Experiment output

A total of three subjects wore the two subjects system at the same time, and their physiological signals were simultaneously measured during sleep. After the test, we imported the EDF file to Alice Sleepware and the clinician was asked to scoring sleep stage. Fig. 5-11 to Fig. 5-15 showed the variation of subject’s bio-signals during difference sleep stage. According to the 2007 AASM standards, there were listed five different stages of sleep, following by Stage W (Wakefulness), Stage N1 (NREM1), Stage N2 (NREM2), Stage N3 (NREM3) and Stage R (REM). After completing the sleep experiment, a 'scorer' would analyze these data by reviewing 30-second epochs to make up a hypnogram for overnight sleep and to summarize sleep structure. The top of Alice Sleepware showed sleep stage, and 30-second physiological signals were shown in the main window. These physiological signals listed from top to bottom respectively were EOG-left, EOG-right, EEG(C4-M1), EEG(O2-M1), EMG, Airflow and ECG..

Fig. 5- 11: Stage W: note the eye movements with high chin tone, 30-s epoch

Fig. 5- 12: Stage N1: SEMs are seen with gradual drop out of alpha,30-s epoch

Fig. 5- 13: Stage N2:K complexes and sleep spindles seen, 30-s epoch

Fig. 5- 14: Stage N3: Delta slow waves are seen, 30-s epoch

Fig. 5- 15: Stage R: Note saw tooth waves with REMS

In statistics, correlation (often measured as a correlation coefficient, ρ) indicates the strength and direction of a linear relationship between two random variables. In general statistical usage, correlation or co-relation refers to the departure of two random variables from independence. In this broad sense there are several coefficients, measuring the degree of correlation, adapted to the nature of the data.

The measure of linear association between i and j shows inequation 5-1.

(5-1)

where is the mathematical expectation and .

[R, P]=corrcoef (...) also returns P, a matrix of p-values for testing the hypothesis of no correlation. Each p-value is the probability of getting a correlation as large as the observed value by random chance, when the true correlation is zero. If P (i, j) is

Small, say less than 0.05, then the correlation R (i, j) is significant, Fig. 5-16 shows Correlation examples.

In order to verify the validity of bio-signals obtained by our PSG system, we random selected 30-second raw physiological signals obtained by our PSG system and Alice 5 Diagnostic Sleep system, and compared to each other. The two sets of physiological signals looked very similar, and owned the same obvious features.

Therefore, a more quantitative comparison was then performed by using cross correlation and correlation coefficients function in MATLAB to obtain the linear correlation of the two sets of physiological signals. Fig. 5-17 to Fig. 5-23 showed the comparison of 30-second raw physiological signal data in time domain and their correlation in every 1 second. From the above results, we found that physiological signals obtained by our PSG system and the reference system in the time domain were highly similar. Therefore, our PSG system can be viewed to own a high level of reliability.

Fig. 5- 17: Comparison of 30-second raw data (EEG C4-M1), and their correlation in every 1 second

Fig. 5- 18: Comparison of 30-second raw data (EEG O2-M1), and their correlation in every 1 second

Fig. 5- 19: Comparison of 30-second raw data (EOG-Left), and their correlation in

Fig. 5- 20: Comparison of 30-second raw data (EOG-Right), and their correlation in every 1 second

Fig. 5- 21: Comparison of raw data (ECG), and their correlation in every 1 second

Fig. 5- 22: Comparison of raw data (EMG-Right)

Fig. 5- 23: Comparison of raw data (Nasal airflow)

The hypnogram built on the two sets of records were shown from Fig. 5-24 to Fig. 5-26. We found that the night patterns evaluated by the expert clinician for the two sets of data were similar. Comprehensive view of subjects 1 and 3 at the beginning of recorded time were different, but this did not affect the trend of the whole sleep architecture. The major difference in interpretation between the reference system and our PSG system occurred in the case of subject2. This is because of that the electrode lead fell off at 5 o’clock. For the case of subject 3, the sleep stage interpretation of our PSG system was the most similar to that of the reference system.

Fig. 5- 24: Hypnograms of the two sets of records (subject 1)

Fig. 5- 25: Hypnograms of the two sets of records (subject 2)

Fig. 5- 26: Hypnograms of the two sets of records (subject 3)

We rejected the segment of subject 1 and subject 2 records obtained by the

segment of subject 2 records obtained by our PSG system, which was after the occurrence of falling off electrode lead. Then, we observed the time duration of each sleep stage which cumulated sleep hours for three subjects (Table 9), and the percentages of each sleep stage over night in two systems (Fig. 5-27). In the view of above, we found that the duration of Stage W, Stage N1 and Stage REM in two systems had similar interpretation, but had a gap between Stage N2 and Stage N3.

According to the explanation of clinical expert, the proportions of the combining duration of Stage N2 and Stage N3 to the whole duration for two systems were the same. Thus, here Stage N2 may be interpreted as Stage N3 if some segments of physiology signals were extremely similar. And this caused the difference of interpretation between the two systems for Stage N2 and Stage N3.

Fig. 5- 27: Percentages of each sleep stage over night

Table 11: Relative errors on times spent in each sleep stage

Sleep stage Ref. Our

proposed system

Rel. Err.

Awakening 156.5 150.5 3%

REM 90 93.5 3%

Stage N1 52.5 56.5 7%

Stage N2 337.5 372.5 9%

Stage N3 194.5 128 51%