| Sleep is an important physiological phenomenon and a necessary physiological process. Adequate sleep have certain influences on the brain information processing, memory enhancement, body development and emotional fluctuation. Polysomnography(PSG) is a technique to explore activity of brain. It acquires brain waves by electrodes, and real-time monitors the brain activity, so that to help people understand the process of sleep. PSG has become the gold standard for monitoring sleep, but its experiment need high environmental requirements and high expense. Besides, it disrupts the sleep of subjects. So it can not be applied to large-scale population and day to day to collect EEG data continuously.With the development of wearable device, the application of actigraphy(ACT) to monitor the quality of sleep has been very common. Because of human sleep habit, there is the characteristic of the wrist to remain immobility in sleep state. ACT collects wrist activity intensity data with applying this characteristic to judge this epoch is sleeping or awake. ACT is easy to wear and has a low price. It solves the polysomnographic technology limitations.ACT is not based on the gold standard for monitoring sleep, and it has many brands,which have great difference among monitoring data. Then how is the performance of it to monitor sleep? As used in this paper, the ACT is a Philips product(Actiwatch 2,wristwatch). The device of PSG is an Easy III product. The experiment lasts for more than six months. Finally, we collect 51 subjects data of PSG and ACT successfully. ACT uses the Epoch-by-Epoch analysis with epoch length of 15 seconds and 30 seconds. It applies the weighted value of each epoch to compute the total activity, then compares to the corresponding threshold to determine whether it is the sleep(stored as 0) or awake(stored as 1). PSG data is divided into a 30 s time intervals as one sleep stage. Each epoch of ACT compare with PSG sleep state at the same time, to calculate performance parameters of ACT, included sensitivity(Sens=0.96), specificity(Spec=0.24), and accuracy(Acc=0.88). Sensitivity refers to the proportion of the numbers of ACT and PSG are sleep at the same time account for the total numbers of sleep state in the PSG. Specificity refers to the proportion of the numbers of ACT and PSG are wake state at the same time account for the total numbers of wake state in the PSG. Accuracy refers to the proportion of the numbers of ACT and PSG are sleep state and wake state at the same time account for the overall numbers in the PSG. The capacity of monitoring about ACT is very weak during wake-sleep-wake transition. This paper discusses the data of two hours before going to bed and two hours before awake. We find that the accuracy of two hours before bedtime is very low(Acc = 0.75), and low specificity(Spec = 0.15) of two hours before waking. Since the impact on sleep with a PSG equipment, this paper analyzes the difference between the data of other days wearing the ACT and the experiment day. In the whole, sleep quality of all subjects is not effected by the PSG equipment. The experiment ACT threshold set has a variety of options, this paper analyzes five thresholds, included 5,10,20,40 and auto threshold, and finds the threshold of 10 is more suitable for monitoring the subjects who are poor sleep quality. In total sleep time(TST), sleep latency(SOL), sleep efficiency(SE), after sleep awake(WASO) and other sleep parameters, the threshold value of 40 only has a significant difference(P = 0.001) when compares to PSG on total sleep time, and the ACT excessively estimates 38.4 minutes of total sleep time, other thresholds have two parameters with significant differences, On the overall level, the threshold value of 40 is more suitable for monitoring the healthy subjects. Then we discuss the influence of some parameters, such as body mass index(BMI), age of the subjects, intelligence, memory, gender and other factors on ACT monitoring sleep parameters and performance parameters. Finally, we draw a conclusion, changes in age have a correlation with total sleep time and sleep efficiency(PTST=0.001, r TST=-0.46; PSE=0.023, r SE=-0.32). With the increase of age, the ability of sleep monitoring of ACT is abate; the change of BMI, intelligence and memory have no correlation with ACT monitoring sleep; about gender, the male and female subjects have significant difference(P = 0.02) in the monitoring TST and female subjects is weaker than male subjects. In terms of performance parameters, ACT monitoring results have no correlation with age, BMI, intelligence, memory; there is also no significant difference in gender. During the experiment, Two subjects catch a cold, this paper studys the result of ACT monitoring two subjects particularly and finds that a great impact of cold on the ACT. When catch a cold, there is not accurate for monitoring sleep. Then according to the behaviors during sleep, such as more body motion, snoring, watching mobile phones, we analyze the impact of each behavior on ACT monitoring sleep, and find playing mobile phone has a significant effect on the accuracy(P = 0.003); and body motion has a significant effect on the variance(P = 0.003) and the mean(P = 0.000) of sensitivity, the mean of specificity(P = 0.000), the variance(P = 0.035) and the mean(P = 0.006) of accuracy. On sleep parameters, the variance of SOL(P = 0.028) and SE(P =0.006) also have a significant impact; Snoring has no significant effect on the experiment. Finally, according to the subjects’ sleep diary, we discuss the impact of a nap, bedtime pressure, and excessive exercise in the day time on ACT monitoring sleep. A nap has a significant impact on the the variance of sensitivity(P = 0.045), the variance(P = 0.001) and the mean(P = 0.008) of accuracy; bedtime stress and excessive exercise in the daytime have no significant influences.Summing up the above, ACT has a low specificity and an excessive total sleep time estimate, but it still can be applied to healthy subjects. Those adult subjects have a better effect on monitoring who have less body motion during sleep,not play mobile phone before sleep, and not have a nap. |
PDF Full Text Request