-Amid ......... DATA CTION USE OF LINC-8 FOR EEG REDU i/6-7/67 USE OF LINC-8 FOP, EF-G DATA RZDUCTION Baci<; @round Interpretation of EEG Waveforms for information relating to subject state has been an area of long@ interest to people outside L' curophysiological-field. A dozen years ago., Norbert o the r W-.ener cc@r@3ucte-d i-avestigations of brain records. He sought to apply to E-EG interpretation, some of the statistical communica. tio-,i theory developed by his circle at MIT. This tleory has been I set forth by Y. W. -Iee. In 1954 Dr. Weiner was particularly interesteel in "a nominal 10 cps clock pulse" which he hypothe- 2 sized. formed the basic timing for serial rd-order-:ng of the brain processing organization, he noted state differences associated with this part of the spectrum in different subjects. Deta ilgd specific,_statis.t-ical reduction techn-ques are laid out by by Dr. 2 he former-attacks reduction in a classical statistical appro dealing wit@-i power spectral estimates, phase functions, auto- covariance and @cross covariance functions, transfer functions, ccrerenc%r, etc, He uses these descri,)tors to determine signifi- cant factor!; associated with EEG responses. A large number of 2 variables can be treated sirnultaneouily. The latte'r attacks the problem principally from a "period analysis" standppint. The basic waveform is clipped and zero- ciossover inter@Vals are scr,,,tinized. Similarly, the first and second der'Lvatives are processed for clipped periods. The inves- tigator utilizes all three of these on t2he same time base to ex=ine the sp4ictrum for state evaluation. The references also give summaries of other recent tech- niques applied to EEG analysis. It is intended here to preseat the results of a modest study car7ie,i out on @k set of data furnished through the corrtesy of Drs. or educationa8l purposes. The subject was recorded, in three states, on 1/411, 2 track, FM tape run at 7 1/2"Isec as follows: (al) Drowsy S m in. (a-2) I kcps tone (bl) Alpha S min. (b-2) 13 (c 1) Light sleep and Alpha (C-22) Editing and Reduction The tape was played, into the A/D che-anels of the LINC - 8 for editinv, using the SCOPE-8 program. The three 5 minu'te samples were examined for gross character-.stics. Comments were as ;ollows: (a) Drowsy 30-60 secon2d arnplitude (long term) modulation was noted to be present in addition to the 7-10 cps dominant component band. (b) Aloha: more extreme excursion; less pro- nounced long term modulation; pronounced short term modulation or beating. (c) Lig2ht Sleep: less pronounced excursions. The data was visually edited again and a run made to store the data. This is given in the colmniilabeued Run I of Table 1. It was -at-tempted to recover representative samples of high, into-r- mediate and low signal power. Block storage numbers are shown. 2 Next tl-.; stored data was examined for spectral content using the proaram FROANA; scope camera pictures -wire taken for com- 0 parison in matrix form for significant differences in the v-arious states. Reloeats of data taking was next performed in Runs TI and 2 Ill. R-.ln III was sampled at a lower rate; this gives a longer sample interval. Next, selected blocks were differentiated once and stored in ILzations corresponding to start point 100 000 waveform s ample, I 0 0 ii r s t de r ivative, e tc. -3 It w@(S feasible to make certain preliminary evaluations of the data at this point. These were, 1'. Signal content for discrirnina-tion between states was highest for the high points of the long term mbdulation. 2 ncy resolution was (I/Z cps) more 1. A f ine r f reque relevant than tl'--- I cps resOlution- c made for oscillosc-ope pictures Finally a set of runs wer to sbo.,v the below format: 2 1. sampled waveforra 2. frequency analysis of (1) ve of (1) 3. first derivati 4. frequency analysis of ,d Conclusions Figures 1, 21 and 3 show thr2ee reprerentative outputs whicl,. appeared tentatively as lity-pical": 1. ELr@ this is sample 060, 160 of the data, taken at a t@igh point. The waveforrn is shown at the top; shown out to its spectrum is 2immediately below 32 cps. A dominant spil-,e shows up at. 9 cps; a plateau frorn 8-13 cps is present. A "characteristir-11 bump rhomed up in the, region 14-20 cps. AlPba: this is sample 062. 162; a red-action of the hump at 14-20 cps was noted. k bifarcation or "forking" appeared in the alpha region at the depression zt 9 cps. This would explain the bea--ing or short terrn rnodulatiol. this is sample 072, 172; 3. DI 2 a general predominance of the alpha tones is noted with shift in energy to the lower end of the spectrum. 'hows For all three sample$ the first derivative spectrum s -up as expected; namely, the "bluing" of the spectr4um by applying a derivative function. It would appear that a. parametric stud-I of BW, frequency, accenting functions (S. S2 etc* examination of phase relation- -4- nd ot ships, reducflon techniqueso a her processing would be meaningiul, "his rnight permit the design of j. real time EEG analyzer which would not require a high level of training for operation. REFERENCES MIT Report i 181, Y. W. Lee 2. 3. Li qt th CO co ti A. C,P Al 5 4.P ea fA ke 40% Romeo t ce 40 AGh