IVied. & Biol. Eng. & Comput., 1977, 15,374-380
Automated visual evoked.responsesystem D. G. Childers Department of Electrical Engineering University of Florida Gainesville, Fla. 32611, USA
Abstract--An automated system capable of simultaneously monitoring, processing, and displaying 16 channels of human-scalp electrode-electroencephalographic (e,e.g.) data evoked by visual stimulation is described. 3-dimensional displays of the scalp spatio-temporal potential field are automatically generated. Other data-processing features include spectrum analysis various time-domain displays and phase measurements. The results of a study to distinguish" dyslexic from normal children are presented.
1 Introduction
IN MAN, a major medical and scientific goal is to sense remotely brain functioning without surgically insulting or causing discomfort to the patient or subject. The experiments in our laboratory stimulate the brain through one sensory modality; namely, vision. An electrode array placed on the scalp monitors simultaneously the topological characteristics of the r evoked by various visual stimuli. The subjects have either normal vision or a diagnosed visual dysfunction. We have previously demonstrated that the potential field patterns monitored by such an array of electrodes are affected by changes in the stimulus pattern projected on to the retina of the normal human subject (BOURNE et al., 1971). Our investigations were then extended to further test the hypothesis that we were sampling visual cortical functioning by enlarging our subject population to include subjects with visual dysfunctions (CHILDERS et aL, 1972). This latter subject group included amblyopia, which can be broadly defined as impaired acuity when an eye without apparent pathology is tested monocularly. The locus of amblyopia is usually attributed more to central processes (higher cortical centres) than to peripheral processes (retinal and primary visual pathways). This assumption is based largely on the elimination of peripheral possibilities rather than the isolation of central mechanisms. Another subject with visual dysfunction was an intermittent exotrope. This visual dysfunction is characterised by a deviation in which the relative position of the eyes is more divergent than the normal position for a given fixation distance. This condition can be overcome by fusion and usually results in the individual seeing a double image, unless the vision from one eye is suppressed. The suppression is not continuous as that presumed for First received 16th Februaryandin final form 9th April 1976
374
amblyopia, which also suppresses vision. Functionally then, the amblyope and intermittent exotrope differ. We demonstrated that these dysfunctions could be distinguished from each other and from the normal individual by evoking visual responses to retinal stimulation; these responses were monitored via an array of scalp electrodes (CHILDERS et ak, 1972). This demonstration was achieved primarily through the application of computer technology to reduce, organise and display massive amounts of data to assist the investigator to distinguish the normal from the abnormal subject and to aid in the establishment of cortical correlates of dysfunction. Our laboratory automation is a result of these previous studies but is applied here to the study of specific developmental dyslexia which has been defined by the World Federation of Neurology as a disease entity in which children fail to learn to read with normal proficiency despite conventional instruction, a culturally stimulating environment, adequate motivation, intact senses, average intelligence, and freedom from gross neurological and psychiatric problems. A symptom of this disease entity of children is reading retardation. As a result of investigative efforts, it is now hypothesised that the familial reading dysfunction in specific developmental dyslexia may be due to a specific neurological lesion, an inherited trait, a failure to develop cerebral dominance, a maturation lag or visual defect (Ross et aL, 1973; BENDER and SCHILDER, 1951; HERMANN and NORDIE, 1958). But these possible causes are diffuse and elusive. As might be expected, the clinical evaluation of each patient includes a personal history, physical and neurological examinations, educational examinations, intelligence quotient, reading examinations (Ross et al., 1973; CHILDERSet al., 1972). Recently, the investigation of the effect of various visual stimuli upon the electroencephalogram, as monitored by scalp electrodes, was added to the clinical
Medical and Biological Engineering & Computing
July 1977
evaluation procedure for a small select group of these dyslexic patients. A summary of the results of this evaluation is presented within the context of our laboratory automation procedures.
2 The need for laboratory automation Our needs for laboratory automation and a minicomputer interactive facility were pluralistic. Our major concern arose from our projection that we would be collecting and processing massive amounts of data; primarily from patients, but also from matched control subjects, i.e. subjects which match a particular patient in terms of selected parameters, e.g. intelligence quotient, age, sex, grade in school etc. Both the normal subjects and the patient populations would be admitted to the laboratory on a n appointment basis. Some would travel considerable distance across the state; this presented us with the requirement that a large amount of data would have to be processed rapidly before the patient was allowed to return home. It would be necessary to process automatically the data to reduce the number of human operator errors which might be introduced as a result of attempting to analyse the data rapidly under the pressure of a time limitation. If the operator could therefore be relieved of as many repetitive functions as possible, we felt the data would be more reliable. In addition, the laboratory would frequently be run by a laboratory assistant who would not be trained in esoteric computer operation. We therefore felt the need for an interactive system which would provide questions for the operator to answer. This would eliminate the dependence on the operator's memory to sequence properly the necessary operations to analyse the data. The team was interdisciplinary, which also meant that many would not be familiar with computer technology; thus we wanted as many of the software details of the system to be as invisible as possible. The system was to provide processed data output, not raw data output. All of these requirements, when incorporated into the system, made the operation of the system extremely easy for the experienced investigator as well. A major advantage of an automated-interactive system is that once the system is initialised the processing proceeds automatically with little or no operator supervision.
the scalp. Interelectrode resistance was of the order of 10 kO. Once the subject was seated in the screen room, the scalp-electrode leads were connected monopolarly (with the ear lobe as reference) to the amplifiers, which were solid-state and of our own design, basically a differential instrumentation circuit using two f.e.t, input operational amplifiers with an input impedance of 107 f~. The input differential f.e.t, operational amplifiers were followed by another low-gain operational amplifier. The overall gain was l0 s. The amplifier bandpass bandwidth was designed for 0.2 to 30 Hz with an attenuation rate of 40 dB/decade at either end. The system has the capability of providing various visual stimuli to the subject, including letters and words via a storage oscilloscope.* These letters can be presented monocularly or binocularly and frequently the letters used are b and d, because of the classic dyslexia-reversal tendency with these letters. The 16 data channels are digitised by an AX-08 laboratory peripheral of a PDP-8/I computer. While the rate of digitisation is variable, it is normally set at 100 samples per second per channel, which satisfies the sampling theorem with approximately 2 ~ r.m.s, error (CHILDERS, 1962). Computer averaging of the evoked responses is performed with software available from Digital Equipment Corp. as part of the Lab-8 averaging package. Typically 80 individual evoked responses to a like number of light flashes are averaged (PERRYand CH~LDERS, 1969). The frequency of visual stimulation is usually 2 Hz. The averaged data can be stored on magnetic tape or disc for further processing or display.
3 Experimental system
3.1 Hardware Prior to the subject entering the Faraday-shielded screen room, 16 electrodes were applied to the scalp in a prescribed fashion, dependent upon the protocol of the experiment to be conducted, usually as shown in Fig. 1. The electrodes were silver-silver chloride and were held in place by bands of masking tape around the subject's head. Beckman electrode paste was used at the interface between the electrode and Medical and Biological Engineering & Computing
Fig. 1 Typical rectangular electrode array. 20ram above the inion with a 20 turn spacing between adjacent electrodes " Tektronix 611
July 1977
375
The computer and peripheral equipment in the system is similar to that used in other laboratories (L]N and ACRAWAL, 1973; SCLABASSIand HARPER, 1973). 3.2 Software The heart of the automated interactive system is the software package for data collection and analysis. This package provides for the collection and real-time averaging of 16 data channels and the display of such data in several special formats9 ANTERIOR
25I r \ ....,:~"~ ~ " : ' v ,.... :."r
.... " ","':; "--~" ",-../"..':""
pv
,P'...a-. . ,'~"-~'....
i
". :~
.."
'-.~.. ,
.."
RIGHT
LEFT e. ", : k:
-.,
,".,
.~. ~, , 'k."
k."
/,.p
~',.'
:..
"..j
'...
e.
\. V
"v:
inion X a ANTERIOR .,-
,,'. ! %," '.:
/"..
,,.
,-'
~
:S"-.-
i....
LEFT
,".. , ,
RIGHT
,'. ,.
.
i'. ,'.
!
,,.,
.
.
:
,
, f/, ,.
\
,,,.
.,....,.,j,".j;.:
if' ,,..
,,.
X inion b Fig. 2 Visual evoked responses for a "b" stimulus, (a) normal subject (b) dyslexic subject. The 16 v.e.r.s are arranged according to Fig. 1. Positive deflection o f the active electrode is upward. The time duration o f each v.e.r, is 500 ms. A l l other parameters are as described in the text
376
The data-collection programs are modifications and extensions of the Lab-8 advanced averager program, which selects the trigger-input channel, provides for channel assignments, and establishes the parameters required for the averaging program. These parameters are entered by the operator as replies to prompting question by the computer program via the graphics terminal.* A channel number is assigned to each of the sixteen averages in accordance with the desired display configuration. Both the channel number and the number of averages to be performed by the computer are specified by the operator. The averaging parameters include the data record length (e.g. 500 ms), samples per record (e.g. 50) and summations per average (e.g. 80). As a result of these specifications, a completed average consists of 80 summations of a data record, 500 ms in duration, described by 50 sample points (sampling frequency of 100Hz). Fig. 2 shows typical average evoked-responses for the 16channel array shown in Fig. I. The data-collection parameters described above can be changed by the operator by calling the appropriate program from the system device (disc or tape). A minimum amount of information is required to be committed to memory, since the program asks the operator for the proper parameter specifications. Once an experimental protocol has been established, the data-collection parameters are, however, seldom changed until a new experimental protocol is designed. The averaging program asks for timing and trigger-level adjustments prior to averaging. During this phase of operation, the operator must first adjust the timing of the AX4)8 clock and adjust the level of the Schmidt trigger to obtain a sweep trace on the oscilloscope. When these adjustments have been made, the program is controlled through the operating system (or monitor) and the operator is free to call the next desired program. The averaging program performs certain bookkeeping tasks while averaging the data in real time. The bookkeeping information is acquired through the operator's response to computer generated enquiries concerning the subjects initials, the datatape number, the data experiment number and date of the experiment. The responses of these questions and the averaged data are then stored for future labelling and display in any of several possible formats. Usually, the investigator now desires a display of the data. Should the initial display not be properly scaled, the investigator may change the scale through keyboard action. Invisible to the investigator is a subprogram which calculates and removes the average value (mean) from each evoked response. The data is then moved to the proper location for access by the display package. 9 Tektronix 4010
Medical and Biological Engineering & Computing
July 1977
There are three display formats for the averaged visual evoked response (v.e.r.) data (PERRY and CHILDERS, 1969). The most fundamental display is that depicted in Fig. 2 which shows the averaged evoked responses for each channel as a function of time. The other two displays plot the averaged evoked potentials as a function of the array spatial co-ordinates for a fixed instant of time. In these displays, the values of the evoked potential for a selected time sample are plotted for each electrode. Linear interpolation is used to interconnect the points of the potential function. Thus, if v(t, x, y) is the potential as a function of time and spatial co-ordinates (electrode positions), then, for a selected to, v(to, x, y) takes on 16 different discrete values corresponding to the 16 different potentials measured at each electrode. These values are plotted as points above the (x, y) (electrode) plane and represent samples of the potential field monitored by the electrode array. These samples are then linearly interpolated and plotted to represent approximately the continuous potential field. A typical example is the 3-dimensional spatial plot of averaged data shown in Fig. 3 at 420 ms after the presentation of the visual stimulus. When viewed in temporal sequence, these plots show how the scalp potential field elicited by a visual stimulus fluctuates. An alternate format is to plot this same data in a basic contour format by using grey levels as the demarcations between quantised amplitude levels, Such a plot is shown in Fig. 4. The 3-dimensional plots of the data have been of particular interest to the interdisciplinary team investigating dyslexia. They require considerable computer time to generate; consequently, we automated this phase as follows: the basic program allows for either manual or automatic modes of operation; labels are included on the displays to provide subject and data identification.
INION
~
~':~'~G:"
Fig. 3 3-dimensional potential field plot of data monitored by electode array of Fig. 1
In an attempt to provide flexibility in data processing, two basic operation modes are provided to the operator. If an affirmative response is made to the query AUTO?, the display package enters the automatic mode and the displays are generated automatically in sequence at an increment specified Medical and Biological Engineerinq & Computing
by the operator. A negative reply places the display package in the manual mode and the operator must specify which display in the temporal sequence is to be generated. During the operation in the manual mode, the operator may select any display from 1 to 50 in any sequence and, with the exception of the 50th display, as often as desired. Once the 50th
Fig. 4 Contour (Gray-level) plot of potential field data monitored by electrode array of Fig. 1
display is generated in either mode, control is returned to the monitor. In the automatic mode, the last display to be generated will be the display whose number when added to the increment exceeds 50. Included on each display is a label, which contains subject identification information, the date of the data collection or experiment and the page number. This page or display number corresponds to the time sample point in milliseconds, i.e. page 1 is zero ms--page 2 is 10 ms--page 3 is 20 ms etc. The display package contains automatic-erase and optional automatic-copy capabilities which are independent of the operating mode, the automaticerase capability guarantees the screen will be erased prior to generating a display; when the optional automatic copy has also been selected, a copy will automatically be obtained for each display generated. If the copy option is not selected, no copy will be obtained and the program will halt after each display. The program will continue as per the operating mode once the continue key on the console has been depressed. Complete automatic operation is available by selecting the AUTO mode and the copy option. In this mode it is possible for the operator to select the data to be processed, start the display package and obtain the desired number of hard copies of the 3-dimensional potential field plots at the preselected sampling intervals without further intervention. (We have also filmed these displays directly off the screen to be used in a film for presentation at conferences.) During this phase of operation, the July 1977
377
operator is not needed and may undertake other tasks. Control will be returned to the monitor once the preselected data has been processed. The contour or grey-level plots can also be processed in the fully automatic mode. Other forms of data processing, which include spectrum analysis, phase measurements between selected electrode pairs and computation of bipolar-electrode temporal plots of the evoked-response data from the monocular temporal plots, are also available in the automatic or manual modes. Spectrum analysis is performed by a fast-Fouriertransform (f.f.t.) algorithm identical to the one in the IBM scientific subroutine package known as H a r m (COOLEY and TUKEY, 1965). Rectangular, Hanning or Tukey windowing is available. The latter window uses cosine smoothing on the first and last 10% of the data with unity weighting in between (BINGHAM et aL, 1967; CHILDERS and DURLING, 1975). The phase between selected electrode pairs is calculated from the arctangent of the ratio of the imaginary to the real part of the complex discrete Fourier coefficients. All data are originally measured monopolarly, i.e. the potential difference between the electrode and the reference ear electrode is monitored. Many investigators report bipolar potentials, i.e. the potential between scalp electrodes. We developed an algorithm to compute the bipolar potentials from the monopolar potentials for an arbitrary reference electrode so that our data could be compared with that from other laboratories. This requires that the reference ear electrode be truly indifferent, which may be difficult to achieve (PERRY and CHILDERS, 1969). Digital filtering of various forms can also be performed including notch, Butterworth, and Chebyshev. The latter two include lowpass, bandpass, or highpass (CHILDERSand DURLING,1975). Other specialised signal-processing algorithms include cepstrum analysis, complex demodulation, inverse filtering, and adaptive filtering (CmLDERS and DURLING, 1975; CHILDERS, 1974; HALPENY and CHILDERS, 1975).
4 Dyslexia study Dyslexia is a neurological symptom known to medicine for over half a century and manifests itself as an inability on the part of the individual to acquire reading and spelling skills commensurate with the individual's intelligence. The most commonly observed specific deficiencies are left-right confusion, word reversals, and letter confusions such as 'b' and 'd'. We wished to determine if an electrophysiological correlate of dyslexia could be established. The specific reason for attempting to add another test to aid in the diagnosis of dyslexia is that frequently 378
it is hard to distinguish the specific developmental dyslexic child from one with a language dysfunction, minimal brain damage, social-emotional problems, cultural-education deprivation, other disease entities, such as epilepsy, or a mixture of etiologies. The results of our early study appear in CHmDERS et aL (1972). Well over 50 patients have now been simultaneously tested by our group. The ages within the patient group vary widely. F o r the purposes of this case study, we shall discuss the results from 11 children. They are all males between the ages of 8 and 13. Four of these boys had no reading problem and were used as models for the normal visual-evoked response. They were also given a battery of tests to ensure that there were no undiagnosed problems. The seven remaining boys demonstrated reading problems. Six were diagnosed as being dyslexic by the interdisciplinary group and one was diagnosed as having a reading problem because of cultural and educational deprivation. These children were then divided into two study groups, test-1 and test-2. The v.e.r, data from seven of the children, the test-1 group, were processed, examined and used to establish characteristics which could be used to distinguish the dyslexic from the normal child. This group consisted of the following children: 4 normal, 2 dyslexic and 1 culturally and educationally deprived. The second, or test-2, group consisted of four children (all dyslexic) and was used to test our diagnostic criteria. The members of these groups were selected at random. Visual-evoked-response data were collected as follows: 13 individual records were analysed. Evoked responses to red and green stimuli as well as to such letters as 'b' and 'd' were collected; also included were control trials wherein no stimulus was presented to the subject. The various v.e.r, averaging parameters were the same as discussed previously. 4" 1 Summary o f results F o r each of the numerous data-analysis procedures, a criterion for diagnosis was developed, which, in the interests of brevity, we omit here but will gladly provide to any interested reader. This decision criterion was based on the examination of data from the test-1 group. The criteria were then applied to the four members of the test-2 group to determine the success of the test criteria. The degree of success was determined by calculating the average number of 'correct classifications, divided by the total number of subjects. All tests were administered for data obtained for the same stimulus which, for the sake of discussion here, is limited to the letters 'b' and 'd'. It should be emphasised that a classification is termed correct if it agrees with the diagnosis of the board of multidisciplinary examiners. There were a few cases when the diagnosis of the board was not unanimous, which was the
Medical and Biological Engineering & Computing
July 1977
primary reason for attempting to establish another test to aid the board in its diagnosis procedures. In some cases it was relatively easy to separate the normal from the dyslexic child, as shown in Fig. 2, which presents the averaged evoked responses for the letter 'b' for two different subjects for the same stimulus conditions. It was not difficult, under these conditions, to establish a criterion to separate the normal from the dyslexic child. Another illustrative example is a spectral
LL
0
50Hz
analysis shown in Fig. 5 for the same subjects and the same stimulus conditions. The spectra are arranged according to the corresponding electrode positions shown in Fig. l. The criteria used for the spectral data to classify the subject population had an average accuracy that ranged over 75 + 1 0 ~ for the stimulus letter b. Other stimuli, including the letter d, did not yield as good results. Similar criteria were developed for the other data-analysis procedures previously described. On the whole, criteria for spectral-analysis data were more consistent. 4" 2 Discussion
k_L. L_L_
LL:_._ LL_.
LLLL •
LLLL
_LLL x b
Fig, 5 Spectrum analysis of the (a) normal and (b) dyslexic data shown in Fig. 2, The spectrum was obtained by an f.f.t, algorithm with each data record being rectangularly windowed, Cosine tapered windowing did not significantly alter the spectrum. Zeros were added to the data record before transforming. The interval between d,f.t, coefficients is 1,56 Hz. The upper most frequency shown is 50 Hz. Only the first 33 d,ft, coefficients are displayed in a vector graph mode
Medical and Biological Engineering & Computing
The fact that spectral analysis was the m o s t consistent in separating the normal from the dyslexic child in this case study is not surprising in the light of SKLAR et al. (1973) who also found that a correct classification to the normal or the dyslexic group could be made from spectral analysis of the data. These investigators, however, used a different stimulus condition from the one discussed here. It has been known for some time that a certain percentage of dyslexics have abnormal e.e.g.s (ToRRES and AYRES, 1968) as distinct from the averaged evoked response described here. The dyslexic subjects who participated in our study all had complete e.e.g, evaluations; only one was found to have a 'mild abnormality', while the others were judged completely normal. Thus the proper classification of normals and dyslexics would be much more difficult and a far lower percentage of success would be achieved based on unaveraged e.e.g, analysis alone; i.e. the probability of successful classifications is increased by using an averaging procedure with evoked responses. 5 Conclusion As a result of automating the numerous steps in the data collection and analysis procedures described above, it is now possible for an investigator who did not participate in the original study to enter the laboratory, be given a few instructions on how to use the special programs and then proceed to examine any desired subset of data. If the investigator selects the automatic mode the data are displayed repetitively and hard copies are automatically generated. Alternately, the investigator may select the manual mode and skip about within the data records in search of unusual characteristics. We consider these to be extremely desirable system features and possibly ones which should be included in future data banks. One disadvantage we have noted as a result of automating the laboratory is that the professional investigative team as well as the technical support staff become resistant to changes necessitated as a result of wishing to ask new scientific questions.
July 1977
379
The tendency is to become infatuated with the data-processing capability and to slight the contributions that the data analyses are making to our understanding of a disease entity. Computer systems recently made available are capable of simultaneously running foreground and background programs. Here data collection could be performed under control of the foreground program, while simultaneously data processing is taking place in the background program. It is possible that the data could be displayed immediately upon completion of data collection in some situations. Such systems will contribute to a more complete automation of the future laboratory. Acknowledgment--The author acknowledges the many discussions with. John J. Ross on various aspects of this research. He is very appreciative of the excellent assistance of Frank Harwood in assembling the various hardware and software components of the system described. Daniel Maloney deserves special mention for his co-operation and discussions on various aspects of the analysis of the data from the dyslexic study. This research was supported in part by USPHS grant NS-HD 10097-01 from the National Institute of Neurological Diseases and Stroke, and by grant 74-2-9 from the Alfred P. Sloan Foundation. References BENDER, L. and SCHILDER, P. (1951) Graphic art as special ability in children with reading disability J. Clin. Exp. Psychopath. 12, 147-157. BOURNE, J. R., CHILDERS,D. G. and PERRY, N. W. Jun. (1971), Topological characteristics of the visual evoked response in man. Electroenceph. clin. Neurophysiol. 30, 423-436. BINGHAM, C., GODFREY,M. D. and TUKEY,J. W. (1967) Modern techniques of power spectrum estimation. IEEE Trans., AU-15, 56-66. CHILDERS,D. G. (1962) Study and experimental investigation on sampling rate and aliasing in time-division telemetry systems. Ibid. SET-8, 267-283.
CHILDERS,D. G., PERRY,N. W. Jun., HALPENY,O. S. and BOURNE, J. R. (1972) Spatio-temporal measures of cortical functioning in normal and abnormal vision. Comp. Biomed. Res. 5, 114-130. CH1LDERS, D. G., ROSS, J. J., PERRY, N. W. and NEViS, A. H. (1972) Dyslexia and cortical correlates, in Symposium on electroretinography (A. WroTH, Ed.) 35-51, Pacini, Pisa, Italy. CHILDERS,D. G. (1974) Decomposition of brain waves. J. Franklin Inst. 297, 103-125. CHILDERS, D. G. and DURLING, A. E. 0975) Digital filtering and signal processing. West Publishing Co., St. Paul, Minn. Appendix 4, pp. 433-440. COOLEY,J. and TUKEY, J. (1965) An algorithm for the machine calculation of complex Fourier series. Math. Comput. 9, 297-301. HALPENY, O. and CHILDERS, D. G. (1975) Composite wavefront decomposition via multidimensional digital filtering of array data. IEEE Trans., CAS-22, 552-563. HERMANN,K. and NORDIE, E. 0958) Is congenital word blindness a hereditary type of Gestmann's syndr0i~?. Psychiat. Neurol. 136, 50--73. LIN, W. C. and AGRAWAL,A. (1973) Minicomputer-based laboratory for speech-intelligibility research. Proc. IEEE, 61, 1583-1588. PERRY, N. W. Jun. and CHILDERS, D. G. (1969) The human visual evoked response: method and theory. Charles C. Thomas, Springfield, II1., Chap. 5. Ross, J. J., CHILDERS, D. G. and PERRY, N. W. Jun. (1973) The natural history and electrophysiological characteristics of familial language dysfunction. In The disabled learner (P. Satz and J. J. Ross, Eds.), Chap. 7, pp. 149-174, Rotterdam University Press. SCLABASSI,R. J. and HARPER,R. M. (1973) Laboratory computers in neurophysiology, Proc. IEEE, 61, 1602-1614. SKLAR, B., HANLEY, J. and SIMMONS, W. W. (1973) A computer analysis of LEG spectral signatures from normal and dyslexic children. 1EEL Trans., BME-20, 20-26. TORRES, F. and AYRES, F. W. (1968) Evaluation of the electroencephalogram of dyslexic children. Electroenceph, clin. Neurophysiol. 24, 287-297.
Syst~me automatis6 de rdponse t~ dvocation visuelle Sommaire--L'article d6erit un syst~me automatis6 capable en m~me temps de contr61er, traiter, et afficher 16 voles de donn6es 61ectro-enc6phalographiques fournies par des 61ectrodes implant6es dans le cl free humain, ces donn6es 6tant 6voqu6es par stimulation visuelle. Le syst~me produit automatiquement des affichages tridimensionnels du champ de potentiel spatio-temporal du crgme. Les autres caract6ristiques informatiques comprennent notamment l'analyse des spectres, les diff6rents affichages temps-domaine et les mesures de phase. L'article pr6sente aussi les r6sultats d'une 6tude cherchant h faire la distinction entre les enfants dyslexiques et les enfants normaux.
Ein automatisiertes, visuell aktiviertes Ansprechsystem Zusammenfassuag--Beschrieben wird ein automatisiertes System, das Kopfhaut-Elektroden-EEG-Daten, die durch visudle Simulierung aktiviert wurden, gleichzeitig kontrolliert, verarbeitet und auf 16 Kan~ilen darstellt. Die dreidimensionale Anzeige des Skalpraum-Temporal-Potentialfelds wird automatisch erzeugt. Zu anderen Datenverarbeitungsmerkmalen z~ihlen die Spektralanalyse, verschiedene Oberbereich-Anzeigen und Phasenmessungen. Auflerdem werden die Ergebnisse einer Studie zur Unterscheidung zwischen dyslexischen und normalen Kindern unterbreitet.
380
Medical and Biological Engineering & Computing
July 1977
Automated visual evoked.responsesystem D. G. Childers Department of Electrical Engineering University of Florida Gainesville, Fla. 32611, USA
Abstract--An automated system capable of simultaneously monitoring, processing, and displaying 16 channels of human-scalp electrode-electroencephalographic (e,e.g.) data evoked by visual stimulation is described. 3-dimensional displays of the scalp spatio-temporal potential field are automatically generated. Other data-processing features include spectrum analysis various time-domain displays and phase measurements. The results of a study to distinguish" dyslexic from normal children are presented.
1 Introduction
IN MAN, a major medical and scientific goal is to sense remotely brain functioning without surgically insulting or causing discomfort to the patient or subject. The experiments in our laboratory stimulate the brain through one sensory modality; namely, vision. An electrode array placed on the scalp monitors simultaneously the topological characteristics of the r evoked by various visual stimuli. The subjects have either normal vision or a diagnosed visual dysfunction. We have previously demonstrated that the potential field patterns monitored by such an array of electrodes are affected by changes in the stimulus pattern projected on to the retina of the normal human subject (BOURNE et al., 1971). Our investigations were then extended to further test the hypothesis that we were sampling visual cortical functioning by enlarging our subject population to include subjects with visual dysfunctions (CHILDERS et aL, 1972). This latter subject group included amblyopia, which can be broadly defined as impaired acuity when an eye without apparent pathology is tested monocularly. The locus of amblyopia is usually attributed more to central processes (higher cortical centres) than to peripheral processes (retinal and primary visual pathways). This assumption is based largely on the elimination of peripheral possibilities rather than the isolation of central mechanisms. Another subject with visual dysfunction was an intermittent exotrope. This visual dysfunction is characterised by a deviation in which the relative position of the eyes is more divergent than the normal position for a given fixation distance. This condition can be overcome by fusion and usually results in the individual seeing a double image, unless the vision from one eye is suppressed. The suppression is not continuous as that presumed for First received 16th Februaryandin final form 9th April 1976
374
amblyopia, which also suppresses vision. Functionally then, the amblyope and intermittent exotrope differ. We demonstrated that these dysfunctions could be distinguished from each other and from the normal individual by evoking visual responses to retinal stimulation; these responses were monitored via an array of scalp electrodes (CHILDERS et ak, 1972). This demonstration was achieved primarily through the application of computer technology to reduce, organise and display massive amounts of data to assist the investigator to distinguish the normal from the abnormal subject and to aid in the establishment of cortical correlates of dysfunction. Our laboratory automation is a result of these previous studies but is applied here to the study of specific developmental dyslexia which has been defined by the World Federation of Neurology as a disease entity in which children fail to learn to read with normal proficiency despite conventional instruction, a culturally stimulating environment, adequate motivation, intact senses, average intelligence, and freedom from gross neurological and psychiatric problems. A symptom of this disease entity of children is reading retardation. As a result of investigative efforts, it is now hypothesised that the familial reading dysfunction in specific developmental dyslexia may be due to a specific neurological lesion, an inherited trait, a failure to develop cerebral dominance, a maturation lag or visual defect (Ross et aL, 1973; BENDER and SCHILDER, 1951; HERMANN and NORDIE, 1958). But these possible causes are diffuse and elusive. As might be expected, the clinical evaluation of each patient includes a personal history, physical and neurological examinations, educational examinations, intelligence quotient, reading examinations (Ross et al., 1973; CHILDERSet al., 1972). Recently, the investigation of the effect of various visual stimuli upon the electroencephalogram, as monitored by scalp electrodes, was added to the clinical
Medical and Biological Engineering & Computing
July 1977
evaluation procedure for a small select group of these dyslexic patients. A summary of the results of this evaluation is presented within the context of our laboratory automation procedures.
2 The need for laboratory automation Our needs for laboratory automation and a minicomputer interactive facility were pluralistic. Our major concern arose from our projection that we would be collecting and processing massive amounts of data; primarily from patients, but also from matched control subjects, i.e. subjects which match a particular patient in terms of selected parameters, e.g. intelligence quotient, age, sex, grade in school etc. Both the normal subjects and the patient populations would be admitted to the laboratory on a n appointment basis. Some would travel considerable distance across the state; this presented us with the requirement that a large amount of data would have to be processed rapidly before the patient was allowed to return home. It would be necessary to process automatically the data to reduce the number of human operator errors which might be introduced as a result of attempting to analyse the data rapidly under the pressure of a time limitation. If the operator could therefore be relieved of as many repetitive functions as possible, we felt the data would be more reliable. In addition, the laboratory would frequently be run by a laboratory assistant who would not be trained in esoteric computer operation. We therefore felt the need for an interactive system which would provide questions for the operator to answer. This would eliminate the dependence on the operator's memory to sequence properly the necessary operations to analyse the data. The team was interdisciplinary, which also meant that many would not be familiar with computer technology; thus we wanted as many of the software details of the system to be as invisible as possible. The system was to provide processed data output, not raw data output. All of these requirements, when incorporated into the system, made the operation of the system extremely easy for the experienced investigator as well. A major advantage of an automated-interactive system is that once the system is initialised the processing proceeds automatically with little or no operator supervision.
the scalp. Interelectrode resistance was of the order of 10 kO. Once the subject was seated in the screen room, the scalp-electrode leads were connected monopolarly (with the ear lobe as reference) to the amplifiers, which were solid-state and of our own design, basically a differential instrumentation circuit using two f.e.t, input operational amplifiers with an input impedance of 107 f~. The input differential f.e.t, operational amplifiers were followed by another low-gain operational amplifier. The overall gain was l0 s. The amplifier bandpass bandwidth was designed for 0.2 to 30 Hz with an attenuation rate of 40 dB/decade at either end. The system has the capability of providing various visual stimuli to the subject, including letters and words via a storage oscilloscope.* These letters can be presented monocularly or binocularly and frequently the letters used are b and d, because of the classic dyslexia-reversal tendency with these letters. The 16 data channels are digitised by an AX-08 laboratory peripheral of a PDP-8/I computer. While the rate of digitisation is variable, it is normally set at 100 samples per second per channel, which satisfies the sampling theorem with approximately 2 ~ r.m.s, error (CHILDERS, 1962). Computer averaging of the evoked responses is performed with software available from Digital Equipment Corp. as part of the Lab-8 averaging package. Typically 80 individual evoked responses to a like number of light flashes are averaged (PERRYand CH~LDERS, 1969). The frequency of visual stimulation is usually 2 Hz. The averaged data can be stored on magnetic tape or disc for further processing or display.
3 Experimental system
3.1 Hardware Prior to the subject entering the Faraday-shielded screen room, 16 electrodes were applied to the scalp in a prescribed fashion, dependent upon the protocol of the experiment to be conducted, usually as shown in Fig. 1. The electrodes were silver-silver chloride and were held in place by bands of masking tape around the subject's head. Beckman electrode paste was used at the interface between the electrode and Medical and Biological Engineering & Computing
Fig. 1 Typical rectangular electrode array. 20ram above the inion with a 20 turn spacing between adjacent electrodes " Tektronix 611
July 1977
375
The computer and peripheral equipment in the system is similar to that used in other laboratories (L]N and ACRAWAL, 1973; SCLABASSIand HARPER, 1973). 3.2 Software The heart of the automated interactive system is the software package for data collection and analysis. This package provides for the collection and real-time averaging of 16 data channels and the display of such data in several special formats9 ANTERIOR
25I r \ ....,:~"~ ~ " : ' v ,.... :."r
.... " ","':; "--~" ",-../"..':""
pv
,P'...a-. . ,'~"-~'....
i
". :~
.."
'-.~.. ,
.."
RIGHT
LEFT e. ", : k:
-.,
,".,
.~. ~, , 'k."
k."
/,.p
~',.'
:..
"..j
'...
e.
\. V
"v:
inion X a ANTERIOR .,-
,,'. ! %," '.:
/"..
,,.
,-'
~
:S"-.-
i....
LEFT
,".. , ,
RIGHT
,'. ,.
.
i'. ,'.
!
,,.,
.
.
:
,
, f/, ,.
\
,,,.
.,....,.,j,".j;.:
if' ,,..
,,.
X inion b Fig. 2 Visual evoked responses for a "b" stimulus, (a) normal subject (b) dyslexic subject. The 16 v.e.r.s are arranged according to Fig. 1. Positive deflection o f the active electrode is upward. The time duration o f each v.e.r, is 500 ms. A l l other parameters are as described in the text
376
The data-collection programs are modifications and extensions of the Lab-8 advanced averager program, which selects the trigger-input channel, provides for channel assignments, and establishes the parameters required for the averaging program. These parameters are entered by the operator as replies to prompting question by the computer program via the graphics terminal.* A channel number is assigned to each of the sixteen averages in accordance with the desired display configuration. Both the channel number and the number of averages to be performed by the computer are specified by the operator. The averaging parameters include the data record length (e.g. 500 ms), samples per record (e.g. 50) and summations per average (e.g. 80). As a result of these specifications, a completed average consists of 80 summations of a data record, 500 ms in duration, described by 50 sample points (sampling frequency of 100Hz). Fig. 2 shows typical average evoked-responses for the 16channel array shown in Fig. I. The data-collection parameters described above can be changed by the operator by calling the appropriate program from the system device (disc or tape). A minimum amount of information is required to be committed to memory, since the program asks the operator for the proper parameter specifications. Once an experimental protocol has been established, the data-collection parameters are, however, seldom changed until a new experimental protocol is designed. The averaging program asks for timing and trigger-level adjustments prior to averaging. During this phase of operation, the operator must first adjust the timing of the AX4)8 clock and adjust the level of the Schmidt trigger to obtain a sweep trace on the oscilloscope. When these adjustments have been made, the program is controlled through the operating system (or monitor) and the operator is free to call the next desired program. The averaging program performs certain bookkeeping tasks while averaging the data in real time. The bookkeeping information is acquired through the operator's response to computer generated enquiries concerning the subjects initials, the datatape number, the data experiment number and date of the experiment. The responses of these questions and the averaged data are then stored for future labelling and display in any of several possible formats. Usually, the investigator now desires a display of the data. Should the initial display not be properly scaled, the investigator may change the scale through keyboard action. Invisible to the investigator is a subprogram which calculates and removes the average value (mean) from each evoked response. The data is then moved to the proper location for access by the display package. 9 Tektronix 4010
Medical and Biological Engineering & Computing
July 1977
There are three display formats for the averaged visual evoked response (v.e.r.) data (PERRY and CHILDERS, 1969). The most fundamental display is that depicted in Fig. 2 which shows the averaged evoked responses for each channel as a function of time. The other two displays plot the averaged evoked potentials as a function of the array spatial co-ordinates for a fixed instant of time. In these displays, the values of the evoked potential for a selected time sample are plotted for each electrode. Linear interpolation is used to interconnect the points of the potential function. Thus, if v(t, x, y) is the potential as a function of time and spatial co-ordinates (electrode positions), then, for a selected to, v(to, x, y) takes on 16 different discrete values corresponding to the 16 different potentials measured at each electrode. These values are plotted as points above the (x, y) (electrode) plane and represent samples of the potential field monitored by the electrode array. These samples are then linearly interpolated and plotted to represent approximately the continuous potential field. A typical example is the 3-dimensional spatial plot of averaged data shown in Fig. 3 at 420 ms after the presentation of the visual stimulus. When viewed in temporal sequence, these plots show how the scalp potential field elicited by a visual stimulus fluctuates. An alternate format is to plot this same data in a basic contour format by using grey levels as the demarcations between quantised amplitude levels, Such a plot is shown in Fig. 4. The 3-dimensional plots of the data have been of particular interest to the interdisciplinary team investigating dyslexia. They require considerable computer time to generate; consequently, we automated this phase as follows: the basic program allows for either manual or automatic modes of operation; labels are included on the displays to provide subject and data identification.
INION
~
~':~'~G:"
Fig. 3 3-dimensional potential field plot of data monitored by electode array of Fig. 1
In an attempt to provide flexibility in data processing, two basic operation modes are provided to the operator. If an affirmative response is made to the query AUTO?, the display package enters the automatic mode and the displays are generated automatically in sequence at an increment specified Medical and Biological Engineerinq & Computing
by the operator. A negative reply places the display package in the manual mode and the operator must specify which display in the temporal sequence is to be generated. During the operation in the manual mode, the operator may select any display from 1 to 50 in any sequence and, with the exception of the 50th display, as often as desired. Once the 50th
Fig. 4 Contour (Gray-level) plot of potential field data monitored by electrode array of Fig. 1
display is generated in either mode, control is returned to the monitor. In the automatic mode, the last display to be generated will be the display whose number when added to the increment exceeds 50. Included on each display is a label, which contains subject identification information, the date of the data collection or experiment and the page number. This page or display number corresponds to the time sample point in milliseconds, i.e. page 1 is zero ms--page 2 is 10 ms--page 3 is 20 ms etc. The display package contains automatic-erase and optional automatic-copy capabilities which are independent of the operating mode, the automaticerase capability guarantees the screen will be erased prior to generating a display; when the optional automatic copy has also been selected, a copy will automatically be obtained for each display generated. If the copy option is not selected, no copy will be obtained and the program will halt after each display. The program will continue as per the operating mode once the continue key on the console has been depressed. Complete automatic operation is available by selecting the AUTO mode and the copy option. In this mode it is possible for the operator to select the data to be processed, start the display package and obtain the desired number of hard copies of the 3-dimensional potential field plots at the preselected sampling intervals without further intervention. (We have also filmed these displays directly off the screen to be used in a film for presentation at conferences.) During this phase of operation, the July 1977
377
operator is not needed and may undertake other tasks. Control will be returned to the monitor once the preselected data has been processed. The contour or grey-level plots can also be processed in the fully automatic mode. Other forms of data processing, which include spectrum analysis, phase measurements between selected electrode pairs and computation of bipolar-electrode temporal plots of the evoked-response data from the monocular temporal plots, are also available in the automatic or manual modes. Spectrum analysis is performed by a fast-Fouriertransform (f.f.t.) algorithm identical to the one in the IBM scientific subroutine package known as H a r m (COOLEY and TUKEY, 1965). Rectangular, Hanning or Tukey windowing is available. The latter window uses cosine smoothing on the first and last 10% of the data with unity weighting in between (BINGHAM et aL, 1967; CHILDERS and DURLING, 1975). The phase between selected electrode pairs is calculated from the arctangent of the ratio of the imaginary to the real part of the complex discrete Fourier coefficients. All data are originally measured monopolarly, i.e. the potential difference between the electrode and the reference ear electrode is monitored. Many investigators report bipolar potentials, i.e. the potential between scalp electrodes. We developed an algorithm to compute the bipolar potentials from the monopolar potentials for an arbitrary reference electrode so that our data could be compared with that from other laboratories. This requires that the reference ear electrode be truly indifferent, which may be difficult to achieve (PERRY and CHILDERS, 1969). Digital filtering of various forms can also be performed including notch, Butterworth, and Chebyshev. The latter two include lowpass, bandpass, or highpass (CHILDERSand DURLING,1975). Other specialised signal-processing algorithms include cepstrum analysis, complex demodulation, inverse filtering, and adaptive filtering (CmLDERS and DURLING, 1975; CHILDERS, 1974; HALPENY and CHILDERS, 1975).
4 Dyslexia study Dyslexia is a neurological symptom known to medicine for over half a century and manifests itself as an inability on the part of the individual to acquire reading and spelling skills commensurate with the individual's intelligence. The most commonly observed specific deficiencies are left-right confusion, word reversals, and letter confusions such as 'b' and 'd'. We wished to determine if an electrophysiological correlate of dyslexia could be established. The specific reason for attempting to add another test to aid in the diagnosis of dyslexia is that frequently 378
it is hard to distinguish the specific developmental dyslexic child from one with a language dysfunction, minimal brain damage, social-emotional problems, cultural-education deprivation, other disease entities, such as epilepsy, or a mixture of etiologies. The results of our early study appear in CHmDERS et aL (1972). Well over 50 patients have now been simultaneously tested by our group. The ages within the patient group vary widely. F o r the purposes of this case study, we shall discuss the results from 11 children. They are all males between the ages of 8 and 13. Four of these boys had no reading problem and were used as models for the normal visual-evoked response. They were also given a battery of tests to ensure that there were no undiagnosed problems. The seven remaining boys demonstrated reading problems. Six were diagnosed as being dyslexic by the interdisciplinary group and one was diagnosed as having a reading problem because of cultural and educational deprivation. These children were then divided into two study groups, test-1 and test-2. The v.e.r, data from seven of the children, the test-1 group, were processed, examined and used to establish characteristics which could be used to distinguish the dyslexic from the normal child. This group consisted of the following children: 4 normal, 2 dyslexic and 1 culturally and educationally deprived. The second, or test-2, group consisted of four children (all dyslexic) and was used to test our diagnostic criteria. The members of these groups were selected at random. Visual-evoked-response data were collected as follows: 13 individual records were analysed. Evoked responses to red and green stimuli as well as to such letters as 'b' and 'd' were collected; also included were control trials wherein no stimulus was presented to the subject. The various v.e.r, averaging parameters were the same as discussed previously. 4" 1 Summary o f results F o r each of the numerous data-analysis procedures, a criterion for diagnosis was developed, which, in the interests of brevity, we omit here but will gladly provide to any interested reader. This decision criterion was based on the examination of data from the test-1 group. The criteria were then applied to the four members of the test-2 group to determine the success of the test criteria. The degree of success was determined by calculating the average number of 'correct classifications, divided by the total number of subjects. All tests were administered for data obtained for the same stimulus which, for the sake of discussion here, is limited to the letters 'b' and 'd'. It should be emphasised that a classification is termed correct if it agrees with the diagnosis of the board of multidisciplinary examiners. There were a few cases when the diagnosis of the board was not unanimous, which was the
Medical and Biological Engineering & Computing
July 1977
primary reason for attempting to establish another test to aid the board in its diagnosis procedures. In some cases it was relatively easy to separate the normal from the dyslexic child, as shown in Fig. 2, which presents the averaged evoked responses for the letter 'b' for two different subjects for the same stimulus conditions. It was not difficult, under these conditions, to establish a criterion to separate the normal from the dyslexic child. Another illustrative example is a spectral
LL
0
50Hz
analysis shown in Fig. 5 for the same subjects and the same stimulus conditions. The spectra are arranged according to the corresponding electrode positions shown in Fig. l. The criteria used for the spectral data to classify the subject population had an average accuracy that ranged over 75 + 1 0 ~ for the stimulus letter b. Other stimuli, including the letter d, did not yield as good results. Similar criteria were developed for the other data-analysis procedures previously described. On the whole, criteria for spectral-analysis data were more consistent. 4" 2 Discussion
k_L. L_L_
LL:_._ LL_.
LLLL •
LLLL
_LLL x b
Fig, 5 Spectrum analysis of the (a) normal and (b) dyslexic data shown in Fig. 2, The spectrum was obtained by an f.f.t, algorithm with each data record being rectangularly windowed, Cosine tapered windowing did not significantly alter the spectrum. Zeros were added to the data record before transforming. The interval between d,f.t, coefficients is 1,56 Hz. The upper most frequency shown is 50 Hz. Only the first 33 d,ft, coefficients are displayed in a vector graph mode
Medical and Biological Engineering & Computing
The fact that spectral analysis was the m o s t consistent in separating the normal from the dyslexic child in this case study is not surprising in the light of SKLAR et al. (1973) who also found that a correct classification to the normal or the dyslexic group could be made from spectral analysis of the data. These investigators, however, used a different stimulus condition from the one discussed here. It has been known for some time that a certain percentage of dyslexics have abnormal e.e.g.s (ToRRES and AYRES, 1968) as distinct from the averaged evoked response described here. The dyslexic subjects who participated in our study all had complete e.e.g, evaluations; only one was found to have a 'mild abnormality', while the others were judged completely normal. Thus the proper classification of normals and dyslexics would be much more difficult and a far lower percentage of success would be achieved based on unaveraged e.e.g, analysis alone; i.e. the probability of successful classifications is increased by using an averaging procedure with evoked responses. 5 Conclusion As a result of automating the numerous steps in the data collection and analysis procedures described above, it is now possible for an investigator who did not participate in the original study to enter the laboratory, be given a few instructions on how to use the special programs and then proceed to examine any desired subset of data. If the investigator selects the automatic mode the data are displayed repetitively and hard copies are automatically generated. Alternately, the investigator may select the manual mode and skip about within the data records in search of unusual characteristics. We consider these to be extremely desirable system features and possibly ones which should be included in future data banks. One disadvantage we have noted as a result of automating the laboratory is that the professional investigative team as well as the technical support staff become resistant to changes necessitated as a result of wishing to ask new scientific questions.
July 1977
379
The tendency is to become infatuated with the data-processing capability and to slight the contributions that the data analyses are making to our understanding of a disease entity. Computer systems recently made available are capable of simultaneously running foreground and background programs. Here data collection could be performed under control of the foreground program, while simultaneously data processing is taking place in the background program. It is possible that the data could be displayed immediately upon completion of data collection in some situations. Such systems will contribute to a more complete automation of the future laboratory. Acknowledgment--The author acknowledges the many discussions with. John J. Ross on various aspects of this research. He is very appreciative of the excellent assistance of Frank Harwood in assembling the various hardware and software components of the system described. Daniel Maloney deserves special mention for his co-operation and discussions on various aspects of the analysis of the data from the dyslexic study. This research was supported in part by USPHS grant NS-HD 10097-01 from the National Institute of Neurological Diseases and Stroke, and by grant 74-2-9 from the Alfred P. Sloan Foundation. References BENDER, L. and SCHILDER, P. (1951) Graphic art as special ability in children with reading disability J. Clin. Exp. Psychopath. 12, 147-157. BOURNE, J. R., CHILDERS,D. G. and PERRY, N. W. Jun. (1971), Topological characteristics of the visual evoked response in man. Electroenceph. clin. Neurophysiol. 30, 423-436. BINGHAM, C., GODFREY,M. D. and TUKEY,J. W. (1967) Modern techniques of power spectrum estimation. IEEE Trans., AU-15, 56-66. CHILDERS,D. G. (1962) Study and experimental investigation on sampling rate and aliasing in time-division telemetry systems. Ibid. SET-8, 267-283.
CHILDERS,D. G., PERRY,N. W. Jun., HALPENY,O. S. and BOURNE, J. R. (1972) Spatio-temporal measures of cortical functioning in normal and abnormal vision. Comp. Biomed. Res. 5, 114-130. CH1LDERS, D. G., ROSS, J. J., PERRY, N. W. and NEViS, A. H. (1972) Dyslexia and cortical correlates, in Symposium on electroretinography (A. WroTH, Ed.) 35-51, Pacini, Pisa, Italy. CHILDERS,D. G. (1974) Decomposition of brain waves. J. Franklin Inst. 297, 103-125. CHILDERS, D. G. and DURLING, A. E. 0975) Digital filtering and signal processing. West Publishing Co., St. Paul, Minn. Appendix 4, pp. 433-440. COOLEY,J. and TUKEY, J. (1965) An algorithm for the machine calculation of complex Fourier series. Math. Comput. 9, 297-301. HALPENY, O. and CHILDERS, D. G. (1975) Composite wavefront decomposition via multidimensional digital filtering of array data. IEEE Trans., CAS-22, 552-563. HERMANN,K. and NORDIE, E. 0958) Is congenital word blindness a hereditary type of Gestmann's syndr0i~?. Psychiat. Neurol. 136, 50--73. LIN, W. C. and AGRAWAL,A. (1973) Minicomputer-based laboratory for speech-intelligibility research. Proc. IEEE, 61, 1583-1588. PERRY, N. W. Jun. and CHILDERS, D. G. (1969) The human visual evoked response: method and theory. Charles C. Thomas, Springfield, II1., Chap. 5. Ross, J. J., CHILDERS, D. G. and PERRY, N. W. Jun. (1973) The natural history and electrophysiological characteristics of familial language dysfunction. In The disabled learner (P. Satz and J. J. Ross, Eds.), Chap. 7, pp. 149-174, Rotterdam University Press. SCLABASSI,R. J. and HARPER,R. M. (1973) Laboratory computers in neurophysiology, Proc. IEEE, 61, 1602-1614. SKLAR, B., HANLEY, J. and SIMMONS, W. W. (1973) A computer analysis of LEG spectral signatures from normal and dyslexic children. 1EEL Trans., BME-20, 20-26. TORRES, F. and AYRES, F. W. (1968) Evaluation of the electroencephalogram of dyslexic children. Electroenceph, clin. Neurophysiol. 24, 287-297.
Syst~me automatis6 de rdponse t~ dvocation visuelle Sommaire--L'article d6erit un syst~me automatis6 capable en m~me temps de contr61er, traiter, et afficher 16 voles de donn6es 61ectro-enc6phalographiques fournies par des 61ectrodes implant6es dans le cl free humain, ces donn6es 6tant 6voqu6es par stimulation visuelle. Le syst~me produit automatiquement des affichages tridimensionnels du champ de potentiel spatio-temporal du crgme. Les autres caract6ristiques informatiques comprennent notamment l'analyse des spectres, les diff6rents affichages temps-domaine et les mesures de phase. L'article pr6sente aussi les r6sultats d'une 6tude cherchant h faire la distinction entre les enfants dyslexiques et les enfants normaux.
Ein automatisiertes, visuell aktiviertes Ansprechsystem Zusammenfassuag--Beschrieben wird ein automatisiertes System, das Kopfhaut-Elektroden-EEG-Daten, die durch visudle Simulierung aktiviert wurden, gleichzeitig kontrolliert, verarbeitet und auf 16 Kan~ilen darstellt. Die dreidimensionale Anzeige des Skalpraum-Temporal-Potentialfelds wird automatisch erzeugt. Zu anderen Datenverarbeitungsmerkmalen z~ihlen die Spektralanalyse, verschiedene Oberbereich-Anzeigen und Phasenmessungen. Auflerdem werden die Ergebnisse einer Studie zur Unterscheidung zwischen dyslexischen und normalen Kindern unterbreitet.
380
Medical and Biological Engineering & Computing
July 1977
