjects on the basis of a single measurement from their topographic maps. All tumor patients showed marked asymmetries overlying the tumor in their VEP BEAM; most showed more than one episode of asymmetry at the same location lsee F i g 6). In this figure we plot the total length of time for all period of asymmetry at the locus of the maximal asymmetry for 7 normal subjects and 7 tumor patients. Note that the tumor patients had a much greater period of asymmetry than the normal persons. Furthermore, a line drawn at 75 msec serves t o separate all normal (< 75 msec)from all tumor subjects I> 130 msec).
usually consists of measuring amplitudes and latencies of various waveform components recorded from several electrodes. The temporal cartooning of EP BEAM images formed from multichannel data allows the visualization of traveling wavefronts of positive or negative electrical activity. Thus waveform components appearing as differing latencies at different scalp sites may, in fact, represent the same wavefront slowly progressing across the scalp surface (see Fig 6). BEAM does not create new data but merely presents the original spectral, EP, or EEG data in a more concise and intuitively comprehensible format. Despite the fact that their routine clinical EEGs were normal or nonlocalizing, all 7 tumor patients examined in this study were differentiated from normal subjects on the basis of a simple measurement obtained from their BEAM studies (see Fig 7). Although BEAM may not prove to be as accurate or consistent as the CT scan in defining anatomical lesions, its success rate exceeds that of the routine EEG. We suggest that BEAM may significantly increase the percentage of space-occupying lesions that can be electrophysiologically defined. In certain instances (see Fig 8) BEAM may provide more clinically useful information than is found on CT scan. The ability of BEAM to identify functional differences in the absence of anatomical lesions was also demonstrated. An EP BEAM test on an adolescent sociopath, chosen for having a normal CT scan, revealed prominent frontal lobe asymmetries (see Fig
318 Annals of Neurology Vol 5
N o 4 April 1979
9). Some authors have invoked frontal lobe dysfunction as the cause underlying abnormal behavior in sociopaths [26, 271. Functional changes were also observed in 3 adolescent dyslexics of normal intelligence and with normal EEGs. Although standard spectral and EP BEAM studies were within normal limits, alpha activity was not suppressed over the left hemisphere by listening to speech. Many investigators have implicated the left parietal lobe as the abnormal site in dyslexia on the basis of behavioral observation and testing and as an analogy to aphasia [2, 7, 8, 13, 331. The findings suggest that BEAM may provide a useful electrophysiological tool for studying the genesis of functional abnormalities of the brain. BEAM is not the first method to be used for the topographic representation of EEG or EP activity. In the 1950s and 1960s devices called toposcopes or encephaloscopes, developed by pioneers such as Harold Shipton and Grey Walter [31, 32, 391, used multiple-intensity modulated cathode ray tubes to display data in space and time. In recent years, others have topographically displayed computer-processed EEG or EP data via techniques varying from handdrawn contour plots [ l , 14, 15, 381 to computerdrawn contour or gray-scale plots [4, 5, 11, 16, 18, 25, 29, 301. The compressed spectral array of Bickford et al[4, 51 and the canonogram of Gotman, Gloor, and Ray Cl61 have proved to be clinically useful in the demonstration of localized brain lesions. Two of the most sophisticated display systems were implemented by Harris et al [ 181and Estrin and Uzgalis [l 11 as early as 1969. Both systems involved computer plotting of electrophysiological data. It is surprising that many of the more recent articles have used much simpler display systems, including handdrawn plots. In a 1976 review article, Petsche [25] F i g 8 . This figure illustrates how BEAM may add additional information to that provided by radiographic evaluation. Five selectedframes of 128 computed images are displayed as in Figure 4. The patient was a 13-year-old girl with a history of multiple episodes of transient neurological dysfunction and a more recent history of headaches and lethargy (see text). The C T scan lsee F i g 2) and angiogram suggested a right frontal lobe tumor. However, both neutvpsychological testing and thc VEP BEAM revealed bilateral parietal dysfunction as well as bifrontaldisease. The VEP BEAM, shown in this figure, reveals: (1) long-lasting bilateral prefrontal waves of both positive (1 64 msec) and negative 1292 msec) polarity, 12) a peculiar absence of activity (48, 116 , 164 msec) of the midparietal region, ( 3 ) an unusual period of polarity reversal between the left and right parietal regions and the midline parietal region (84 msec),and (4) a lesser deficiency of activation over the midfrontal region I1 64 and 292 msec). No tumor was found at the time of craniotomy. The neuropathologicaldiagnosis was lymphomatoid granulomatosis, a process that may involve the CNS in a multifocal manner.
F i x 9. This figure shou1.r a functional abnormality demonstrated by V E P topography in the presence of a normal C T scan (no anatomical lesion). Three selected frames of 128 computed images are displayed as in Figure 4. The patient was an 18-year-old man referred for evaluation with a hi.itory of sociopathic behavior. Although his C T .scan was within normal limits, his behavior suggested frontal lobe dysIysfirnction. His V E P plot. shown here, demonstrates recurring asymmetries in the frontal lobes of both positive and negatire activity. 7'hi.r. we believe, represents the neurophysiologicalcorrelate of-the expectedfrontal lobe beharioral dy.rfunction seen i n the presence of a normal C T scun. Such topographic maps may protme usefir1 i n finding electrnphysiologicalcorrelates of behacioral disorders.
Fig 8
suggested features to be incorporated into a topographic display system of the future. Many of these features, however, had already been implemented by Harris et a1 [18] in 1969. T h e reasons for the lack of application of technology already developed may be twofold. First, the early systems may have been operationally difficult, inconvenient, or expensive to use for other than a circumscribed research project. Second, the better systems appear to have been developed by engineers, and the failure to see these systems widely applied to clinical problems may have reflected a comprehension or credibility gap between engineers and physicians. The BEAM system described in this report takes full advantage of modern advances in solid-state electronics technology (see Scienti$c American, Sept 1977). A key feature of BEAM is a simple and easy to use interface between a small biomedical computer and a standard color television set. At present, computation time for a single topographic map is under 4 seconds; thus the 128 maps of the usual EP sequence can be created from the initial 20 EP curves in approximately 9 minutes. EEG technicians have been trained to create BEAM images within a day's instruction. Furthermore, the prototype computed video interface costs little more than an eight-channel polygraph. Therefore, only a small investment would be required for a laboratory already performing EPs or collecting spectral data to begin to display their results topographically. The results of the current feasibility study suggest that topographic display of electrophysiological data may greatly expand their clinical utility, especially as regards functional lesions. In this sense, data from BEAM appear to complement the anatomical definition provided by the C T scan.
References 1. Allison T, Matsumiya Y ,Goff GD, et al: The scalp topography of human visual evoked potentials. Electroencephalogr Clin Neurophysiol 42:185-197, 1977 2. Benton AL: Developmental dyslexia: neurological aspects. Adv Neurol 7:l-74, 1975 3 . Bickford RG: New trends in clinical electroencephalography. Practitioner 217: 100- 107, 1976 4. Bickford RG, Billinger TW, Fleming NI, et al: The compressed spectral array ( C S A t a pictorial EEG, in Proceedings of the San Diego Biomedical Symposium, Feb 2-4, 1972. La Jolla, CA, Simulation Councils, 1972, vol 11, pp 365-470 5. Bickford RG, Brimmer J , Berger L, et al: Application of compressed spectral array in clinical EEG, in Kellaway P, Petersen J (eds): Automation of Clinical Electroencephalography. New Yark, Raven, 1973, pp 55-64 6. Davis H : Principles of electric response audiometry. Ann Otol Rhino1 Laryngol 85:Suppl 28:l-96, 1976 7. Denckla MB: Minimal brain dysfunction and dyslexia: beyond
320
Annals of Neurology
Vol 5
No 4
April 1979
diagnosis by exclusion, in Blau ME, Rapin I, Kinsbourne M (eds): Child Neurology. New York, Spectrum, 1977 8. Denckla MB, Rude1 RG: Naming of pictured objects by dyslexic and other learning-disabled children. Brain Language 3:1-15, 1976 9. Duchowny MS. Weiss IP, Majlessi H. et al: Visual evoked responses in childhood cortical blindness after head trauma and meningitis. Neurology (Minneap) 24933-940, 1974 10. Duffy FH, Rengstorff RH: Ametropia measurements from the visual evoked response. Am J Optom 48:717-728, 1971 11. Estrin T, Uzgalis R: Computer display of spatio-temporal EEG patterns. IEEE Trans Biomed Eng 16:192-196, 1969 12. Galen D, Ornstein R: Lateral specialization of cognitive mode: an EEG study. Psychophysiology 9:412-418. 1972 13. Geschwind N: The anatomy of acquired disorders of reading, in Money J (ed): Reading Disability: Progress and Research Needs in Dyslexia. Baltimore, Johns Hopkins University Press, 1962, pp 115-129 14. Goff G D , Matsumiya Y ,Allison T, et al: The scalp topography of human somatosensory and auditory evoked potentials. Electroencephalogr Clin Neurophyriol 42:57-76, 1977 15. Goff WR, Matsumiya Y ,Allison T, et al: Cross-modality comparisons of averaged evoked potentials, in Douchin E, Lindsley D B (eds): Average Evoked Potentials: Methods, Results, and Evaluations (NASA SP-191). Washington, DC, US Government Printing Office, 1969, pp 95-141 16. Gotman J, Gloor P, Ray WF: A quantitative comparison of traditional reading of the EEG and interpretation of computer-extracted features in patients with supratentorial brain lesions. Electroencephalogr Clin Neurophysiol 38: 623-639, 1975 17. Hanley J, Sklar B: Electroencephalographic correlates of developmental reading dyslexias: computer analyses of recordings from normal and dyslexic children, in Leisman G (ed): Basic Visual Processes and Learning Disability. Springfield, IL, Thomas, 1976, p 221 18. Harris JA, Melry GM, Bickford RG: Computer-controlled multidimensional display device for investigation and modeling of physiologic systems. Comp Biomed Res 2:519-535, 1969 19. Hughes JR, Park GG: Electro-clinical correlation in dyslexic children. Electroencephalogr Clin Neurophysiol26:117-121, 1969 20. Kokmen E, Billman JR, Abell MR: Lymphomatoid granulomatosis clinically confined to the CNS. Arch Neurol 34:782-784, 1977 21. McKee G, Humphrey B, McAdam OW: Scaled lateralization of alpha activity during linguistic and musical tasks. Psychophysiology 10:44 1-443, 1973 22. Melner BA, Regan D, Heron JR: Differential diagnosis of multiple sclerosis by visual evdted potential recording. Brain 97:755-772, 1974 23. Moore WH, Iang MK: Alpha asymmetry over the right and left hemispheres of stutterers and control subjects preceding massed oral readings: a preliminary investigation. Percept Mot Skills 44:223-230, 1977 24. Pena CE: Lymphomatoid granulomatosis with cerebral involvement. Acta Neuropathol (Bed) 37:193-197, 1977 25. Petsche H : Topography of the EEG: survey and prospects. Clin Neurol Neurosurg 7915-28, 1976 26. Pontius AA: Neurological aspects in some types of delinquency, especially among juveniles. Adolescence 7:289-308. 1972 27. Pontius AA, Ruttiger KF: Frontal lobe system maturational lag in juvenile delinquents shown in narrative test. Adolescence 11:509-518, 1976
28. Regan D, Melner BA, Heron JR: Delayed visual perception and delayed evoked potentials in spinal form of multiple sclerosis and in retrobulbar neuritis. Brain 97:43-66. 1976 29. Rimond A: Orientations et tendences des mtthodes topographiques dans i'Ptude de I'activitC tlectrique du cerveau. Rev Neurol (Paris) 93:397-410, 1755 30. Rtmond A. Gsevre N, Joseph JP, et al: The alpha average: I. Methodology and description. Electroencephalogr Clin Neurophysiol 26:245-285, 1969 31. Shipton HW: A new electrotoposcope using a helical scan. Proc Electrophysiol Techno1 Assn 2:2-11, 1956 32. Shipton HW: A new frequency-selective toposcope for electroencephalography. Med Electron Biol Eng 1:483-475, 1763 33. Sobotka KR,May JG: Visual evoked potentials and reaction time in normal and dyslexic children. Psychophysiology 14:18-24, 1777
34. Sokol S. Bloom B: Visually evokcd cortical responses of amblyopia to a spatially alternating stimulus. Invest Ophthalmol 12:736-939, 1773 35. Starr A, Achar J: Auditory brain stem responses in neurological disease. Arch Ncurol 32:761-768, 1975 36. Symann-Louert N , Gascon GG, Marsumiya Y ,et al: Wave form difference in visual evoked responses between normal and reading disabled children. Neurology (Minneap) 27: 156- 159, 197 7 37. Torres R,Ayers F W :Evaluation of the electroencephalogram of dyslexic children. Electroencephalogr Clin Neurophysiol 24:281-294, 1968 38. Vaughn HG Jr, Ritter W: The sources of auditory evoked responses recorded from the human scalp. Electroencephalogr Clin Neurophysiol 28:360-378, 1970 37. Walter WG, Shipton HW: A new toposcopic display system. Electroencephalogr Clin Neurophysiol 3:281-292, 195 1
Duffy et al: Brain Electrical Activity Mapping
321
Brain Electrical Activity Mapping (BEAM): A Method for Extending the Clinical Utility of EEG and Evoked Potential Data Frank H. Duffy, MD, James I. Burchfiel, PhD, and Cesare T. Lombroso, MD
The difficulties inherent in extracting clinically useful information by visual inspection alone from the massive amounts of data contained in multichannel polygraphic recordings have placed limits on the accuracy and range of utility of electroencephalography and evoked potentials. A method for condensing and summarizing the spatiotemporal information contained in recordings from multiple scalp electrodes is described. Data dimensionality is reduced and visibility increased by computer-controlled topographic mapping and display of data as color television images. Examples are given in which such brain electrical activity mapping (BEAM) (1) localizes tumors in patients with normal or nondiagnostic EEGs, ( 2 ) adds additional information to that visible on computerized axial tomography, and (3) demonstrates electrophysiological abnormalities in patients with functional lesions but normal CT scans. A sensitivity to the functional component of a neurological lesion suggests that BEAM may provide complementary information to the anatomical definition provided by the CT scan. Duffy FH, Burchfiel JL, Lombroso C T Brain electrical activity mapping (BEAM):a method for extending the clinical utility of EEG and evoked potential data. Ann Neurol 5:309-321,1979
Computerized axial tomography-the C T scan-has largely replaced cerebral angiography and pneumoencephalography for the diagnosis of anatomical abnormalities of the brain. The impact of the CT scan upon electroencephalography has been more variable. This stems from the use of EEG to diagnose not only anatomical lesions (e.g., tumor, abscess) but functional lesions as well (e.g., epilepsy). Whereas EEG may fail to diagnose 15% of brain tumors [3], it remains the best or only test for epilepsy and certain other functional disorders (e.g., Reye syndrome, brain death). The clinical utilization of evoked potentials (EP) to assist in the diagnosis of various sensory disorders [6, 9, 10, 341, brainstem lesions [351, and demyelinating diseases [22, 281 has increased. Nevertheless, the clinical value of EEG and EP in the diagnosis of certain forms of purely functional lesions is still limited. For instance, there are many reports of EEG and EP abnormalities in children with dyslexia 117, 19, 33, 36, 371; however, these findings are too nonspecific to form useful diagnostic constellations. We do not believe that EEG and EP suffer from an
~
inherent insensitivity to underlying brain dysfunction. O n the contrary, we propose that brain electrical activity presents not too little, but too much information to be easily grasped and assimilated by visual inspection alone. The approach we have taken t o enhance its utility in the diagnosis of both anatomical and functional lesions is based upon topographic mapping. In this technique, EEG and EP data recorded from multiple scalp electrodes are graphically displayed o n a computer-driven color video screen. Values between electrodes are obtained by interpolation. We call this technique brain electrical activity mapping, or BEAM. This report describes the methods and results of a feasibility study illustrating the ability of BEAM to define both anatomical and functional lesions in selected cases.
Method Data from 24 standard EEG scalp electrodes are amplified via a 20-channel polygraph (Grass Model 78) and recorded
on a 28-channel FM analogue tape recorder (Honeywell 5600E) for off-line computer processing (Digital Equip-
~~~~
From the Seizure Unit and Division of Neurophysiology, Department of Neurology, Children’s Hospital Medical Center and Harvard Medical School, Boston, MA.
Address reprint requests to Dr F. H. Duffy, Seizure Unit, Children’s Hospital Medical Center, 300 Longwood Ave, Bosron, MA 02115.
Accepted for publication Aug 21, 1978.
0 3 6 4 - ~ 1 3 4 / 7 ~ / 0 4 0 3 0 9 - 1 3 $ 0 1 .@ 2 5 1978 by Frank
H.Duffy 309
ment Corp PDP-12). Data analysis of three types is performed under the SIGSYS-12 biomedical software package (Agrippa Data Systems). First, spectral analysis from 0 to 32 Hz is performed o n background EEG using the fast Fourier transform (FFT) technique. To avoid artifacts from muscle and mains (60 Hz) activity, signals are tightly filtered at 24 d b per octave using active Butterworth filters (EEG Associates Mark 4 x 24). The amount of energy in each classic EEG band (e.g., delta, theta, alpha, and beta) is determined for each electrode and stored for future display. Second, visual EPs of 5 12 msec duration are formed from stimulation by randomly presented strobe flashes (Grass PS-2) set 20 cm from the subject's eyes at intensity level 2. Trials containing muscle or blink artifact are automatically excluded. The average EP voltage over 128 epochs each lasting 4 msec is determined for each electrode and stored for later display. Third, epochs of raw EEG signals containing epileptic spikes are digitized. Next, a 5to 20-msec epoch containing the feature of interest (spike) is averaged and stored for later display. By means of a hardware digital interface (Agrippa Data Systems 8ETV3) between the computer and an ordinary Sony color television set, results are displayed within a graphic outline of the head. Individual topographic displays of four sorts are commonly prepared: (1) spectral energy in any individual EEG band, (2) EP voltage at any 4-msec epoch after stimulus onset, ( 3 ) EEG amplitude during the peak of an epileptic spike, and ( 4 )results of statistical analysis on any of the preceding features (e.g., standard deviation). For each display, scalp areas around the original 24 electrode values are filled by linear three-dimensional interpolation based on the values at the three nearest electrodes (see Figure 1 for details of this process). Furthermore, the interface hardware is capable of displaying a new video image every 100 msec; accordingly, images may be cartooned at variable rates of up to 10 frames per second in an endless-loop manner. Thereby, for example, one can observe an EP in both space (a single topographic video image) and time (a cartooned series of images representing 512 msec in real time).
Subjects and Tests Twelve patients (7 with tumors, 1 with a suspected tumor, 3 with dyslexia, and 1 with a sociopathic personality) of the many referred for clinical EEGs and EPs were evaluated using the described techniques; the results of these evaluations are included in this report. These patients were carefully chosen to demonstrate the clinical utility of BEAM: they were selected to compare the ability of BEAM to define anatomical lesions (tumors) and functional lesions (dyslexia and behavioral disorders). Seven right-handed subjects with normal EEGs and normal neurological examinations were used as controls. The 12 patients were right-handed and had either normal or nondiagnostically abnormal EEGs containing no focal or lateralizing features. Seven ambulatory patients with unoperated supratentorial brain tumors diagnosed by C T scan were studied with spectral and visual EP topography. The tumors included gliomas of all four lobes of the right hemisphere and all but the frontal lobe of the left hemisphere. One patient, sus-
310 Annals of Neurology
Vol 5
No 4 April 1979
pected by C T scan of having a right frontal tumor, was similarly studied. This 13-year-old girl was admitted with a one and one-half year history of brief periods of abnormal C N S function. These periods included episodes of bizarre behavior, focal seizures of the right side of the face, transient left-sided hemianopia, and nominal aphasia. Due to a more recent history of headache and lethargy, a CT scan was performed which revealed areas of density in the right frontal lobe with shift of the midline structures from right to left (Fig 2). An angiogram confirmed a right frontal avascular mass. The patient was admitted for surgical exploration. Neuropsychological testing, however, revealed that in addition to nondominant frontal abnormalities, she showed bilateral parietal deficits. Three neurologically intact adolescents of normal intelligence, referred with the diagnosis of specific reading disability (dyslexia), were evaluated. In addition to resting spectral and visual EP studies, their spectral topographies were examined while they were listening to music or speech. O n e neurologically intact adolescent patient with a history of sociopathic behavior was studied with resting spectral and visual EP topography. His EEG was considered borderline due to excessive theta slowing in the anterior quadrants; however, his C T scan was within normal limits.
Results Controls Figure 3 demonstrates a BEAM spectral plot for a normal subject in the eyes-open and the eyes-closed state. For this subject and all the control subjects, spatial distributions were those one might predict using classic EEG analysis. For example, beta activity was maximal anteriorly (eyes open), and alpha activity tended to be maximal in the midline leads and posterior leads with a slight right-sided predominance (eyes closed). Figure 4 demonstrates five representative frames of a 128-frame visual evoked potential (VEP) examination in which each frame represents 4 msec. The symmetrical appearance of activity in the -occipital regions should be noted; the sequential appearance of symmetrical vertex waves is also illustrated. All normal subjects demonstrated spectral and VEP plots similar to those shown in Figures 3 and 4. Tumor Patients Figure 5 demonstrates the abnormal spectral plot for a patient with a left occipital glioma (his CT scan is shown in Figure 2 for comparison). Note the increased delta, decreased beta, and decreased alpha activity over the tumor site. Six of the 7 tumor patients showed the same abnormalities that this subject demonstrated, namely, increased slow activity and decreased fast activity overlying the tumor site. Although the seventh subject had a normal spectral plot, the spatial distribution of the standard deviation
P3 p4
F7 T4
73
4-
$
T5 01
A
B
C
D
F i g I . Example of the construction of a topographic map for E P data. Mean EPs are formed from each of 20 recordiq sites. Each EP is divided into 128 4 - m e c intervals, and the mean voltage value for each interval is calculated. In (A) the individual EPs are shown for the electrode locations indicated on the head diagram. In IB) the mean voltage values at these locations are shown for the interval beginning I 9 2 msec after the stimulus (the vertical line in A indicates this time on the EPs). Next the head region is treated as a 64 x 64 matrix; the resulting 4,096 spatial domains are illustrated in (C). Each domain is assigned a voltage value by linear interpolation from the three nearest known points. Finally, for display, the raw voltage values are fitted t o a discrete-level equal-interval intensity scale as shown in ( D ) .An illustration of topographic maps in two-color display for other selected intervals over the time course of the EP is shown in Figure 6.
Duffy et al: Brain Electrical Activity Mapping
311
Fig 2. Each photograph represents the resnlts of a computerized axial toniographic radiographic examination. ( A ) The primary lesion i n the patient u!ith the left occipital glionia shown in Figures 5 and G is marked with un X. Note the dilatation of the lefi lateral ventricle; this may explain the disruption of forzrjard spread of actitJitynoted in the VEP BEAM ( F i g 6 ) . ( B ) The lesion in the patient with lymphomatoid granuloniatosis is marked by an arrow. C T scan lesions were limited to the right frontal lobe. No abnormalities were seen i n the parietal regions to correlate with the biparietal abnormalities obserred by neurop.ryc-hologii-altesting and V E P BEAM (see F i g 8).
312
Annals of Neurology Vol 5
No 4
April 1979
F i g .3. Example of a normal BEAM spectral plot in the eyesopen (EO; A-D) and eyes-closed (EC; E-H) state, formed according to the method outlined in Figure 1 ; however, spectral fiinctions instead of EPs are used as input. Each spec~irmis dirfided u p into the classic E E G spectral bands (i.e.9 delta = 0 l o 4 H z , theta = 4 t o 8 H z , rrlphu = 8 t o 12 HL,und beta = 12 t o 20 Hz). Each plot is scaled i n d i d u a l I y . Note the ex-
pected asymmetry of alpha activity, which is greater over the right occipzrt for this right-handed subject. Note the general shift i n energy from the anterior to the posterior regionsfor all frequencies with EC, a frequentjnding in normal persons. Note the posterior beta asymmetry with EC; this i s seen when beta includes the 12 t o 16 H z range, adjacent t o alpha. I t is not seen fnr the 16 to 3 2 H z range (not shown here).
Duffy et
al:
Brain Electrical Activity Mapping 313
Fig 4
of his spectral energy (not illustrated) demonstrated localization to the site of his tumor. Figure 6 demonstrates a typical BEAM VEP plot for tumor subjects. All 7 tumor patients showed VEP plots demonstrating: (1) depression or absence of early electrical activity at the site of the tumor; (2) abnormally persistent late activity overlying the tumor; (3) asymmetrical spread of activity; and (4) diminished or absent vertex waves. In every patient, the topographically demonstrated abnormalities extended beyond the strict anatomical limits of the lesion on C T scan (for example, see the lesion size in Figure 6 as compared to the C T scan lesion in Figure 2). The ability of measurements taken from topographic plots to differentiate tumor patients from normal subjects is shown in Figure 7. The time spent in asymmetry at the locus of maximum asymmetry is plotted for all 7 controls and the 7 tumor patients. Note that a line drawn at 7 5 msec completely separates these two groups.
Patient with Suspected Tumor Figure 8 demonstrates the abnormal BEAM VEP plot for the 13-year-old girl with the suspected right frontal lobe tumor by CT scan and angiogram. This VEP BEAM showed repetitive midline parietal ab~
F i g 4. Example of a normal BEAM VEP plot formed according t o the method described in Figure 1. Five representative plots of the total of 128 plots spanning 512 msec are shown. Each plot represents a 4-msec averaged epoch. In this and all other VEP plots, red represents positive and blue represents negative activity with respect t o linked ear references. At 40 msec note the frontal electsoretinogram IERG). The primary EP wave begins with maximum amplitude overlying PZ, just forward of the occipital region 152 msec); this is a normal variant. This activity spreads to the “vertex”at CZ for the first vertex wave (1 08 msec). Note the slight vertex wave asymmetry, which is within normal limits. The next positive component,just beginning at 108 msec in the occipital region, sweeps up t o CZ t o form the next vertex wave (168 msec). Much kater, another vertex waveforms (440 msec). Normal VEPs consist of a series of vertex waves with positive or negative wavefronts converging on the vertex region. Asymmetries of early vertex waves are not uncommon. The ERG is a common contaminant (40 msec), as is some temporalis muscle activity (i.e., left ear at 52 msec). F i g 5. Example of a tumor (left occipital glioma) demonstrated & a BEAM spectral plot, formed as for Figure 3 in the EC state. Note the increased leji occipital delta, markedly decreased /eft occipital alpha, and slightly decreased lefi occipital beta activity. Theta shows only a slight increase on the left. Although alpha asymmetries are seen in normal subjects ( F i g 31, a marked depression of alpha activity overlying one occiput, as in thisjigure, is not. (See also Figure 6 for the VEP plot of this subject.)
normalities, possibly involving both medial parietal lobes, and bifrontal abnormalities. Thus, in contrast to the radiographic studies, BEAM demonstrated multifocal abnormalities that were more in line with the clinical history and with neuropsychological testing. At operation no tumor was found. The subsequent diagnosis by examination of biopsied material was lymphomatoid granulomatosis, a disease entity that may involve the CNS in a multifocal manner [20, 241. Dyslexics
None o f the dyslexic patients showed abnormalities on standard spectral plots. However, when they were studied under special circumstances, asymmetries were noted (not illustrated). When these subjects were asked to listen to music, alpha activity was suppressed over the right hemisphere as in many normal subjects [12, 21, 231. When they were asked to listen to speech, however, the left hemisphere alpha suppression often seen in normal subjects was not noted [12, 21, 231. The standard VEP plots of the dyslexics were normal. Unusual visual EP features in other special test situations will be the subject of a future report.
Patient with Behavioral Abnormality Although the spectral plot for the sociopathic patient was within normal limits, his visual EP topography showed consistent abnormalities consisting of recurring asymmetries of activity overlying the frontal lobes (Fig 9). Discussion Interpretation of EEG and EP records requires correlation of large volumes of data across both space and time. We propose that the difficulties involved in making spatiotemporal correlations by unaided visual inspection of multichannel polygraphic recordings constrain the clinical utility of EEG and EP. Accordingly, we have developed a technique for the topographic display of scalp-recorded signals on a color television monitor. Such topographic mapping reduces the dimensionality of data in a twofold manner. First, topographic mapping presents spatial information in a more concise and summarized form. For example, one must ordinarily examine many pages of EEG to appreciate subtle slow-wave asymmetries. With spectral BEAM plots, such asymmetries can be summarized in a single image (see Figs 3 , 5 ) . Second, linking together individual images, as in endless-loop cinematography (cartooning), enables the visualization of temporal patterns of brain activity that are not otherwise easily perceived. For example, EP analysis
Duffy et al: Brain Electrical Activity Mapping 315
Fig 6. BEAM VEP plot, showing 42 of the 128 frames generated from the 512-msec VEP of a patient with a kji occipital brain tumor, formed as in Figure 4. Note the initial negativity beginning at PZ (72 msec), which then sweeps forward primarily on the left. The first occipital positivity appears only on the right 188 to I12 msec). In a similar manner, the second occipital positivity appears asymmetrically on the right. Next, it sweeps simultaneously i n two directions: ( 1 ) forward on the right side exclusively, and (2) across the midline t o activate the left occipital region (I 52 to 21 6 msec). The kji occipital positivity persists despite disappearance of the right occipital positivity (I 60 t o 296 msec), reaching a maximum between 200 and 224 msec. The delayed activation of the &Jt occipital region overlying the tumor and the prolonged retention of activity i n this region, once activated, are typical of abnormalities associated with tumors. Note also the lack of forward spread of activity from the leji occipital region to the kji frontal region; we hypothesized that this might represent the electrical correlate of a functional disconnection between the k f t occipital and frontal lobes caused by the tumor (i.e., interrup-
tion of the left occipitofrontalpatbway). The late waves then
repeat the same sequence. Note the right occipital negativity developing at 344 msec, which then sweeps forward on the right only (424 to 472 msec); notice also that it grws in amplitude on the right (432 t o 448 msec) well before it spreads t o the kji occipital region 1456 to 504 msec). Again, the left side i s late and persistently activated (504 msec +). Thus the process offirst appearance in the right occipital region, spread forward on the right only, and late activation of the lefi occipital region from the right was repeated twice. It is of interest t o note that this patient had a completeb normal standard neurological examination, including visual field studies; the E E G was also normal. The chical impression was migraine headache. The significance of the peak of positivity appearing in the left frontal region ( I 5 2 to 224 msec) is not known (perhaps it represents the effect of deafferentationfrom the leji occipital region). Neuropsychological tests were not performed on this patient; therefore, one cannot comment about the possibility of a defect of frontal lobe function. (See also Figure 5 for the spectral plot of this subject.)
Duffy et al: Brain Electrical Activity Mapping
317
250
1
200
-
175
150
-
-
125
25 5o
1
0 NORMALS
TUMORS
F i g 7 . This graph demonstrates the possibility of differentiating patients with tumor from normal subjects on the basis of a single measurement from their topographic maps. All tumor patients showed marked asymmetries overlying the tumor in their VEP BEAM; most showed more than one episode of asymmetry at the same location lsee F i g 6). In this figure we plot the total length of time for all period of asymmetry at the locus of the maximal asymmetry for 7 normal subjects and 7 tumor patients. Note that the tumor patients had a much greater period of asymmetry than the normal persons. Furthermore, a line drawn at 75 msec serves t o separate all normal (< 75 msec)from all tumor subjects I> 130 msec).
usually consists of measuring amplitudes and latencies of various waveform components recorded from several electrodes. The temporal cartooning of EP BEAM images formed from multichannel data allows the visualization of traveling wavefronts of positive or negative electrical activity. Thus waveform components appearing as differing latencies at different scalp sites may, in fact, represent the same wavefront slowly progressing across the scalp surface (see Fig 6). BEAM does not create new data but merely presents the original spectral, EP, or EEG data in a more concise and intuitively comprehensible format. Despite the fact that their routine clinical EEGs were normal or nonlocalizing, all 7 tumor patients examined in this study were differentiated from normal subjects on the basis of a simple measurement obtained from their BEAM studies (see Fig 7). Although BEAM may not prove to be as accurate or consistent as the CT scan in defining anatomical lesions, its success rate exceeds that of the routine EEG. We suggest that BEAM may significantly increase the percentage of space-occupying lesions that can be electrophysiologically defined. In certain instances (see Fig 8) BEAM may provide more clinically useful information than is found on CT scan. The ability of BEAM to identify functional differences in the absence of anatomical lesions was also demonstrated. An EP BEAM test on an adolescent sociopath, chosen for having a normal CT scan, revealed prominent frontal lobe asymmetries (see Fig
318 Annals of Neurology Vol 5
N o 4 April 1979
9). Some authors have invoked frontal lobe dysfunction as the cause underlying abnormal behavior in sociopaths [26, 271. Functional changes were also observed in 3 adolescent dyslexics of normal intelligence and with normal EEGs. Although standard spectral and EP BEAM studies were within normal limits, alpha activity was not suppressed over the left hemisphere by listening to speech. Many investigators have implicated the left parietal lobe as the abnormal site in dyslexia on the basis of behavioral observation and testing and as an analogy to aphasia [2, 7, 8, 13, 331. The findings suggest that BEAM may provide a useful electrophysiological tool for studying the genesis of functional abnormalities of the brain. BEAM is not the first method to be used for the topographic representation of EEG or EP activity. In the 1950s and 1960s devices called toposcopes or encephaloscopes, developed by pioneers such as Harold Shipton and Grey Walter [31, 32, 391, used multiple-intensity modulated cathode ray tubes to display data in space and time. In recent years, others have topographically displayed computer-processed EEG or EP data via techniques varying from handdrawn contour plots [ l , 14, 15, 381 to computerdrawn contour or gray-scale plots [4, 5, 11, 16, 18, 25, 29, 301. The compressed spectral array of Bickford et al[4, 51 and the canonogram of Gotman, Gloor, and Ray Cl61 have proved to be clinically useful in the demonstration of localized brain lesions. Two of the most sophisticated display systems were implemented by Harris et al [ 181and Estrin and Uzgalis [l 11 as early as 1969. Both systems involved computer plotting of electrophysiological data. It is surprising that many of the more recent articles have used much simpler display systems, including handdrawn plots. In a 1976 review article, Petsche [25] F i g 8 . This figure illustrates how BEAM may add additional information to that provided by radiographic evaluation. Five selectedframes of 128 computed images are displayed as in Figure 4. The patient was a 13-year-old girl with a history of multiple episodes of transient neurological dysfunction and a more recent history of headaches and lethargy (see text). The C T scan lsee F i g 2) and angiogram suggested a right frontal lobe tumor. However, both neutvpsychological testing and thc VEP BEAM revealed bilateral parietal dysfunction as well as bifrontaldisease. The VEP BEAM, shown in this figure, reveals: (1) long-lasting bilateral prefrontal waves of both positive (1 64 msec) and negative 1292 msec) polarity, 12) a peculiar absence of activity (48, 116 , 164 msec) of the midparietal region, ( 3 ) an unusual period of polarity reversal between the left and right parietal regions and the midline parietal region (84 msec),and (4) a lesser deficiency of activation over the midfrontal region I1 64 and 292 msec). No tumor was found at the time of craniotomy. The neuropathologicaldiagnosis was lymphomatoid granulomatosis, a process that may involve the CNS in a multifocal manner.
F i x 9. This figure shou1.r a functional abnormality demonstrated by V E P topography in the presence of a normal C T scan (no anatomical lesion). Three selected frames of 128 computed images are displayed as in Figure 4. The patient was an 18-year-old man referred for evaluation with a hi.itory of sociopathic behavior. Although his C T .scan was within normal limits, his behavior suggested frontal lobe dysIysfirnction. His V E P plot. shown here, demonstrates recurring asymmetries in the frontal lobes of both positive and negatire activity. 7'hi.r. we believe, represents the neurophysiologicalcorrelate of-the expectedfrontal lobe beharioral dy.rfunction seen i n the presence of a normal C T scun. Such topographic maps may protme usefir1 i n finding electrnphysiologicalcorrelates of behacioral disorders.
Fig 8
suggested features to be incorporated into a topographic display system of the future. Many of these features, however, had already been implemented by Harris et a1 [18] in 1969. T h e reasons for the lack of application of technology already developed may be twofold. First, the early systems may have been operationally difficult, inconvenient, or expensive to use for other than a circumscribed research project. Second, the better systems appear to have been developed by engineers, and the failure to see these systems widely applied to clinical problems may have reflected a comprehension or credibility gap between engineers and physicians. The BEAM system described in this report takes full advantage of modern advances in solid-state electronics technology (see Scienti$c American, Sept 1977). A key feature of BEAM is a simple and easy to use interface between a small biomedical computer and a standard color television set. At present, computation time for a single topographic map is under 4 seconds; thus the 128 maps of the usual EP sequence can be created from the initial 20 EP curves in approximately 9 minutes. EEG technicians have been trained to create BEAM images within a day's instruction. Furthermore, the prototype computed video interface costs little more than an eight-channel polygraph. Therefore, only a small investment would be required for a laboratory already performing EPs or collecting spectral data to begin to display their results topographically. The results of the current feasibility study suggest that topographic display of electrophysiological data may greatly expand their clinical utility, especially as regards functional lesions. In this sense, data from BEAM appear to complement the anatomical definition provided by the C T scan.
References 1. Allison T, Matsumiya Y ,Goff GD, et al: The scalp topography of human visual evoked potentials. Electroencephalogr Clin Neurophysiol 42:185-197, 1977 2. Benton AL: Developmental dyslexia: neurological aspects. Adv Neurol 7:l-74, 1975 3 . Bickford RG: New trends in clinical electroencephalography. Practitioner 217: 100- 107, 1976 4. Bickford RG, Billinger TW, Fleming NI, et al: The compressed spectral array ( C S A t a pictorial EEG, in Proceedings of the San Diego Biomedical Symposium, Feb 2-4, 1972. La Jolla, CA, Simulation Councils, 1972, vol 11, pp 365-470 5. Bickford RG, Brimmer J , Berger L, et al: Application of compressed spectral array in clinical EEG, in Kellaway P, Petersen J (eds): Automation of Clinical Electroencephalography. New Yark, Raven, 1973, pp 55-64 6. Davis H : Principles of electric response audiometry. Ann Otol Rhino1 Laryngol 85:Suppl 28:l-96, 1976 7. Denckla MB: Minimal brain dysfunction and dyslexia: beyond
320
Annals of Neurology
Vol 5
No 4
April 1979
diagnosis by exclusion, in Blau ME, Rapin I, Kinsbourne M (eds): Child Neurology. New York, Spectrum, 1977 8. Denckla MB, Rude1 RG: Naming of pictured objects by dyslexic and other learning-disabled children. Brain Language 3:1-15, 1976 9. Duchowny MS. Weiss IP, Majlessi H. et al: Visual evoked responses in childhood cortical blindness after head trauma and meningitis. Neurology (Minneap) 24933-940, 1974 10. Duffy FH, Rengstorff RH: Ametropia measurements from the visual evoked response. Am J Optom 48:717-728, 1971 11. Estrin T, Uzgalis R: Computer display of spatio-temporal EEG patterns. IEEE Trans Biomed Eng 16:192-196, 1969 12. Galen D, Ornstein R: Lateral specialization of cognitive mode: an EEG study. Psychophysiology 9:412-418. 1972 13. Geschwind N: The anatomy of acquired disorders of reading, in Money J (ed): Reading Disability: Progress and Research Needs in Dyslexia. Baltimore, Johns Hopkins University Press, 1962, pp 115-129 14. Goff G D , Matsumiya Y ,Allison T, et al: The scalp topography of human somatosensory and auditory evoked potentials. Electroencephalogr Clin Neurophyriol 42:57-76, 1977 15. Goff WR, Matsumiya Y ,Allison T, et al: Cross-modality comparisons of averaged evoked potentials, in Douchin E, Lindsley D B (eds): Average Evoked Potentials: Methods, Results, and Evaluations (NASA SP-191). Washington, DC, US Government Printing Office, 1969, pp 95-141 16. Gotman J, Gloor P, Ray WF: A quantitative comparison of traditional reading of the EEG and interpretation of computer-extracted features in patients with supratentorial brain lesions. Electroencephalogr Clin Neurophysiol 38: 623-639, 1975 17. Hanley J, Sklar B: Electroencephalographic correlates of developmental reading dyslexias: computer analyses of recordings from normal and dyslexic children, in Leisman G (ed): Basic Visual Processes and Learning Disability. Springfield, IL, Thomas, 1976, p 221 18. Harris JA, Melry GM, Bickford RG: Computer-controlled multidimensional display device for investigation and modeling of physiologic systems. Comp Biomed Res 2:519-535, 1969 19. Hughes JR, Park GG: Electro-clinical correlation in dyslexic children. Electroencephalogr Clin Neurophysiol26:117-121, 1969 20. Kokmen E, Billman JR, Abell MR: Lymphomatoid granulomatosis clinically confined to the CNS. Arch Neurol 34:782-784, 1977 21. McKee G, Humphrey B, McAdam OW: Scaled lateralization of alpha activity during linguistic and musical tasks. Psychophysiology 10:44 1-443, 1973 22. Melner BA, Regan D, Heron JR: Differential diagnosis of multiple sclerosis by visual evdted potential recording. Brain 97:755-772, 1974 23. Moore WH, Iang MK: Alpha asymmetry over the right and left hemispheres of stutterers and control subjects preceding massed oral readings: a preliminary investigation. Percept Mot Skills 44:223-230, 1977 24. Pena CE: Lymphomatoid granulomatosis with cerebral involvement. Acta Neuropathol (Bed) 37:193-197, 1977 25. Petsche H : Topography of the EEG: survey and prospects. Clin Neurol Neurosurg 7915-28, 1976 26. Pontius AA: Neurological aspects in some types of delinquency, especially among juveniles. Adolescence 7:289-308. 1972 27. Pontius AA, Ruttiger KF: Frontal lobe system maturational lag in juvenile delinquents shown in narrative test. Adolescence 11:509-518, 1976
28. Regan D, Melner BA, Heron JR: Delayed visual perception and delayed evoked potentials in spinal form of multiple sclerosis and in retrobulbar neuritis. Brain 97:43-66. 1976 29. Rimond A: Orientations et tendences des mtthodes topographiques dans i'Ptude de I'activitC tlectrique du cerveau. Rev Neurol (Paris) 93:397-410, 1755 30. Rtmond A. Gsevre N, Joseph JP, et al: The alpha average: I. Methodology and description. Electroencephalogr Clin Neurophysiol 26:245-285, 1969 31. Shipton HW: A new electrotoposcope using a helical scan. Proc Electrophysiol Techno1 Assn 2:2-11, 1956 32. Shipton HW: A new frequency-selective toposcope for electroencephalography. Med Electron Biol Eng 1:483-475, 1763 33. Sobotka KR,May JG: Visual evoked potentials and reaction time in normal and dyslexic children. Psychophysiology 14:18-24, 1777
34. Sokol S. Bloom B: Visually evokcd cortical responses of amblyopia to a spatially alternating stimulus. Invest Ophthalmol 12:736-939, 1773 35. Starr A, Achar J: Auditory brain stem responses in neurological disease. Arch Ncurol 32:761-768, 1975 36. Symann-Louert N , Gascon GG, Marsumiya Y ,et al: Wave form difference in visual evoked responses between normal and reading disabled children. Neurology (Minneap) 27: 156- 159, 197 7 37. Torres R,Ayers F W :Evaluation of the electroencephalogram of dyslexic children. Electroencephalogr Clin Neurophysiol 24:281-294, 1968 38. Vaughn HG Jr, Ritter W: The sources of auditory evoked responses recorded from the human scalp. Electroencephalogr Clin Neurophysiol 28:360-378, 1970 37. Walter WG, Shipton HW: A new toposcopic display system. Electroencephalogr Clin Neurophysiol 3:281-292, 195 1
Duffy et al: Brain Electrical Activity Mapping
321
