- Email: [email protected]
Changes in frequency and amplitude of delta activity during sleep
Electroencephalography and Clinical Neurophysiology, 1975, 39:1 7 © Elsevier Scientific Publishing Company, Amsterdam - Printed in The Netherlands
CHANGES DURING
IN FREQUENCY
AND
AMPLITUDE
OF DELTA
1
ACTIVITY
SLEEP 1
M. W. CHURCH 2 J. D. MARCH, S.
HIBI, K.
BENSON, C. CAVNESS AND I. FEINBERG
Veterans Administration Hospital, San Francisco, Calif. 94121, and University of California, San Francisco, Calif. 94143
(U.S.A.) (Accepted for publication: January 20, 1975)
Visual scoring of all-night sleep records shows that stage E (Loomis et al. 1937) or stage 4 EEG (Dement and Kleitman 1957) declines across the night. These changes are statistically reliable (cf Williams et aL 1974) and are sometimes non-linear (Feinberg et al. 1967, 1969; Feinberg 1974). Since the classification of a sleep epoch as stage 4 depends largely upon the presence of profuse, high-voltage delta activity, we thought it of interest to investigate the characteristics of delta activity itself across a night's sleep. Here we report results of an application of period (Burch 1959) and amplitude analysis to this problem. Previous investigators have applied methods of period analysis to other aspects of the sleep EEG (Itil et al. 1969; Roessler et al. 1970; Lessard et al. 1974), and the general role of the computer in sleep research has been the subject of a recent symposium (Johnson 1972). A. Subjects
Fourteen young adult males aged 22.0-27.8 (mean 24.6) years were studied. Ten were paid medical student volunteers studied at the sleep laboratories at the San Francisco Veterans Administration Hospital. Four were prisoners convicted of felonies studied at a California State prison. The prisoners had no histories of epilepsy or head injury, no evidence of organic brain impairment on psychological testing, and no evidence of EEG abnormalities during alltSupported by research funds from the U.S. Veterans Administration. Reprint requests to I. Feinberg, VA Hospital, 4150 Clement St., San Francisco, Calif. 94121, U.S.A. 2Current address: Dept. of Psychology, San Diego State University, San Diego, Calif. 92110, U.S.A.
night sleep recording; visual scoring of sleep stages gave results similar to the norms for their age group. The prisoners were also screened by two psychiatrists and were considered clearly non-psychotic. Each subject underwent 3 or 4 consecutive nights of baseline recording. The total time in bed varied between 7.5 and 8 h. Two EEG and two eye-movement channels were recorded continuously for each subject. B. Recording
For EEG analysis a C3-mastoid derivation was recorded continuously on a Beckman Type R dynograph with a paper speed of 15 mm/sec, a gain of 8 mm/50 uV and a time constant of 0.3 sec. The pre-amplifier output of the dynograph was amplified to IRIG levels and recorded (FM) on an Ampex Model SP 700 or a Vetter Model A run at 1-7/8 and 15/16 ips respectively. Our methods for recording and visually scoring eye movement and EEG during sleep have been described elsewhere (Feinberg et al. 1967). Our scoring system differs from that subsequently proposed by Rechtschaffen and Kales (1968) mainly in that we include transient periods of non-rapid eye movement sleep (non-SREM) stage 1 with stage 2 sleep. C. Analysis
The tape-recorded signals were played back at 4 times recorded speed into the analog-todigital converter of the Digital Equipment Corp. PDP-12 computer at an effective sampling rate of 200 times/see through a high pass filter (TC=0.7 sec) which eliminated baseline drift. The modified period analysis (MPA) program
2
M.W. CHURCHet al.
developed in our laboratory (PANV-31) computes the following measures from the digitized signal and stores them on linc tape for subsequent analysis: a. The number of zero crossings (major period), time in band and integrated amplitude for frequencies of 0-0.2; 0.24).5; 0.5-1; 1-1.5; 1.5-2; 2-3; 3 4 ; over 4 c/sec. The integrated amplitude measure is the sum of the rectified amplitudes of all sample points which lie on any half-wave which falls in the given frequency band. Since no calibration was done in this study, no inferences can be drawn regarding the physiological amplitudes involved. Inter-subject variation in amplitude measures may result, in part, from differences in amplifier settings. However, within subject, across-night amplitude changes are not affected by these variations, since amplifier gains were constant for each subject. b. The number of zero counts of the first derivative (minor period) for frequency bands 0-25 and over 25 c/sec. RESULTS
Fig. 1 shows the percent time spent by one subject in 0-3 c/sec activity for consecutive 5 min epochs across the night. It illustrates the waxing and waning of delta activity across the night. The rhythmicity in amplitude measures resembles a "damped sinusoid" (Koga 1965; Agnew et al. 1967; Kripke 1972; Sinha et al. 1972; Lubin et al. 1973). The "damping" was less apparent to us in amplitude-free measures 90 80 70
~ 6o 0 c_ 5 O
~ 4o ~_ 3 0
~ 20 6'o
i
12o
1;o
Time (min) for
24o' 3;0
3~o
'
420
.~o
$ # 435 N @ 3
Fig. 1. Percent time spent in 0-3 c/sec activity, plotted for 5 min epochs across a single night for one subject. The upper segment of each peak would correspond to visually scored stage 4. While all subjects we have studied show such peaks, there is considerable variability in their number and time of occurrence.
such as time in band. Examination of such patterns in each of 10 subjects revealed that all showed peaks of delta activity across the night. However, there was considerable variation in the number, size and time of occurrence of these peaks. Such peaks were also apparent in the other measures we investigated, viz. integrated amplitude and number of zero crossings in the 0-3 c/sec band. Fig. 2 represents a sample of explorations intended to determine appropriate band-widths for the investigation of delta activity during sleep. It shows the cumulative output by 0.5 c/sec increments within 0-3 c/sec for consecutive 5 min epochs across the night for a single subject. We investigated the across-night patterns for zero crossings and integrated amplitude within 0-3 c/sec by similar 0.5 c/sec increments, and found that the conventionally accepted delta band limit, 0-3 c/sec, included the frequencies which show the most marked rhythmicity across a night's sleep. In agreement with previous workers (Johnson et al. 1969) we found that most of this variation occurred between 0.5 and 2 c/sec. Fig. 2 also suggests changes in delta characteristics, which we shall describe in greater detail below. Thus, considering the 0-3 c/sec activity for slow-wave sleep periods (SSWPs) 1-3 shown in Fig. 1-3, it will be seen that the low frequency activity (0-1 c/sec, band 3) changed little or increased whereas the higher frequencies (2-3 c/sec, bands 5 and 6) declined, i.e., the proportion of low frequency activity within 0-3 c/sec increased across the night. Table I shows changes in 0-3 c/sec activity across successive SSWPs for the 11 subjects who completed 4 SSWPs. The mean value for the 3 main MPA variables and for visually scored stage 4 EEG are shown. Each MPA variable showed a statistically significant decline across SSWPs. Variation among subjects in these trends was statistically significant only for integrated amplitude. ~As noted above, this variation may be partly the result of variations in amplification.) The rate of change for visually scored stage 4 EEG was greater than for the MPA measures. All the trends shown were linear (P < 0.001), none showing statistically significant curvature. The data in Table I are consistent with previous investigations (Johnson et al. 1969)
3
DELTA FREQUENCY AND AMPLITUDE CHANGES DURING SLEEP
10090 SSWP1 7 SSWP2 7 80
6
SSWP37
~
40
SSWP5
4=0-1.5Hz 5=0- 2 Hz '6=0- 3Hz I7=AII bonds
6H 5/
70
.~
iSSWP4== 0_-. t HzHZ 7
,~
+e 60 120 180 240 300 Sleep time (min) for S # 435 N :#:3
360
420
480
Fig. 2. Cumulative output of delta activity for 6 frequency bands plotted by 0.5 c/sec increments in 5 min epochs across the night for a single subject. Activity within 0-3 c/sec effectively reflects the waxing and waning of delta activity. "All bands" designates all recorded EEG activity. TABLEI Mean values for stage 4 EEG and MPA variables for activity in 0-3 c/sec for SSWPs 1 4 for subjects who completed 4 SSWPs. N = l l . MPA variables
Duration (min)
Mean values for 0-3 c/sec activity
Analysis of variance
SSWP1
SSWP2
SSWP3
SSWP4
66.5
74.0
77.7
66.7
39.86 116.73 523 35.73
37.22 109.44 422 28.09
31.65 100.16 288 9.82
29.84 98.35 255 8.09
SSWPs F
Subjects F
15.30" 20.09* 34.94* 15.54"
1.16 1.60 4.48* 0.91
0-3 c/sec Sec/min Xings/min Int. amp/min Total stage 4 (min/SSWP) * P < 0.001.
which show that during sleep, most of the EEG activity (or "intensity") falls within the delta band. This is true even for the late SSWPs when there is little or no stage 4 indicated by visual scoring. We next investigated the characteristics of delta activity in epochs scored as stage 4 EEG. For this analysis, we employed a computer classification scheme based upon the above MPA variables. The scoring method, which will
be described elsewhere, attempts to minimize pattern recognition--first visually classifying the EEG record into periods of slow wave (SSW) and rapid eye movement (SREM) sleep. This can be accomplished quickly and with relatively high reliability. Computer analysis is then directed toward classification of SSW into stages 2, 3 and 4. With the present data, the computer method showed 94 % agreement with visual scoring for stage 4 EEG. Results were also
M . W . CHURCH el al.
4 TABLE II
Mean values and ANOVA for activity in 0-3 c/sec for epochs scored by computer as stage 4 EEG during SSWPs 1-3. Computed for those subjects showing some stage 4 in each of the first SSWPs. N = 8. MPA variables
Sec/min Xings/min Am~/min x 103 (units)
Analysis of variance
0-3 c/sec activity SSWP 1
SSWP 2
SSWP3
SSWPs F
Subjects F
47.50 129.83 766.00
47.08 121.20 712.00
46.39 116.59 610.00
0.859 44.803** 8.932**
4.127" 6.296** 10.685"*
* P < 0.05. ** P < 0.005. 50
45 40
SWP~
o
~ 35
swPa -7- 30 '7 o
;swPI
i
5 llO ll5 210 Time from stage 4 onset (min)
i
25
Fig. 3. Percent 0-1/0-3 c/sec activity for time in band for consecutive 5 min blocks of stage 4 EEG within slow-wave sleep periods 1-3. This value increases within SSWPs. While the difference between the first and second 5 min blocks appears greatest for each SSWP, the trends do not depart significantly from linearity. The trends for SSWPs 1 and 2 are highly significant; the trend for SSWP 3 is based on only 4 subjects and does not reach statistical significance.
good for stage 2 (91~o) but poor for stage 3 (43~), as has been the case in previous studies with spectral analysis (Lubin et al. 1969; Martin et al. 1972). Table II gives results for MPA variables for computer-scored stage 4 for those subjects who showed some stage 4 EEG in each of the first 3 SSWPs. This table shows that for epochs of st~ige 4 EEG, time in 0-3 c/sec did not change significantly across SSWPs. However, both the number of crossings and the integrated amplitude declined. Since the time occupied by waves of 0-3 c/sec remained constant, but the number of zero crossings and the integrated amplitude in this band declined, the data in Table II imply that delta frequencies slowed and amplitudes diminished as stage 4 EEG p/'ogressed.
That this is the case was shown by the data in Table III. For each MPA variable, 0-1 c/sec activity made up a larger proportion of the delta (0-3 c/sec) band in the successive SSWPs. These increases were highly significant. All were linear except for amplitude which showed some curvature. The ANOVA also revealed that subjects showed significant variation in these trends. We next examined the changing characteristics of stage 4 EEG within the first 3 SSWPs as reflected by MPA variables. We chose subjects for this analysis who showed at least 20 min of computer-scored stage 4 EEG within a SSWP. Eleven subjects met this criterion for SSWPI, 10 for SSWP2 and only 4 for SSWP 3. For each, subject, the mean percentage of time in 0-1/0-3 :/sec was computed for successive 5 min blocks of stage 4 EEG. The results are shown in Fig. 3. ANOVAs of these data revealed that trends within SSWPs 1 and 2 were highly significant and linear (P<0.001). SSWP~ showed a pattern similar to that for the first 2 SSWPs but the trend did not reach statistical significance, perhaps because the N was small. ANOVAs again revealed considerable heterogeneity among subjects (F=16.6 and 21.0; dr= 10, 40; 9, 27) for SSWPs 1 and 2 ( P < 0.001). The same analysis was carried out for zero crossings and integrated amplitude. The same trends within SSWPs were found for these measures and were again highly significant for SSWPs 1 and 2. It is of interest to note that there is a suggestion ofnon-linearity in the curves shown in Fig. 3, i.e.. the greatest change occurs from the first to the
DELTA FREQUENCY AND AMPLITUDE CHANGES DURING SLEEP
5
TABLE III
Mean values of MPA variables for the percentage 0-1/0-3 c/sec in epochs scored by computer as stage 4 in SSWPs 1-3; ANOVA and tests for trends across SSWPs. N =8. MPA variables
Analysis of variance
0-1/0-3 c/sec (~o) SSWP~
SSWP 2
SSWP3
SSWPs
Subjects
Trends F
Sec/min Xings/min Amp/min
32.35 19.05 37.61
40.31 24.28 47.82
43.71 26.20 51.78
30.299** 17.734"* 58.707**
(linear)
F (Quad)
57.514"* 33.103"* 100.283"*
3.085 2.364 7.289*
13.715"* 9.990** 25.922**
*P< 0.025. **P< 0.001.
TABLE IV
Mean differences for 0-1/0-3 c/sec activity between visually scored stage 4 epochs at the beginning and end of SSWPs 1 and 2. N=8.
Mean difference Beginning-end SSWP,
Time in band (sec) Zero crossings (No.) Integrated amplitude (units)
SSWP2
0-1/0-3
2-3/0-3
0-1/0-3
2-3/0-3
-9.49*** ~.95"**
7.95*** 7.41"**
-8.05** -5.06*
1.83 1.56
9.69***
6.94***
-5.46
1.69
Two-tailed "t" tests. * P < 0.05. **P< 0.01. ***P< 0.001. (-) sign indicates value higher at end of SSWP.
second 5 min block of stage 4. However, the test for curvature was uniformly non-significant except for zero crossings for SSWP1 (F=5.3; dr= 1, 40; P < 0.05). This figure also shows that the level 0-1/0-3 c/sec at the beginning of SSWPs 2 and 3 was quite close to that at the termination of the preceding SSWPs, although a SREMP had intervened. Although the high concordance (94 ~ agreement) between visual and computer scoring of stage 4 rendered it unlikely, we wished to rule out the possibility that the changes in MPA characteristics of stage 4 were an artifact of our
method of computer scoring. We therefore measured 0-3 c/sec activity in visually scored stage 4 epochs at different points in the night. Twelve of the 14 subjects in this study exhibited slowing in the 0-3 c/sec band across the night. 'lhe records of 8 of these 12 were randomly selected for further analysis. From each of the selected records, the first and last 5 min blocks of visually scored stage 4 for SSWPs 1 and 2 were subjected to MPA. Time in band, zero crossings and integrated amplitude were determined for 0-1, 0-2 and 0-3 c/sec. The absolute values as well as these values as percentages of the 0-3
6 c/sec band were compared across the different time samples with paired "t" tests. The results for visually scored stage 4 were entirely consistent with those obtained with computer scoring. A sample of these findings is presented in Table IV. This analysis also suggested that the slowing of frequencies within 0-3 c/sec occurs at a faster rate earlier in the night, especially for the 2-3 c/sec component. DISCUSSION
Application of period and amplitude analysis to all-night EEG recording reveals that, within the 0-3 c/sec band, integrated amplitude declines and frequencies become slower across the night. These changes are apparent in mean values for the successive complete periods of slow-wave sleep which include SSW stages 2~4. They were seen in all 14 subjects studied. The changes appear more pronounced within epochs of SSW classified as stage 4 EEG, where they occur within as well as across SSWPs. These latter changes also appear highly reliable. They were manifested by 12 of the 14 subjects. It may be of interest that the 2 subjects who did not show slowing frequencies and diminishing amplitude in 0-3 c/sec activity of stage 4 within SSWPs were both prisoners. Our present findings supplement the observations of previous investigators who have demonstrated variations in delta activity (e.g., peaks late in the night) not apparent in visually scored stage 4 EEG (Agnew et al. 1967; Johnson et al. 1969; Martin et al. 1972). The physiological significance of delta activity during sleep remains unknown. For this reason, it would be idle to engage in extensive speculation regarding the significance of the decline in delta frequencies and amplitudes across the night. Our hunch is that changes in EEG sleep patterns during sleep may provide clues to the kinetics of the underlying metabolic processes (Sinha et al. 1972; Feinberg 1974). If this is the case, such changes could perhaps yield new measures of "goodness of sleep". Whatever their biological significance, these trends could provide additional empirical measures of the normative EEG phenomena of sleep. If the present results are confirmed by studies in
M . W . C H U R C H el al.
progress in our laboratory, we will seek variations in these patterns as a function of age, the administration of drugs, and in "organic" and "functional" mental disorders. SUMMARY
Modified period analysis was applied to all-night sleep recordings from 14 young adult males. The modifications involved addition of measures of integrated amplitude and of time in frequency band to the zero crossings and zero counts of the first derivative. The analysis was directed toward changes in the characteristics of delta activity (0-3 c/sec) across the night. Delta shifted toward lower frequencies and decreased in amplitude as sleep progressed. These trends were apparent in mean values for successive periods of slowwave sleep (SSWPs). For epochs of record classified as stage 4 EEG, these trends were seen both within and across SSWPs. The physiological significance of these changes is unknown. We suggest that they may reflect the kinetics of the metabolic processes underlying sleep. Whatever their theoretical significance, the present results provide new, normative features of the sleep EEG which are not evident in sleep stage classification. The trends we observed appear sufficiently reliable to merit empirical investigation as a function of age, clinical state, or the administration of drugs in addition to further studies aimed at determining their biological significance. RESUME CHANGEMENTS DE FREQUENCE ET D'AMPLITUDE DE L'ACTIVITE DELTA AU COURS DU SOMMEIL
Une analyse de p6riode modifi6e a 6t6 appliqu6e aux enregistrements du sommeil de toute la nuit de 14jeunes adultes de sexe masculin. Les modifications impliquent l'adjonction de mesures de l'amplitude int6gr6e et du temps en bande de fr6quence par rapport aux croisements de la ligne de base et aux comptages des z6ros de la premiere d6riv6e. Cette analyse cherche fi d6celer des modifications des caract6ristiques de l'activit~ delta (de
DELTA FREQUENCY AND AMPLITUDE CHANGES DURING SLEEP
0-3 c/sec) tout au long de la nuit. L'activit6 delta 6volue vers des fr6quences plus basses et diminue d'amplitude au fur et ~t mesure que le sommeil progresse. Ces tendances apparaissent duns les valeurs moyennes de p6riodes successives de sommeil ~ ondes lentes (SSWPs). En ce qui concerne les 6poques de trac6s class6s comme stade IV LEG, ces tendances s'observent aussi bien l'int6rieur des SSWPs que d'une SSWPs fi l'autre. La signification physiologique de ces modifications est inconnue. Les auteurs font l'hypoth+se qu'elle puisse refl6ter la mouvance des processus m6taboliques qui sous-tendent le sommeil. Quelle que soit leur signification th6orique, les r6sultats actuels fournissent des donn6es nouvelles et normatives de I'EEG du sommeil qui ne sont pus 6videntes dans la classification en stades de sommeil. Les tendances que les auteurs ont observ6 paraissent suffisamment fiables pour m6riter une investigation empirique en fonction de l'fige, de l'6tat clinique ou de l'administration de drogues en plus d'6tudes ult6rieures tendant b. expliciter leur signification biologique. REFERENCES
AGNEW, H. W., PARKER,J. C., WEBa, W. B. and WILLIAMS, R. L. Amplitude measurement of the sleep electroencephalogram. Electroenceph. clin. Neurophysiol., 1967, 22: 84-86. BURCH, N. R. Automatic analysis of electroencephalogram : A review and classification of systems. Electroenceph. clin. Neurophysiol., 1959, I I: 827-834. DIiMENT. W. and KLEITMAN, N. Cyclic variations in LEG during sleep and their relation to eye movements, body motility, and dreaming. Electroenceph. clin Neurophysiol., 1957, 9: 673~590. FEINBERG, I. Changes in sleep cycle patterns with age. J. psTchiat. Res., 1974, 10: 283-306. FEINBERG, I., BRAUN, M. and SHULMAN, E. EEG sleep patterns in mental retardation. Electroenceph. clin. Neurophysiol., 1969, 27: 128-141. FEINBERG, I., KORESKO, R. L. and HELLER, N. EEG sleep patterns as a function of normal and pathological aging
7
in man. J. psychiat. Res., 1967, 5: 107-144. ITIL, T. M., SHAPIRO, D. M., PINK, M. and KASSEBAUM,D. Digital computer classifications of LEG sleep stages. Electroenceph. clin. Neurophysiol., 1969, 27: 76-83. JOHNSON, L. Computers in sleep research. In M. H. CHASE (Ed.), The sleeping brain. Brain Information Service, Brain Research Institute, Univ. of California, Los Angeles, 1972: 277-321. JOHNSON, L., LUBIN,A., NAITOH,P., NUTE, C. and AUSTIN,M. Spectral analysis of the LEG of dominant and nondominant alpha subjects during waking and sleeping. Electroeneeph. clin. Neurophysiol., 1969, 26: 361-370. KOGA, E. A new method of LEG analysis and its application to the study of sleep. Foliapsychiat. neurol.jap., 1965,19 : 269-278. KRIPKE, D. F. An ultradian biological rhythm associated with perceptual deprivation and REM sleep. Psychosom. Med., 1972, 34: 221-234. LESSARD, C. S., SANCES, JR., A. and LARSON, S. J. Period analysis of LEG signals during sleep and post-traumatic coma. Aerospace Med., 1974, 45 : 664-668. LOOMIS, A. L., HARVEY, E. N. and HOBART, G. A fferebral states during sleep as studied by human brain potentials. J. exp. Psychol., 1937, 21: 127-144. LUalN, A., JOHNSON, L. C. and AUSTIN, M. T. Discrimination among states of consciousness using LEG spectra. Psychophysiology, 1969, 6: 122-132. LUBIN, A., NUTE, C., NAITOH, P. and MARTIN, W. B. LEG delta activity during human sleep as a damped ultradian rhythm. Psychophysiology, 1973, 10: 27-35. MARTIN, W. B., JOHNSON,L. C., VIGLIONE,S. S., NAITOH,P., JOSEPH, R. D. and MOSES, J. D. Pattern recognition of LEG-LOG as a technique for all night sleep stage scoring. Electroenceph. clin. Neurophysiol., 1972, 32: 417~,27. RECHTSCHAEEEN, A. and KALES, A. (Eds.). A manual of
standardized terminology , techniques, and scorin9 system for sleep stages of human subjects. National Institutes of Health Publication No. 204. U.S. Govt. Printing Office. Washington, D.C., 1968, 58 p. ROESSLER,R., COLLINS,F. and OSTMAN,R. A period analysis classification of sleep stages. Electroenceph. clin. Neurophysiol., 1970, 29: 358-362. S1NHA, A. K., SMYTHE,H., ZARCONE,V.P., BARCHAS,J. D. and DEMENT, W. C. Human sleep-electroencephalogram: a damped oscillatory phenomenon. J. theor. Biol., 1972, 35 : 387-393. WILLIAMS,R. L., KARACAN,I. and HURSCH,C. J. Electroencephalography of human sleep. Wiley. New York, 1974, 169 p.
CHANGES DURING
IN FREQUENCY
AND
AMPLITUDE
OF DELTA
1
ACTIVITY
SLEEP 1
M. W. CHURCH 2 J. D. MARCH, S.
HIBI, K.
BENSON, C. CAVNESS AND I. FEINBERG
Veterans Administration Hospital, San Francisco, Calif. 94121, and University of California, San Francisco, Calif. 94143
(U.S.A.) (Accepted for publication: January 20, 1975)
Visual scoring of all-night sleep records shows that stage E (Loomis et al. 1937) or stage 4 EEG (Dement and Kleitman 1957) declines across the night. These changes are statistically reliable (cf Williams et aL 1974) and are sometimes non-linear (Feinberg et al. 1967, 1969; Feinberg 1974). Since the classification of a sleep epoch as stage 4 depends largely upon the presence of profuse, high-voltage delta activity, we thought it of interest to investigate the characteristics of delta activity itself across a night's sleep. Here we report results of an application of period (Burch 1959) and amplitude analysis to this problem. Previous investigators have applied methods of period analysis to other aspects of the sleep EEG (Itil et al. 1969; Roessler et al. 1970; Lessard et al. 1974), and the general role of the computer in sleep research has been the subject of a recent symposium (Johnson 1972). A. Subjects
Fourteen young adult males aged 22.0-27.8 (mean 24.6) years were studied. Ten were paid medical student volunteers studied at the sleep laboratories at the San Francisco Veterans Administration Hospital. Four were prisoners convicted of felonies studied at a California State prison. The prisoners had no histories of epilepsy or head injury, no evidence of organic brain impairment on psychological testing, and no evidence of EEG abnormalities during alltSupported by research funds from the U.S. Veterans Administration. Reprint requests to I. Feinberg, VA Hospital, 4150 Clement St., San Francisco, Calif. 94121, U.S.A. 2Current address: Dept. of Psychology, San Diego State University, San Diego, Calif. 92110, U.S.A.
night sleep recording; visual scoring of sleep stages gave results similar to the norms for their age group. The prisoners were also screened by two psychiatrists and were considered clearly non-psychotic. Each subject underwent 3 or 4 consecutive nights of baseline recording. The total time in bed varied between 7.5 and 8 h. Two EEG and two eye-movement channels were recorded continuously for each subject. B. Recording
For EEG analysis a C3-mastoid derivation was recorded continuously on a Beckman Type R dynograph with a paper speed of 15 mm/sec, a gain of 8 mm/50 uV and a time constant of 0.3 sec. The pre-amplifier output of the dynograph was amplified to IRIG levels and recorded (FM) on an Ampex Model SP 700 or a Vetter Model A run at 1-7/8 and 15/16 ips respectively. Our methods for recording and visually scoring eye movement and EEG during sleep have been described elsewhere (Feinberg et al. 1967). Our scoring system differs from that subsequently proposed by Rechtschaffen and Kales (1968) mainly in that we include transient periods of non-rapid eye movement sleep (non-SREM) stage 1 with stage 2 sleep. C. Analysis
The tape-recorded signals were played back at 4 times recorded speed into the analog-todigital converter of the Digital Equipment Corp. PDP-12 computer at an effective sampling rate of 200 times/see through a high pass filter (TC=0.7 sec) which eliminated baseline drift. The modified period analysis (MPA) program
2
M.W. CHURCHet al.
developed in our laboratory (PANV-31) computes the following measures from the digitized signal and stores them on linc tape for subsequent analysis: a. The number of zero crossings (major period), time in band and integrated amplitude for frequencies of 0-0.2; 0.24).5; 0.5-1; 1-1.5; 1.5-2; 2-3; 3 4 ; over 4 c/sec. The integrated amplitude measure is the sum of the rectified amplitudes of all sample points which lie on any half-wave which falls in the given frequency band. Since no calibration was done in this study, no inferences can be drawn regarding the physiological amplitudes involved. Inter-subject variation in amplitude measures may result, in part, from differences in amplifier settings. However, within subject, across-night amplitude changes are not affected by these variations, since amplifier gains were constant for each subject. b. The number of zero counts of the first derivative (minor period) for frequency bands 0-25 and over 25 c/sec. RESULTS
Fig. 1 shows the percent time spent by one subject in 0-3 c/sec activity for consecutive 5 min epochs across the night. It illustrates the waxing and waning of delta activity across the night. The rhythmicity in amplitude measures resembles a "damped sinusoid" (Koga 1965; Agnew et al. 1967; Kripke 1972; Sinha et al. 1972; Lubin et al. 1973). The "damping" was less apparent to us in amplitude-free measures 90 80 70
~ 6o 0 c_ 5 O
~ 4o ~_ 3 0
~ 20 6'o
i
12o
1;o
Time (min) for
24o' 3;0
3~o
'
420
.~o
$ # 435 N @ 3
Fig. 1. Percent time spent in 0-3 c/sec activity, plotted for 5 min epochs across a single night for one subject. The upper segment of each peak would correspond to visually scored stage 4. While all subjects we have studied show such peaks, there is considerable variability in their number and time of occurrence.
such as time in band. Examination of such patterns in each of 10 subjects revealed that all showed peaks of delta activity across the night. However, there was considerable variation in the number, size and time of occurrence of these peaks. Such peaks were also apparent in the other measures we investigated, viz. integrated amplitude and number of zero crossings in the 0-3 c/sec band. Fig. 2 represents a sample of explorations intended to determine appropriate band-widths for the investigation of delta activity during sleep. It shows the cumulative output by 0.5 c/sec increments within 0-3 c/sec for consecutive 5 min epochs across the night for a single subject. We investigated the across-night patterns for zero crossings and integrated amplitude within 0-3 c/sec by similar 0.5 c/sec increments, and found that the conventionally accepted delta band limit, 0-3 c/sec, included the frequencies which show the most marked rhythmicity across a night's sleep. In agreement with previous workers (Johnson et al. 1969) we found that most of this variation occurred between 0.5 and 2 c/sec. Fig. 2 also suggests changes in delta characteristics, which we shall describe in greater detail below. Thus, considering the 0-3 c/sec activity for slow-wave sleep periods (SSWPs) 1-3 shown in Fig. 1-3, it will be seen that the low frequency activity (0-1 c/sec, band 3) changed little or increased whereas the higher frequencies (2-3 c/sec, bands 5 and 6) declined, i.e., the proportion of low frequency activity within 0-3 c/sec increased across the night. Table I shows changes in 0-3 c/sec activity across successive SSWPs for the 11 subjects who completed 4 SSWPs. The mean value for the 3 main MPA variables and for visually scored stage 4 EEG are shown. Each MPA variable showed a statistically significant decline across SSWPs. Variation among subjects in these trends was statistically significant only for integrated amplitude. ~As noted above, this variation may be partly the result of variations in amplification.) The rate of change for visually scored stage 4 EEG was greater than for the MPA measures. All the trends shown were linear (P < 0.001), none showing statistically significant curvature. The data in Table I are consistent with previous investigations (Johnson et al. 1969)
3
DELTA FREQUENCY AND AMPLITUDE CHANGES DURING SLEEP
10090 SSWP1 7 SSWP2 7 80
6
SSWP37
~
40
SSWP5
4=0-1.5Hz 5=0- 2 Hz '6=0- 3Hz I7=AII bonds
6H 5/
70
.~
iSSWP4== 0_-. t HzHZ 7
,~
+e 60 120 180 240 300 Sleep time (min) for S # 435 N :#:3
360
420
480
Fig. 2. Cumulative output of delta activity for 6 frequency bands plotted by 0.5 c/sec increments in 5 min epochs across the night for a single subject. Activity within 0-3 c/sec effectively reflects the waxing and waning of delta activity. "All bands" designates all recorded EEG activity. TABLEI Mean values for stage 4 EEG and MPA variables for activity in 0-3 c/sec for SSWPs 1 4 for subjects who completed 4 SSWPs. N = l l . MPA variables
Duration (min)
Mean values for 0-3 c/sec activity
Analysis of variance
SSWP1
SSWP2
SSWP3
SSWP4
66.5
74.0
77.7
66.7
39.86 116.73 523 35.73
37.22 109.44 422 28.09
31.65 100.16 288 9.82
29.84 98.35 255 8.09
SSWPs F
Subjects F
15.30" 20.09* 34.94* 15.54"
1.16 1.60 4.48* 0.91
0-3 c/sec Sec/min Xings/min Int. amp/min Total stage 4 (min/SSWP) * P < 0.001.
which show that during sleep, most of the EEG activity (or "intensity") falls within the delta band. This is true even for the late SSWPs when there is little or no stage 4 indicated by visual scoring. We next investigated the characteristics of delta activity in epochs scored as stage 4 EEG. For this analysis, we employed a computer classification scheme based upon the above MPA variables. The scoring method, which will
be described elsewhere, attempts to minimize pattern recognition--first visually classifying the EEG record into periods of slow wave (SSW) and rapid eye movement (SREM) sleep. This can be accomplished quickly and with relatively high reliability. Computer analysis is then directed toward classification of SSW into stages 2, 3 and 4. With the present data, the computer method showed 94 % agreement with visual scoring for stage 4 EEG. Results were also
M . W . CHURCH el al.
4 TABLE II
Mean values and ANOVA for activity in 0-3 c/sec for epochs scored by computer as stage 4 EEG during SSWPs 1-3. Computed for those subjects showing some stage 4 in each of the first SSWPs. N = 8. MPA variables
Sec/min Xings/min Am~/min x 103 (units)
Analysis of variance
0-3 c/sec activity SSWP 1
SSWP 2
SSWP3
SSWPs F
Subjects F
47.50 129.83 766.00
47.08 121.20 712.00
46.39 116.59 610.00
0.859 44.803** 8.932**
4.127" 6.296** 10.685"*
* P < 0.05. ** P < 0.005. 50
45 40
SWP~
o
~ 35
swPa -7- 30 '7 o
;swPI
i
5 llO ll5 210 Time from stage 4 onset (min)
i
25
Fig. 3. Percent 0-1/0-3 c/sec activity for time in band for consecutive 5 min blocks of stage 4 EEG within slow-wave sleep periods 1-3. This value increases within SSWPs. While the difference between the first and second 5 min blocks appears greatest for each SSWP, the trends do not depart significantly from linearity. The trends for SSWPs 1 and 2 are highly significant; the trend for SSWP 3 is based on only 4 subjects and does not reach statistical significance.
good for stage 2 (91~o) but poor for stage 3 (43~), as has been the case in previous studies with spectral analysis (Lubin et al. 1969; Martin et al. 1972). Table II gives results for MPA variables for computer-scored stage 4 for those subjects who showed some stage 4 EEG in each of the first 3 SSWPs. This table shows that for epochs of st~ige 4 EEG, time in 0-3 c/sec did not change significantly across SSWPs. However, both the number of crossings and the integrated amplitude declined. Since the time occupied by waves of 0-3 c/sec remained constant, but the number of zero crossings and the integrated amplitude in this band declined, the data in Table II imply that delta frequencies slowed and amplitudes diminished as stage 4 EEG p/'ogressed.
That this is the case was shown by the data in Table III. For each MPA variable, 0-1 c/sec activity made up a larger proportion of the delta (0-3 c/sec) band in the successive SSWPs. These increases were highly significant. All were linear except for amplitude which showed some curvature. The ANOVA also revealed that subjects showed significant variation in these trends. We next examined the changing characteristics of stage 4 EEG within the first 3 SSWPs as reflected by MPA variables. We chose subjects for this analysis who showed at least 20 min of computer-scored stage 4 EEG within a SSWP. Eleven subjects met this criterion for SSWPI, 10 for SSWP2 and only 4 for SSWP 3. For each, subject, the mean percentage of time in 0-1/0-3 :/sec was computed for successive 5 min blocks of stage 4 EEG. The results are shown in Fig. 3. ANOVAs of these data revealed that trends within SSWPs 1 and 2 were highly significant and linear (P<0.001). SSWP~ showed a pattern similar to that for the first 2 SSWPs but the trend did not reach statistical significance, perhaps because the N was small. ANOVAs again revealed considerable heterogeneity among subjects (F=16.6 and 21.0; dr= 10, 40; 9, 27) for SSWPs 1 and 2 ( P < 0.001). The same analysis was carried out for zero crossings and integrated amplitude. The same trends within SSWPs were found for these measures and were again highly significant for SSWPs 1 and 2. It is of interest to note that there is a suggestion ofnon-linearity in the curves shown in Fig. 3, i.e.. the greatest change occurs from the first to the
DELTA FREQUENCY AND AMPLITUDE CHANGES DURING SLEEP
5
TABLE III
Mean values of MPA variables for the percentage 0-1/0-3 c/sec in epochs scored by computer as stage 4 in SSWPs 1-3; ANOVA and tests for trends across SSWPs. N =8. MPA variables
Analysis of variance
0-1/0-3 c/sec (~o) SSWP~
SSWP 2
SSWP3
SSWPs
Subjects
Trends F
Sec/min Xings/min Amp/min
32.35 19.05 37.61
40.31 24.28 47.82
43.71 26.20 51.78
30.299** 17.734"* 58.707**
(linear)
F (Quad)
57.514"* 33.103"* 100.283"*
3.085 2.364 7.289*
13.715"* 9.990** 25.922**
*P< 0.025. **P< 0.001.
TABLE IV
Mean differences for 0-1/0-3 c/sec activity between visually scored stage 4 epochs at the beginning and end of SSWPs 1 and 2. N=8.
Mean difference Beginning-end SSWP,
Time in band (sec) Zero crossings (No.) Integrated amplitude (units)
SSWP2
0-1/0-3
2-3/0-3
0-1/0-3
2-3/0-3
-9.49*** ~.95"**
7.95*** 7.41"**
-8.05** -5.06*
1.83 1.56
9.69***
6.94***
-5.46
1.69
Two-tailed "t" tests. * P < 0.05. **P< 0.01. ***P< 0.001. (-) sign indicates value higher at end of SSWP.
second 5 min block of stage 4. However, the test for curvature was uniformly non-significant except for zero crossings for SSWP1 (F=5.3; dr= 1, 40; P < 0.05). This figure also shows that the level 0-1/0-3 c/sec at the beginning of SSWPs 2 and 3 was quite close to that at the termination of the preceding SSWPs, although a SREMP had intervened. Although the high concordance (94 ~ agreement) between visual and computer scoring of stage 4 rendered it unlikely, we wished to rule out the possibility that the changes in MPA characteristics of stage 4 were an artifact of our
method of computer scoring. We therefore measured 0-3 c/sec activity in visually scored stage 4 epochs at different points in the night. Twelve of the 14 subjects in this study exhibited slowing in the 0-3 c/sec band across the night. 'lhe records of 8 of these 12 were randomly selected for further analysis. From each of the selected records, the first and last 5 min blocks of visually scored stage 4 for SSWPs 1 and 2 were subjected to MPA. Time in band, zero crossings and integrated amplitude were determined for 0-1, 0-2 and 0-3 c/sec. The absolute values as well as these values as percentages of the 0-3
6 c/sec band were compared across the different time samples with paired "t" tests. The results for visually scored stage 4 were entirely consistent with those obtained with computer scoring. A sample of these findings is presented in Table IV. This analysis also suggested that the slowing of frequencies within 0-3 c/sec occurs at a faster rate earlier in the night, especially for the 2-3 c/sec component. DISCUSSION
Application of period and amplitude analysis to all-night EEG recording reveals that, within the 0-3 c/sec band, integrated amplitude declines and frequencies become slower across the night. These changes are apparent in mean values for the successive complete periods of slow-wave sleep which include SSW stages 2~4. They were seen in all 14 subjects studied. The changes appear more pronounced within epochs of SSW classified as stage 4 EEG, where they occur within as well as across SSWPs. These latter changes also appear highly reliable. They were manifested by 12 of the 14 subjects. It may be of interest that the 2 subjects who did not show slowing frequencies and diminishing amplitude in 0-3 c/sec activity of stage 4 within SSWPs were both prisoners. Our present findings supplement the observations of previous investigators who have demonstrated variations in delta activity (e.g., peaks late in the night) not apparent in visually scored stage 4 EEG (Agnew et al. 1967; Johnson et al. 1969; Martin et al. 1972). The physiological significance of delta activity during sleep remains unknown. For this reason, it would be idle to engage in extensive speculation regarding the significance of the decline in delta frequencies and amplitudes across the night. Our hunch is that changes in EEG sleep patterns during sleep may provide clues to the kinetics of the underlying metabolic processes (Sinha et al. 1972; Feinberg 1974). If this is the case, such changes could perhaps yield new measures of "goodness of sleep". Whatever their biological significance, these trends could provide additional empirical measures of the normative EEG phenomena of sleep. If the present results are confirmed by studies in
M . W . C H U R C H el al.
progress in our laboratory, we will seek variations in these patterns as a function of age, the administration of drugs, and in "organic" and "functional" mental disorders. SUMMARY
Modified period analysis was applied to all-night sleep recordings from 14 young adult males. The modifications involved addition of measures of integrated amplitude and of time in frequency band to the zero crossings and zero counts of the first derivative. The analysis was directed toward changes in the characteristics of delta activity (0-3 c/sec) across the night. Delta shifted toward lower frequencies and decreased in amplitude as sleep progressed. These trends were apparent in mean values for successive periods of slowwave sleep (SSWPs). For epochs of record classified as stage 4 EEG, these trends were seen both within and across SSWPs. The physiological significance of these changes is unknown. We suggest that they may reflect the kinetics of the metabolic processes underlying sleep. Whatever their theoretical significance, the present results provide new, normative features of the sleep EEG which are not evident in sleep stage classification. The trends we observed appear sufficiently reliable to merit empirical investigation as a function of age, clinical state, or the administration of drugs in addition to further studies aimed at determining their biological significance. RESUME CHANGEMENTS DE FREQUENCE ET D'AMPLITUDE DE L'ACTIVITE DELTA AU COURS DU SOMMEIL
Une analyse de p6riode modifi6e a 6t6 appliqu6e aux enregistrements du sommeil de toute la nuit de 14jeunes adultes de sexe masculin. Les modifications impliquent l'adjonction de mesures de l'amplitude int6gr6e et du temps en bande de fr6quence par rapport aux croisements de la ligne de base et aux comptages des z6ros de la premiere d6riv6e. Cette analyse cherche fi d6celer des modifications des caract6ristiques de l'activit~ delta (de
DELTA FREQUENCY AND AMPLITUDE CHANGES DURING SLEEP
0-3 c/sec) tout au long de la nuit. L'activit6 delta 6volue vers des fr6quences plus basses et diminue d'amplitude au fur et ~t mesure que le sommeil progresse. Ces tendances apparaissent duns les valeurs moyennes de p6riodes successives de sommeil ~ ondes lentes (SSWPs). En ce qui concerne les 6poques de trac6s class6s comme stade IV LEG, ces tendances s'observent aussi bien l'int6rieur des SSWPs que d'une SSWPs fi l'autre. La signification physiologique de ces modifications est inconnue. Les auteurs font l'hypoth+se qu'elle puisse refl6ter la mouvance des processus m6taboliques qui sous-tendent le sommeil. Quelle que soit leur signification th6orique, les r6sultats actuels fournissent des donn6es nouvelles et normatives de I'EEG du sommeil qui ne sont pus 6videntes dans la classification en stades de sommeil. Les tendances que les auteurs ont observ6 paraissent suffisamment fiables pour m6riter une investigation empirique en fonction de l'fige, de l'6tat clinique ou de l'administration de drogues en plus d'6tudes ult6rieures tendant b. expliciter leur signification biologique. REFERENCES
AGNEW, H. W., PARKER,J. C., WEBa, W. B. and WILLIAMS, R. L. Amplitude measurement of the sleep electroencephalogram. Electroenceph. clin. Neurophysiol., 1967, 22: 84-86. BURCH, N. R. Automatic analysis of electroencephalogram : A review and classification of systems. Electroenceph. clin. Neurophysiol., 1959, I I: 827-834. DIiMENT. W. and KLEITMAN, N. Cyclic variations in LEG during sleep and their relation to eye movements, body motility, and dreaming. Electroenceph. clin Neurophysiol., 1957, 9: 673~590. FEINBERG, I. Changes in sleep cycle patterns with age. J. psTchiat. Res., 1974, 10: 283-306. FEINBERG, I., BRAUN, M. and SHULMAN, E. EEG sleep patterns in mental retardation. Electroenceph. clin. Neurophysiol., 1969, 27: 128-141. FEINBERG, I., KORESKO, R. L. and HELLER, N. EEG sleep patterns as a function of normal and pathological aging
7
in man. J. psychiat. Res., 1967, 5: 107-144. ITIL, T. M., SHAPIRO, D. M., PINK, M. and KASSEBAUM,D. Digital computer classifications of LEG sleep stages. Electroenceph. clin. Neurophysiol., 1969, 27: 76-83. JOHNSON, L. Computers in sleep research. In M. H. CHASE (Ed.), The sleeping brain. Brain Information Service, Brain Research Institute, Univ. of California, Los Angeles, 1972: 277-321. JOHNSON, L., LUBIN,A., NAITOH,P., NUTE, C. and AUSTIN,M. Spectral analysis of the LEG of dominant and nondominant alpha subjects during waking and sleeping. Electroeneeph. clin. Neurophysiol., 1969, 26: 361-370. KOGA, E. A new method of LEG analysis and its application to the study of sleep. Foliapsychiat. neurol.jap., 1965,19 : 269-278. KRIPKE, D. F. An ultradian biological rhythm associated with perceptual deprivation and REM sleep. Psychosom. Med., 1972, 34: 221-234. LESSARD, C. S., SANCES, JR., A. and LARSON, S. J. Period analysis of LEG signals during sleep and post-traumatic coma. Aerospace Med., 1974, 45 : 664-668. LOOMIS, A. L., HARVEY, E. N. and HOBART, G. A fferebral states during sleep as studied by human brain potentials. J. exp. Psychol., 1937, 21: 127-144. LUalN, A., JOHNSON, L. C. and AUSTIN, M. T. Discrimination among states of consciousness using LEG spectra. Psychophysiology, 1969, 6: 122-132. LUBIN, A., NUTE, C., NAITOH, P. and MARTIN, W. B. LEG delta activity during human sleep as a damped ultradian rhythm. Psychophysiology, 1973, 10: 27-35. MARTIN, W. B., JOHNSON,L. C., VIGLIONE,S. S., NAITOH,P., JOSEPH, R. D. and MOSES, J. D. Pattern recognition of LEG-LOG as a technique for all night sleep stage scoring. Electroenceph. clin. Neurophysiol., 1972, 32: 417~,27. RECHTSCHAEEEN, A. and KALES, A. (Eds.). A manual of
standardized terminology , techniques, and scorin9 system for sleep stages of human subjects. National Institutes of Health Publication No. 204. U.S. Govt. Printing Office. Washington, D.C., 1968, 58 p. ROESSLER,R., COLLINS,F. and OSTMAN,R. A period analysis classification of sleep stages. Electroenceph. clin. Neurophysiol., 1970, 29: 358-362. S1NHA, A. K., SMYTHE,H., ZARCONE,V.P., BARCHAS,J. D. and DEMENT, W. C. Human sleep-electroencephalogram: a damped oscillatory phenomenon. J. theor. Biol., 1972, 35 : 387-393. WILLIAMS,R. L., KARACAN,I. and HURSCH,C. J. Electroencephalography of human sleep. Wiley. New York, 1974, 169 p.