1. 1 Methods for detection of hidden changes in the EEG H. Hinrikus*, M.Bachmann*, J.Kalda**, M.Säkki**, J.Lass*, R.Tomson* *Biomedical Engineering Center **Institute of Cybernetics Tallinn University of Technology COST B27 Meeting, Swansea, September 16-18, 2006 Conjunct COST B27 and SAN Scientific Meeting, Swansea, UK, 16-18 September 2006

2. 2 Introduction Qualitative analysis of the changes in dynamics of the EEG is complicated due to the irregular nature of the EEG signal. It is difficult to detect small variations in the EEG signals on the background of their high natural variability. The influence of a weak stressor, such as a mental task or low-level microwave radiation, on the EEG activity, is usually very small. Our previous attempts on detection of the effect of microwave radiation on human EEG showed that some traditional methods of the EEG analysis such as quantitative EEG, bispectrum or fractal dimension, did not provide reliable distinction of small changes in the EEG caused by the microwave radiation

3. 3 Introduction (2) Example of the recorded EEG signal (10 s) without and with (after marker) microwave

4. 4 Aim of the study The aim of this study is to select and evaluate methods sensitive to reveal small hidden changes in the EEG signal. Two original methods were considered: the length distribution of low variability periods (LDLVP) the integration of differences in energy of the EEG Both methods were evaluated on the same EEG database recorded in conditions of the exposure to microwave radiation.

5. 5 Database Subjects A group of volunteers, 15 persons (aged 21-24), 8 male and 7 female. Microwave Exposure 450 MHz microwave radiation, 1W output power 100% amplitude modulated at 40 Hz frequency (duty cycle 50%) field power density at the skin 0.16 mW/cm 2 Recording equipment Cadwell Easy II EEG 19 electrodes, EEG channels FP1, FP2, P3, P4, T3, T4, O1, O2 Sampling frequency 400 Hz.

6. 6 Database: Recording protocol Firstly, the reference EEG was recorded over 5 minute. Secondly, modulated with 40 Hz microwave was applied. The second protocol for the sham-exposure included the same steps, except that the microwave generator was switched off.

7. 7 Analysis of the EEG based on the LDLVP method (1) Initially, all the EEG recordings were divided into two sub-signals: the first sub-signal contained all 1 min periods without microwave exposure (all the odd minutes from the initial EEG recording), the second sub-signal contained all minutes with microwave exposure (all even minutes of the initial EEG recording) The scaling analysis utilizing LDLVP method was applied for two sub-signals.

8. 8 Analysis of the EEG based on the LDLVP method (2) The LDLVP analysis consists of several steps: Firstly, we define the local average of the signal in time-window T Secondly, we define the local variability as the deviation of the current value of the signal from the local average. Thirdly, the low-variability periods are defined as continuous intervals with

9. 9 Analysis of the EEG based on the LDLVP method (3) Finally, the number of low-variability periods N exceeding the length T 0 is plotted versus the length T 0 The weighted area under the curve of the function T 0 = T 0 (N) was selected as the non-linear quantitative measure

10. 10 Analysis of the EEG based on the integration of differences method (1) Initially, the signals bands of four basic EEG rhythm frequencies theta (4 – 7 Hz), alpha (8 – 13 Hz), beta1 (15 – 20 Hz) beta2 (22 – 38 Hz), were extracted from the total EEG signal (0.5 – 48 Hz) by filtering. The EEG analysis was performed separately for each EEG rhythms frequency bands

11. 11 Analysis of the EEG based on the integration of differences method (2) The method of integration of differences consists of several steps. Firstly, the average energy of the signal inside a selected comparison segment in a time-window T is calculated as Secondly, relative differences of the average energy are calculated for every cycle f: where s 1f and s 2f are the average energies in a comparison segment without and with stimulation respectively

12. 12 Analysis of the EEG based on the integration of differences method (3) Integration of the differences over ten cycles of exposure for a subject n is applied and characteristic parameter S n is calculated as The first 30 s intervals of 60 s recording half-cycles with and without exposure are selected as the signal segments for comparison. For sham recordings the same parameters are calculated for comparison segments as even and odd minutes of the recordings.

13. 13 Results: the LDLVP method (1) The number of low-variability periods N exceeding the length T 0 for a significant subject: (a)  second sub-signal of exposed recording (intervals with microwave); (b)  first sub-signal of exposed recording (intervals without microwave).

14. 14 Results: the integration of differences method (1) The relative average changes in % (S – parameter) of the EEG rhythms total energy of the recording segments with and without microwave exposure in P – channels for the whole group.

15. 15 Results: the integration of differences method (2) The relative changes (S – parameter in %) of the EEG rhythms total energy in P – channels for a subject

16. 16 Results: the LDLVP method (2) Table 1. The ratio of the computed power difference to the standard deviation of differences and p-values as a result of Bonferroni correction in P-channels

17. 17 Results: the integration of differences method (3) Table 2. Calculated p-values for different EEG rhythms in P-channels as a result of analysis by post – hoc Bonferroni correction

18. 18 Discussion The LDLVP analysis presents the outcome, resulting in significant results for 4 subjects (Table 1). The S-parameter measures exceeded the limit of significant deviation in beta2 frequency band for 2 subjects (Table 2) cases. From those results it is difficult to conclude whether the effect appear rather in intensity (S-parameter) or time variability (LDLVP) of the EEG signals and which measure, LDLVP or S-parameter, is more effective.

19. 19 Conclusion The results show that small changes in the EEG signals hidden in visual inspections can be detected by the LDLVP and integration of differences methods. The modulated with 40 Hz 450 MHz microwave exposure caused statistically significant changes in the time variability (26.7% of the subjects) and energy (13.3% of the subjects) of the EEG signals.