SPECTRAL ANALYSIS OF THE ABNORMAL THETA RHYTHM IN THE EEG OF YOUNG CHILDREN R. Taistra, H. Gerken, H. Doose and /. Willebrand
Taistra, R., Gerken, H., Doose, H. and 'Willebrand, ].: Spectral analysis of the abnormal theta rhythm in the EEG of young children. Neuropa'diatrie 6: 363—376 (1975). The abnormal theta rhythm of the EEG of early childhood was investigated by spectral analysis. The pattern in the power spectrum is characterized by a peak in the thetaband. The relative peak intensity (peak ratio = ratio of peak intensity and average intensity of the spectrum within the frequency range of 0—10 Hz) as well as the right-left coherence proved to be good criteria for the description of the abnormal theta rhythms. When the ratio q J> 2.0 and the coherence Coh ^ 0.7 an abnormal theta rhythm can be considered to be present. When this definition is applied, a high correspendence between the mathematical and visual evaluation of the EEG is present. The abnormal theta rhythm must be regarded as a symptom of a functional anomaly which is — under electroencephalographic aspects — characterized by generators acting in abnormal synchronism and monofrequency. Theta rhythm
spectral analysis
Fourier analysis
The development of the Fast-FourierTransform algorithm by Cooley and Tukey (1965) and the segmentation method by Welch (1967) have led to important progress for the practice of automatic EEG-analysis (Dumermuth 1968, 1969, Dumermuth et al. 1972, Kiinkel 1972 a and b, Persson 1972). In connection with the rapid development of computer techniques, it has become possible to develop clinically applicable, simultaneous multichannel-on lineEEG-analysis (Kleiner et al. 1970, Kiin&e/1972aandb). It is the purpose of the present study to investigate the abnormal theta rhythm of the EEG of early childhood by means of spectral analysis and to define parameters which enable to deReceived: March 21, 1975
EEG
scribe the phenomenon as exactly as possible. Abnormal theta rhythms are defined by the following criteria (Doose 1964, Doose et al. 1966, 1967, 1972): Bilateral-synchronous, monomorphous, mostly monofrequent 4—7/s-rhythms occuring predominantly during early childhood. They usually show a pronounced parietal accentuation, but can also appear synchronously over all brain regions. They are not, or only incompletely, blocked by eyes opening; they occur in groups, sequences and sometimes continuously. Theta rhythms are more often found in boys than in girls. Doubtlessly they are genetically determined, probably by a multifactorial inheritance (Doose et al. 1972). They can be found in up to 15 °/o in brain healthy
Accepted: August 11, 1975
Address: H. D., Pediatric Clinic, University Kiel, Neuropediatric Department, Frobelstr. 15—17, D-2300 Kiel
363
Downloaded by: Universite Laval. Copyrighted material.
Neuropediatric Department of the Pediatric Clinic, University Kiel
children and are especially found in children suffering from seizures, predominantly in patients with primarily generalized epilepsies. The abnormal theta rhythm has therefore been interpreted as a bioelectrical symptom of an increased susceptibility to convulsions of the so-called centrencephalic type (Doose et al. 1966, 1967, 1972, Daute and Klust 1969). Material and methods 15 children (5 boys, 10 girls), aged 3-5 years, were examined electroencephalographically. The tracings were made in waking state with a Schwayzer%channel electroencephalograph according to a standardized scheme in monopolar leads with the right ear as reference electrode. Time constant was 0.1 sec, calibration 50 pV = 5 mm. Two pieces of 40 seconds of tracings with open respectively closed eyes were registered on analog tape (AEG-Telefunken MAS 55). The gauging was 100010 (50 pV = 100 Ole). Tracings disturbed by artefacts were replaced by undisturbed tracings. Visual evaluation Visual evaluation was performed as a blind study, i. e. without knowledge of the personal dates and of the results of the spectral analysis. The theta rhythms were classified according to intensity and location. 3 categories were distinguished: Positive (Fig. 4 and 5) = at least one group of typical theta rhythms within 20 seconds; questionably positive = less definite theta rhythms; negative = no theta rhythms (Fig. 1 and 2).
Vol. 6, No. 4, 1975
1
Ncurop5diatrie
Spectral analysis The analysis of the curves was performed with the biosignal processor "BIO 16" of AEG-Telefunken. The parameters were chosen as follows: 1. Number of segments 10, i. e. each segment has a length of 4 sec. 2. Rectangular data window. 3. Output of the power spectra on the plotter in linear scale. 4. Output of right-left-coherences. Parietal and occipital tracings were recorded with open as well as with closed eyes. This resulted in a total of 120 samples from 15 children, whose power-spectra were computed. Fig. 1- 6 are examples of rhythmic and non-rhythmic EEG records together with the appertaining power spectra. Evaluation and interpretation of the power spectra Frequency ranges with a relatively high intensity result in a peak in the power spectrum. The intensity within a certain frequency range is proportional to the corresponding area of the power spectrum. Accordingly the intensity of the abnormal theta rhythms could be determined by the peak area. The most important characteristic of abnormal theta rhythms is that they are condensed into a very small frequency range. Therefore, the peak height is regarded as a more suitable measure of intensity and this has been used in our investigations. The relative intensity can be expressed by the ratio (q) from the intensity of the peak (Ep) and the average intensity (E) (Fig. 7).
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
Fig. I
l o r n P., $, 5 years; Investigation during open eyes (time constant 0.1 s): Normal EEG
0 1 6 , No. 4, 9 7
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
I Neuropzdiatrie
Downloaded by: Universite Laval. Copyrighted material.
EEG theta rhythm
Fig. 3 Automatic EEG-analysis of the original records demonstrated in Fig. 1 and 2. Output of power spectra in linear scale of the parietal (par.) and occipital (occ.) leads during investigation with open ( and closed ( J- ) eyes
T)
367
Vol. 6, No. 4, 1975
1
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al. NeuropZdiatric
Fig. fi
theta rhythms
Thomas H., $, 3 years; Investigation during closed eyes (time constant 0.1 s): Parietal
Vol. 6, No. 4, 1975
1
Neuropldiatrie
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
Fig. 6 Automatic EEG-analysis of the original records demonstrated in Fig. 4 and 5. Output of the power spectra in linear scale of the parietal (par.) and occipital (occ.) leads during investigation with open ( .T ) and closed ( ) eyes
4
ft=OHz
fr'IOHz
Fig. 7 Method of evaluation of the ratio of the peak (see text)
A peak is regarded as significant if the relative intensity (q) has a value greater 2.0. This is to be motivated as follows: The computation of power spectra by means of the Fast Fourier Transform leads to confidence limits, depending on the number of degrees of freedom, which for 10 segments and a rectangular data window turns out to be about 20 (Welch 1967). Although the variable q includes estimates of Ep and which are statistically dependent, it
can be assumed to approximately follow a chi-squared distribution; for 20 degrees of freedom the lower limit of the 99010 twosided confidence interval is 0.5 (Jenkins and Watts 1968). This value corresponds to a ratio q = 2.0. The bilateral synchrony of the theta rhythm is best seen from the right-leftcoherence (Shaw and Ongley 1972). Considering a probability of 95 010 and 20 degrees of freedom a coherence (Coh) of at least 0.7 can be accepted as sta-
Downloaded by: Universite Laval. Copyrighted material.
EEG theta rhythm
Vol. 6, No. 4, 1975
tistically significant (Jenkinsand Watts 1968). The peaks were evaluated as follows: The area of the power spectrum within the range of 0-10 Hz was determined by planimetry. The average of three values from the same area was taken. The average intensity was obtained by dividing the resulting value by the factor 6.15 (i. e. the distance 0 -1 0 H z in cm). If double peaks occured within a range of k0.5 Hz, only the dominant peak was evaluated. The comparison of the calculated peaks with the result of the visual evaluation was performed as follows: An inaccuracy of f.0.5 H z was allowed to the visual determination of frequency. If by visual evaluation the frequency range could only be determined within certain limits, the peak which fell into this frequency range, was evaluated. If several peaks were found in such a range, only the highest peak was evaluated. Results Table I lists the number of peaks which lie within the frequency range 4.0-7.0 H z and fulfill the required conditions (q 2 2.0; Coh 2 0.7). Table I Number of peaks within the frequency range 4-7 Hz -
>
1
NeuropPdiatrie
A comparison of the results obtained with the spectral analysis method and by visual evaluation revealed the following: out of 56 peaks in the parietal leads, 37 were recognized in the visual evaluation of the original tracings. 10 records were evaluated as "questionably positive", 9 records as negative. Less favorable results were obtained in the occipital area: Only 5 out of 33 peaks were recognized visually, 14 were evaluated as "questionably positive", 14 as "negative". If the visually "questionably positive" and visually "positive" findings are considered together, the following result is obtained (not tabulated): Between the visual evaluation and the automatic analysis of 56 peaks (q 2 2.0; Coh 2 0.7) in the parietal region correspondence is obtained in 47 cases (83.9O/o). For 33 peaks in the occipital region, correspondence is only obtained in 19 cases (57.6 O/o). - It is noteworthy that 7 out of the 9 peaks in the parietal region, which were not recognized visually, have a ratio q of only 2.0 or 2.1, while all records evaluated as "positive" or "questionably positive" showed peaks with a ratio of a t least 2.2. Such an obvious distribution cannot be proved for the occipital tracings.
..
(q 2 2.0; Cwh 0.7) N = total number of peaks n' = visually positive n k visually questionably positive n3 = visually negative leads N n' n h 3 parietal 56 37 10 9 occipital 33 5 14 14
This comparison is repeated in table 11, where only peaks with a ratio of 2.2 or higher were taken into account: If the tracings considered visually as "positive" or "questionably positive" are taken together, 47 out of 49 peaks (q 2 2.2; Coh 2 0.7) were recognized in the parietal area (95.90/0). In the occi-
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
EEG theta rhythm
Discussion Spectral analysis of the EEG so far has rarely been applied to pediatric electroencephalography, especially not to the evaluation of basic rhythm variants, like e. g. the theta rhythm, which up to now has not much been taken into account in clinical electroencephalography (Doose et al. 1972, lit.). It was the purpose of the present study to investigate, whether and to what extent the abnormal theta rhythm in the EEG of early childhood can be evaluated by spectral analysis and which parameters describe the phenomenon in the best possible way.
Table I1 Number of peaks within the frequency range 4-7 (q 2 2.2; Coh N = total number of peaks =
,
2
Hz 0.7)
A
nz = visually questionably positive n3 = visually negative leads parietal occipital
N
n1
n2
n3
49 32
37
10
5
13
2 14
For the visual evaluation, the most significant characteristic of the abnormal theta rhythm is the monofrequency and the strict synchronism. The consequence is that the phenomenon should be best definable by mathematical analysis of these two characteristics. The monofrequency is determined by the relative intensity of the peak, i. e. by the quotient of the intensity of the peak and the average intensity of the spectrum; the synchronism of the rhythms is best assessed by the value of the right-leftcoherence. This leads to the working hypothesis that an abnormal theta rhythm must be considered as given, when a peak is present, whose intensity is statistically significant above the average intensity of the spectrum and which demonstrates an also statistically significant right-left-coherence. Considering the basic parameters of this special computer program, the definition "ratio q 2 2.0; Coh 2 0.7" accords with this requirement. In the EEG tracings of 15 children, which were partly normal and partly characterized by abnormal theta rhythms it was studied whether and to what extent interpretable results could
Downloaded by: Universite Laval. Copyrighted material.
pital area, however, only 18 out of 32 peaks (56.3 O/o) were recognized. Furthermore it was investigated, how tracks peak (q 2 2.0; Coh 2 0.7) were visually classified as "positive". As can be seen from table 111, only one track in the parietal area, which showed no ~ e a kin the automatic analysis, was classified visually as "questionably positive". In the occipital area, however, 25 tracks were visually considered as "questionable" or "definitely positive7' without showing a peak of the above definition in the machine analysis. Therefore, correspondence occured in 82.5 O/o and 32.8 O/o, respectively. Similar conditions result from the application of the peak definition "q 2 2.2; Coh 2 0.7" (table IV): A discrepancy is found only in one case for the parietal area, but in 26 cases for the occipital area. The frequency of a correspondence is therefore 94.0°/o and 3 1.o "10.
Taistra et al.
Table I I I -.-
Vol. 6 , No. 1, 1975
1
Neuropzdiatrie
Correspondence between visual evaluation of original records and results of automatic analysis considering peaks of the definition "q 2 2.0; Coh 2 0.7" - --
--
-.
- -
-
-
A = Peaks in automatic analysis V = abnormal rhythms in visual evaluation only N = number of tracings with a peak in the power spectrum and/or positive resp. questionably positive visual evaluation n = number of correspondences between automatic analysis and visual evaluation
CL
N
n
O/o
(95 Ole)
Parietal
56
1
57
47
82.5
70.191.3
Occipital
33
25
58
19
32.8
21.046.3
be obtained by applying this definition. The question was whether a sufficient correspondence between visual evaluation and mamine analysis with regard to the presence or absence of abnormal theta rhythms was present. The working hypothesis was confirmed by the comparison between the results of visual evaluation and automatical analysis, which was performed according to the blind principle. At least with regard to the parietal traTable IV
p
>
A = Peaks in automatic analysis V = abnormal rhythms in visual evaluation only N = number of tracings with a peak in the power spectrum and/or positive resp, questionably positive visual evaluation n = number of correspondences between automatic analysis and visual evaluation CL A V N n O/o (95 "10) p --
Parietal
49
1
50
47
94.0
Occipital
32
26
58
18
31.0
-
83.598.8 -
19.544.5
< 0.0005
Downloaded by: Universite Laval. Copyrighted material.
V
A
EEG theta rhythm
The results indicate that the abnormal theta rhythms in the EEG of young children can be very well registered and described by automatical analysis. The relative peak intensity and left-rightcoherence prove to be useful parameters for characterizing the phenomenon. Especially, the importance of significant coherence as an expression of an abnormal synchronism could be well demonstrated, because peaks with a high intensity could not be identified visually when a significant coherence was missing ( T a i s t ~ a1974). The present results could already be confirmed by a more extensive investigation. I t could be shown, that abnormal theta rhythms of the defined type occur significantly more frequently in children with febrile convulsions than in healthy control children (Taistva et al. 1975). With regard to previous studies (Doose 1964, Doose et al. 1965, 1966,1967,1972) the abnormal rhythms defined by quantitative analysis can thus be interpreted as an increased susceptibility to convulsions. An important factor in the development of this convulsibility might be a genetically determined functional cerebral anomaly, which is - under electroencephalographic aspects - charac-
terized by generators which are active in abnormal synchronism and monofrequency and are probably triggered by an abnormal central pacemaker.
Zusammenf assung Die abnorme Theta-Rhythmisierung des kindlichen EEGs wird spektralanalytisd untersucht. Das Muster kommt im Power-Spektrum als Peak im Thetaband zur Darstellung. Relative Peak-Intensitat (Peak-Quotient =QUOtient aus Intensitat des Peaks und mittlerer Intensitat des Spektrums innerhalb des Frequenzbereiches 0 -10 Hz) sowie Links-Rechts-Kohkenz erweisen sich als gute Kriterien fiir die Beschreibung der abnormen Theta-Rhythmisierung. Liegt ein Quotient q 2 2,O und eine Kohkenz von Coh > 0,7 vor, so ist eine abnorme Theta-Rhythmisierung des EEGs als gegeben anzusehen. Bei Anwendung dieser Definition besteht eine hohe Ubereinstimmung zwischen maschineller und visueller Auswertung des EEGs. - Die abnorme Theta-Rhythmisierung ist als Symptom einer funktionellen Anomalie aufzufassen, die unter elektroenzephalographischen Aspekten durch in abnormer Synchronie und Monofrequenz tatige Spannungsgeneratoren gekennzeichnet ist. References 1. Cooley, J. W., Tukey, J. W.: An algorithm for the machine calculation of complex Fourier Series. Math. Comput. 19: 297-301 (1965). 2. Daute, K. H., Klust, E.: Der abnorme ThetaRhythmus und seine Beziehung zu den Infektkrampfen. 2. Kinderhk. 105: 108-118 (1969). 3. Doose, H.: Das akinetishe Petit kal. I. und 11. Mitteilung. Arch. Psychiat. Nervenkr. 205: 625-654 (1964).
Downloaded by: Universite Laval. Copyrighted material.
clear whether these findings contain a true information on the phenomenon or whether they are due to the statistical inaccuracy because of the relatively short duration of the analysis. This problem was not investigated particularly within the present study, and for the first eliminated by neglecting the minor one of two peaks occuring within the range of 0.5 Hz.
4. Doose, H., Volzke, E., Scheffner, D.: Verlaufsformen kindlicher Epilepsien mit Spikewave-Absencen. Arch. Psychiat. Nervenkr. 207: 394-415 (1965). 5. Doose, H., Petersen, C. E., Volzke, E., Herzberger, E.: Fieberkrampfe und Epilepsie. Arch. Psychiat. Nervenkr. 208: 400432 (1966). 6. Doose, H., Gerken, H., Volzke, E.: Elektroenzephalographische Untersuchungen uber die Genetik zentrenzephaler Epilepsien. 2. Kinderhk. 101:242-257 (1967). 7. Doose, H., Gerken, H., Volzke, E.: On the genetics of EEG-anomalies in childhood. Neuro~adiatrie3: 386-401 (1972). 8. Dumermuth, G.: Variance Spectra of EEGs in twins. In: P. Kellaway and J. Petersen: Clin. Electroenceph. of children. Stockholm: Almquist and Wiksell1968, pag. 119-154. 9. Dumermuth, G.: Die Anwendung von Varianzspectra fur einen quantitativen Vergleich von EEG bei Zwillingen. Helv. paediatr. acta 24: 45-54 (1969). 10. Dumermuth, G., Waltz, W., Scollo-Lavizzari, G., Kleiner, B.: Spectral analysis of EEG activity in different sleep stages in normal adults. Europ. Neurol. 7: 265-296 (1972). 11. Jenkins, G. and Watts, D. G . : Spectral analysis and its applications. San Francisco: Holden Day 1968.
Vol. 6, No. 4, 1975
1
Neuropidiatrie
12. Kleiner, B., Fluhler, H., Huber, P. J., Dumermuth, G.: Spectrum analysis of the Electroencephalogram. Computer Programs in Biomedicine. 1 : 183-197 (1970). 13. Kiinkel, H.: Die Spektraldarstellung des EEG Z. EEG - EMG 3: 15-24 (1972a). 14. Kiinkel, H.: Simultane Vielkanal-on lineEEG Analyse in Echtzeit Z. EEG - EMG 3: 30-38 (1972 b). 15. Persson, J.: A computerized spectralanalysis of EEG from normal children during the first year of life. Technical report, School of Electrical Engineering, Chalmers University of Technology, GGteborg (1972). 16. Shaw, J. C. and Ongley, C.: The measurement of Synchronization. In: Petsche, H . and Brazier, M. A. B.: Synchronization of EEG Activity in Epilepsy. Springer Verlag Wien, New York, 204-215 (1972). 17. Taistra, R.: Spektralanalytische Untersuchungen uber die abnorme Theta-Rhythmisierung des kindlichen Elektroencephalogramms. Dissertation, Kiel 1974. 18. Taistra, R., Gerken, H., Doose, H.: EEGspectral analysis in children with febrile convulsions European Neurology, in press 1975. 19. Welch, P. D.: The use of Fast Fourier Transform for the estimation of Power Spectra. Transactions on Audio and Elektro-acoustics. Vol. Au - 15, 2: 70-73 (1967).
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
Taistra, R., Gerken, H., Doose, H. and 'Willebrand, ].: Spectral analysis of the abnormal theta rhythm in the EEG of young children. Neuropa'diatrie 6: 363—376 (1975). The abnormal theta rhythm of the EEG of early childhood was investigated by spectral analysis. The pattern in the power spectrum is characterized by a peak in the thetaband. The relative peak intensity (peak ratio = ratio of peak intensity and average intensity of the spectrum within the frequency range of 0—10 Hz) as well as the right-left coherence proved to be good criteria for the description of the abnormal theta rhythms. When the ratio q J> 2.0 and the coherence Coh ^ 0.7 an abnormal theta rhythm can be considered to be present. When this definition is applied, a high correspendence between the mathematical and visual evaluation of the EEG is present. The abnormal theta rhythm must be regarded as a symptom of a functional anomaly which is — under electroencephalographic aspects — characterized by generators acting in abnormal synchronism and monofrequency. Theta rhythm
spectral analysis
Fourier analysis
The development of the Fast-FourierTransform algorithm by Cooley and Tukey (1965) and the segmentation method by Welch (1967) have led to important progress for the practice of automatic EEG-analysis (Dumermuth 1968, 1969, Dumermuth et al. 1972, Kiinkel 1972 a and b, Persson 1972). In connection with the rapid development of computer techniques, it has become possible to develop clinically applicable, simultaneous multichannel-on lineEEG-analysis (Kleiner et al. 1970, Kiin&e/1972aandb). It is the purpose of the present study to investigate the abnormal theta rhythm of the EEG of early childhood by means of spectral analysis and to define parameters which enable to deReceived: March 21, 1975
EEG
scribe the phenomenon as exactly as possible. Abnormal theta rhythms are defined by the following criteria (Doose 1964, Doose et al. 1966, 1967, 1972): Bilateral-synchronous, monomorphous, mostly monofrequent 4—7/s-rhythms occuring predominantly during early childhood. They usually show a pronounced parietal accentuation, but can also appear synchronously over all brain regions. They are not, or only incompletely, blocked by eyes opening; they occur in groups, sequences and sometimes continuously. Theta rhythms are more often found in boys than in girls. Doubtlessly they are genetically determined, probably by a multifactorial inheritance (Doose et al. 1972). They can be found in up to 15 °/o in brain healthy
Accepted: August 11, 1975
Address: H. D., Pediatric Clinic, University Kiel, Neuropediatric Department, Frobelstr. 15—17, D-2300 Kiel
363
Downloaded by: Universite Laval. Copyrighted material.
Neuropediatric Department of the Pediatric Clinic, University Kiel
children and are especially found in children suffering from seizures, predominantly in patients with primarily generalized epilepsies. The abnormal theta rhythm has therefore been interpreted as a bioelectrical symptom of an increased susceptibility to convulsions of the so-called centrencephalic type (Doose et al. 1966, 1967, 1972, Daute and Klust 1969). Material and methods 15 children (5 boys, 10 girls), aged 3-5 years, were examined electroencephalographically. The tracings were made in waking state with a Schwayzer%channel electroencephalograph according to a standardized scheme in monopolar leads with the right ear as reference electrode. Time constant was 0.1 sec, calibration 50 pV = 5 mm. Two pieces of 40 seconds of tracings with open respectively closed eyes were registered on analog tape (AEG-Telefunken MAS 55). The gauging was 100010 (50 pV = 100 Ole). Tracings disturbed by artefacts were replaced by undisturbed tracings. Visual evaluation Visual evaluation was performed as a blind study, i. e. without knowledge of the personal dates and of the results of the spectral analysis. The theta rhythms were classified according to intensity and location. 3 categories were distinguished: Positive (Fig. 4 and 5) = at least one group of typical theta rhythms within 20 seconds; questionably positive = less definite theta rhythms; negative = no theta rhythms (Fig. 1 and 2).
Vol. 6, No. 4, 1975
1
Ncurop5diatrie
Spectral analysis The analysis of the curves was performed with the biosignal processor "BIO 16" of AEG-Telefunken. The parameters were chosen as follows: 1. Number of segments 10, i. e. each segment has a length of 4 sec. 2. Rectangular data window. 3. Output of the power spectra on the plotter in linear scale. 4. Output of right-left-coherences. Parietal and occipital tracings were recorded with open as well as with closed eyes. This resulted in a total of 120 samples from 15 children, whose power-spectra were computed. Fig. 1- 6 are examples of rhythmic and non-rhythmic EEG records together with the appertaining power spectra. Evaluation and interpretation of the power spectra Frequency ranges with a relatively high intensity result in a peak in the power spectrum. The intensity within a certain frequency range is proportional to the corresponding area of the power spectrum. Accordingly the intensity of the abnormal theta rhythms could be determined by the peak area. The most important characteristic of abnormal theta rhythms is that they are condensed into a very small frequency range. Therefore, the peak height is regarded as a more suitable measure of intensity and this has been used in our investigations. The relative intensity can be expressed by the ratio (q) from the intensity of the peak (Ep) and the average intensity (E) (Fig. 7).
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
Fig. I
l o r n P., $, 5 years; Investigation during open eyes (time constant 0.1 s): Normal EEG
0 1 6 , No. 4, 9 7
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
I Neuropzdiatrie
Downloaded by: Universite Laval. Copyrighted material.
EEG theta rhythm
Fig. 3 Automatic EEG-analysis of the original records demonstrated in Fig. 1 and 2. Output of power spectra in linear scale of the parietal (par.) and occipital (occ.) leads during investigation with open ( and closed ( J- ) eyes
T)
367
Vol. 6, No. 4, 1975
1
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al. NeuropZdiatric
Fig. fi
theta rhythms
Thomas H., $, 3 years; Investigation during closed eyes (time constant 0.1 s): Parietal
Vol. 6, No. 4, 1975
1
Neuropldiatrie
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
Fig. 6 Automatic EEG-analysis of the original records demonstrated in Fig. 4 and 5. Output of the power spectra in linear scale of the parietal (par.) and occipital (occ.) leads during investigation with open ( .T ) and closed ( ) eyes
4
ft=OHz
fr'IOHz
Fig. 7 Method of evaluation of the ratio of the peak (see text)
A peak is regarded as significant if the relative intensity (q) has a value greater 2.0. This is to be motivated as follows: The computation of power spectra by means of the Fast Fourier Transform leads to confidence limits, depending on the number of degrees of freedom, which for 10 segments and a rectangular data window turns out to be about 20 (Welch 1967). Although the variable q includes estimates of Ep and which are statistically dependent, it
can be assumed to approximately follow a chi-squared distribution; for 20 degrees of freedom the lower limit of the 99010 twosided confidence interval is 0.5 (Jenkins and Watts 1968). This value corresponds to a ratio q = 2.0. The bilateral synchrony of the theta rhythm is best seen from the right-leftcoherence (Shaw and Ongley 1972). Considering a probability of 95 010 and 20 degrees of freedom a coherence (Coh) of at least 0.7 can be accepted as sta-
Downloaded by: Universite Laval. Copyrighted material.
EEG theta rhythm
Vol. 6, No. 4, 1975
tistically significant (Jenkinsand Watts 1968). The peaks were evaluated as follows: The area of the power spectrum within the range of 0-10 Hz was determined by planimetry. The average of three values from the same area was taken. The average intensity was obtained by dividing the resulting value by the factor 6.15 (i. e. the distance 0 -1 0 H z in cm). If double peaks occured within a range of k0.5 Hz, only the dominant peak was evaluated. The comparison of the calculated peaks with the result of the visual evaluation was performed as follows: An inaccuracy of f.0.5 H z was allowed to the visual determination of frequency. If by visual evaluation the frequency range could only be determined within certain limits, the peak which fell into this frequency range, was evaluated. If several peaks were found in such a range, only the highest peak was evaluated. Results Table I lists the number of peaks which lie within the frequency range 4.0-7.0 H z and fulfill the required conditions (q 2 2.0; Coh 2 0.7). Table I Number of peaks within the frequency range 4-7 Hz -
>
1
NeuropPdiatrie
A comparison of the results obtained with the spectral analysis method and by visual evaluation revealed the following: out of 56 peaks in the parietal leads, 37 were recognized in the visual evaluation of the original tracings. 10 records were evaluated as "questionably positive", 9 records as negative. Less favorable results were obtained in the occipital area: Only 5 out of 33 peaks were recognized visually, 14 were evaluated as "questionably positive", 14 as "negative". If the visually "questionably positive" and visually "positive" findings are considered together, the following result is obtained (not tabulated): Between the visual evaluation and the automatic analysis of 56 peaks (q 2 2.0; Coh 2 0.7) in the parietal region correspondence is obtained in 47 cases (83.9O/o). For 33 peaks in the occipital region, correspondence is only obtained in 19 cases (57.6 O/o). - It is noteworthy that 7 out of the 9 peaks in the parietal region, which were not recognized visually, have a ratio q of only 2.0 or 2.1, while all records evaluated as "positive" or "questionably positive" showed peaks with a ratio of a t least 2.2. Such an obvious distribution cannot be proved for the occipital tracings.
..
(q 2 2.0; Cwh 0.7) N = total number of peaks n' = visually positive n k visually questionably positive n3 = visually negative leads N n' n h 3 parietal 56 37 10 9 occipital 33 5 14 14
This comparison is repeated in table 11, where only peaks with a ratio of 2.2 or higher were taken into account: If the tracings considered visually as "positive" or "questionably positive" are taken together, 47 out of 49 peaks (q 2 2.2; Coh 2 0.7) were recognized in the parietal area (95.90/0). In the occi-
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
EEG theta rhythm
Discussion Spectral analysis of the EEG so far has rarely been applied to pediatric electroencephalography, especially not to the evaluation of basic rhythm variants, like e. g. the theta rhythm, which up to now has not much been taken into account in clinical electroencephalography (Doose et al. 1972, lit.). It was the purpose of the present study to investigate, whether and to what extent the abnormal theta rhythm in the EEG of early childhood can be evaluated by spectral analysis and which parameters describe the phenomenon in the best possible way.
Table I1 Number of peaks within the frequency range 4-7 (q 2 2.2; Coh N = total number of peaks =
,
2
Hz 0.7)
A
nz = visually questionably positive n3 = visually negative leads parietal occipital
N
n1
n2
n3
49 32
37
10
5
13
2 14
For the visual evaluation, the most significant characteristic of the abnormal theta rhythm is the monofrequency and the strict synchronism. The consequence is that the phenomenon should be best definable by mathematical analysis of these two characteristics. The monofrequency is determined by the relative intensity of the peak, i. e. by the quotient of the intensity of the peak and the average intensity of the spectrum; the synchronism of the rhythms is best assessed by the value of the right-leftcoherence. This leads to the working hypothesis that an abnormal theta rhythm must be considered as given, when a peak is present, whose intensity is statistically significant above the average intensity of the spectrum and which demonstrates an also statistically significant right-left-coherence. Considering the basic parameters of this special computer program, the definition "ratio q 2 2.0; Coh 2 0.7" accords with this requirement. In the EEG tracings of 15 children, which were partly normal and partly characterized by abnormal theta rhythms it was studied whether and to what extent interpretable results could
Downloaded by: Universite Laval. Copyrighted material.
pital area, however, only 18 out of 32 peaks (56.3 O/o) were recognized. Furthermore it was investigated, how tracks peak (q 2 2.0; Coh 2 0.7) were visually classified as "positive". As can be seen from table 111, only one track in the parietal area, which showed no ~ e a kin the automatic analysis, was classified visually as "questionably positive". In the occipital area, however, 25 tracks were visually considered as "questionable" or "definitely positive7' without showing a peak of the above definition in the machine analysis. Therefore, correspondence occured in 82.5 O/o and 32.8 O/o, respectively. Similar conditions result from the application of the peak definition "q 2 2.2; Coh 2 0.7" (table IV): A discrepancy is found only in one case for the parietal area, but in 26 cases for the occipital area. The frequency of a correspondence is therefore 94.0°/o and 3 1.o "10.
Taistra et al.
Table I I I -.-
Vol. 6 , No. 1, 1975
1
Neuropzdiatrie
Correspondence between visual evaluation of original records and results of automatic analysis considering peaks of the definition "q 2 2.0; Coh 2 0.7" - --
--
-.
- -
-
-
A = Peaks in automatic analysis V = abnormal rhythms in visual evaluation only N = number of tracings with a peak in the power spectrum and/or positive resp. questionably positive visual evaluation n = number of correspondences between automatic analysis and visual evaluation
CL
N
n
O/o
(95 Ole)
Parietal
56
1
57
47
82.5
70.191.3
Occipital
33
25
58
19
32.8
21.046.3
be obtained by applying this definition. The question was whether a sufficient correspondence between visual evaluation and mamine analysis with regard to the presence or absence of abnormal theta rhythms was present. The working hypothesis was confirmed by the comparison between the results of visual evaluation and automatical analysis, which was performed according to the blind principle. At least with regard to the parietal traTable IV
p
>
A = Peaks in automatic analysis V = abnormal rhythms in visual evaluation only N = number of tracings with a peak in the power spectrum and/or positive resp, questionably positive visual evaluation n = number of correspondences between automatic analysis and visual evaluation CL A V N n O/o (95 "10) p --
Parietal
49
1
50
47
94.0
Occipital
32
26
58
18
31.0
-
83.598.8 -
19.544.5
< 0.0005
Downloaded by: Universite Laval. Copyrighted material.
V
A
EEG theta rhythm
The results indicate that the abnormal theta rhythms in the EEG of young children can be very well registered and described by automatical analysis. The relative peak intensity and left-rightcoherence prove to be useful parameters for characterizing the phenomenon. Especially, the importance of significant coherence as an expression of an abnormal synchronism could be well demonstrated, because peaks with a high intensity could not be identified visually when a significant coherence was missing ( T a i s t ~ a1974). The present results could already be confirmed by a more extensive investigation. I t could be shown, that abnormal theta rhythms of the defined type occur significantly more frequently in children with febrile convulsions than in healthy control children (Taistva et al. 1975). With regard to previous studies (Doose 1964, Doose et al. 1965, 1966,1967,1972) the abnormal rhythms defined by quantitative analysis can thus be interpreted as an increased susceptibility to convulsions. An important factor in the development of this convulsibility might be a genetically determined functional cerebral anomaly, which is - under electroencephalographic aspects - charac-
terized by generators which are active in abnormal synchronism and monofrequency and are probably triggered by an abnormal central pacemaker.
Zusammenf assung Die abnorme Theta-Rhythmisierung des kindlichen EEGs wird spektralanalytisd untersucht. Das Muster kommt im Power-Spektrum als Peak im Thetaband zur Darstellung. Relative Peak-Intensitat (Peak-Quotient =QUOtient aus Intensitat des Peaks und mittlerer Intensitat des Spektrums innerhalb des Frequenzbereiches 0 -10 Hz) sowie Links-Rechts-Kohkenz erweisen sich als gute Kriterien fiir die Beschreibung der abnormen Theta-Rhythmisierung. Liegt ein Quotient q 2 2,O und eine Kohkenz von Coh > 0,7 vor, so ist eine abnorme Theta-Rhythmisierung des EEGs als gegeben anzusehen. Bei Anwendung dieser Definition besteht eine hohe Ubereinstimmung zwischen maschineller und visueller Auswertung des EEGs. - Die abnorme Theta-Rhythmisierung ist als Symptom einer funktionellen Anomalie aufzufassen, die unter elektroenzephalographischen Aspekten durch in abnormer Synchronie und Monofrequenz tatige Spannungsgeneratoren gekennzeichnet ist. References 1. Cooley, J. W., Tukey, J. W.: An algorithm for the machine calculation of complex Fourier Series. Math. Comput. 19: 297-301 (1965). 2. Daute, K. H., Klust, E.: Der abnorme ThetaRhythmus und seine Beziehung zu den Infektkrampfen. 2. Kinderhk. 105: 108-118 (1969). 3. Doose, H.: Das akinetishe Petit kal. I. und 11. Mitteilung. Arch. Psychiat. Nervenkr. 205: 625-654 (1964).
Downloaded by: Universite Laval. Copyrighted material.
clear whether these findings contain a true information on the phenomenon or whether they are due to the statistical inaccuracy because of the relatively short duration of the analysis. This problem was not investigated particularly within the present study, and for the first eliminated by neglecting the minor one of two peaks occuring within the range of 0.5 Hz.
4. Doose, H., Volzke, E., Scheffner, D.: Verlaufsformen kindlicher Epilepsien mit Spikewave-Absencen. Arch. Psychiat. Nervenkr. 207: 394-415 (1965). 5. Doose, H., Petersen, C. E., Volzke, E., Herzberger, E.: Fieberkrampfe und Epilepsie. Arch. Psychiat. Nervenkr. 208: 400432 (1966). 6. Doose, H., Gerken, H., Volzke, E.: Elektroenzephalographische Untersuchungen uber die Genetik zentrenzephaler Epilepsien. 2. Kinderhk. 101:242-257 (1967). 7. Doose, H., Gerken, H., Volzke, E.: On the genetics of EEG-anomalies in childhood. Neuro~adiatrie3: 386-401 (1972). 8. Dumermuth, G.: Variance Spectra of EEGs in twins. In: P. Kellaway and J. Petersen: Clin. Electroenceph. of children. Stockholm: Almquist and Wiksell1968, pag. 119-154. 9. Dumermuth, G.: Die Anwendung von Varianzspectra fur einen quantitativen Vergleich von EEG bei Zwillingen. Helv. paediatr. acta 24: 45-54 (1969). 10. Dumermuth, G., Waltz, W., Scollo-Lavizzari, G., Kleiner, B.: Spectral analysis of EEG activity in different sleep stages in normal adults. Europ. Neurol. 7: 265-296 (1972). 11. Jenkins, G. and Watts, D. G . : Spectral analysis and its applications. San Francisco: Holden Day 1968.
Vol. 6, No. 4, 1975
1
Neuropidiatrie
12. Kleiner, B., Fluhler, H., Huber, P. J., Dumermuth, G.: Spectrum analysis of the Electroencephalogram. Computer Programs in Biomedicine. 1 : 183-197 (1970). 13. Kiinkel, H.: Die Spektraldarstellung des EEG Z. EEG - EMG 3: 15-24 (1972a). 14. Kiinkel, H.: Simultane Vielkanal-on lineEEG Analyse in Echtzeit Z. EEG - EMG 3: 30-38 (1972 b). 15. Persson, J.: A computerized spectralanalysis of EEG from normal children during the first year of life. Technical report, School of Electrical Engineering, Chalmers University of Technology, GGteborg (1972). 16. Shaw, J. C. and Ongley, C.: The measurement of Synchronization. In: Petsche, H . and Brazier, M. A. B.: Synchronization of EEG Activity in Epilepsy. Springer Verlag Wien, New York, 204-215 (1972). 17. Taistra, R.: Spektralanalytische Untersuchungen uber die abnorme Theta-Rhythmisierung des kindlichen Elektroencephalogramms. Dissertation, Kiel 1974. 18. Taistra, R., Gerken, H., Doose, H.: EEGspectral analysis in children with febrile convulsions European Neurology, in press 1975. 19. Welch, P. D.: The use of Fast Fourier Transform for the estimation of Power Spectra. Transactions on Audio and Elektro-acoustics. Vol. Au - 15, 2: 70-73 (1967).
Downloaded by: Universite Laval. Copyrighted material.
Taistra et al.
