Seventy participants aged between 18 and 40 years-old were recruited from the Guangdong Provincial Hospital of Chinese Medicine through posters from May 2016 to November 2017. All subjects were asked to complete a two-week sleep diary followed by a single all-night PSG recording in a sleep laboratory. Personal information was obtained from all subjects, including age, sex, race, place of residence, marital status, family history of insomnia and psychosis. Two self-reported questionnaires, the Pittsburgh Sleep Quality Index (PSQI) [25] and the Symptom Checklist 90 (SCL-90), were given to each participant. Subjective sleep quality was determined by self-reported TST after PSG. The subjects were asked two questions about their perceived sleep within 2 h after PSG completion: (1) “How long did you sleep last night?” and (2) “Did you sleep as usual?”. In this way, the sTST of the patient was obtained. For example, if the patient replied that they slept for 6 h during the previous night, 360 min was his/her sTST.

Subjects were categorized as GS according to the following criteria: (1) reported no difficulty in sleep according to the two-week sleep diary (i.e. sleep onset (SO) < 30 min, wake after sleep onset (WASO) < 40 min, TST between 6.0 and 8.0 h, or sleep efficiency (SE) ≥ 85%); (2) had a PSQI score < 7 [25], SE > 85% or TST > 6 h.

Participants were categorized as insomnia patients if they met the following criteria: (1) diagnosed with chronic insomnia disorder (International Classification of Sleep Disorders, 3rd edition); (2) reported at least three nights per week of sleep difficulty (i.e.SO > 30 min, WASO > 40 min, sTST < 6.0 h, or SE < 85%); (3) had a PSQI score of > 7; (4) had difficulty sleeping for more than 3 months; (5) did not have other medical, psychological, or sleeping disorders and did not take any medications that would affect sleep (e.g. sedative and hypnotic drugs, antidepressants, anti-schizophrenia drugs, etc.).

Insomnia patients were further categorized into two subgroups based on their SOD of TST. These two subgroups comprised patients with low mismatch (IWLM) and patients with high mismatch (IWHM). The SOD of TST was operationalized as the values of the differences between subjective and objective measures (i.e. sTST–oTST value) [8]. IWLM patients were those individuals who met the criteria of chronic insomnia disorder and had an SOD < 60 min in TST. IWHM were defined as patients who met the criteria of chronic insomnia disorder and had normal PSG parameters (i.e. SE > 85% and TST > 6.5 h) and SOD > 120 min in TST. In both the IWHM and IWLM subgroups, patients were excluded if they met one of the following criteria: (1) diagnosed with another Axis I disorder or any other sleeping disorder (e.g. idiopathic insomnia, sleep apnea, which was defined as an apnea–hypopnea index of more than five events per hour using PSG, or restless leg syndrome); (2) affected by other external factors that might affect insomnia (e.g. physical pain caused by medical diseases, drugs affecting sleeping structure, alcohol consumption, other treatments, etc.); (3) go to sleep later than 0:00 am or wake up before 6:00 am, or had irregular sleeping schedules.

Based on the inclusion and exclusion criteria, 47 participants were included in the study: GS group (n = 10; 5 males, 5 females), IWHM group (n = 18; 9 males, 9 females), and IWLM group (n = 19; 3 males, 16 females).

The PSQI is a questionnaire consisting of 21 items and has been commonly used to evaluate subjective sleep quality. The higher the score, the greater the severity of insomnia. A score > 7 indicates abnormal sleeping (severe difficulty in at least two areas or moderate difficulty in more than three areas).

The SCL-90 is one of the most widely used mental health scales in the field of psychiatry. It is a 90-item, self-reported symptom inventory. The score for each item is summed, yielding a total score that covers ten aspects. The higher the total score, the greater the risk of developing psychological distress [26].

In conventional PSG (Nicolet, ONE, EEG 32, USA), the international 10–20 system was used to record EEG. In this study, the grounding electrode was placed on the frontal pole midline point and the bilateral ear electrodes were used as the reference. All electrographic electrodes were placed according to the AASM 2.6 recommended guidelines. The impedance was kept below 5 kΩ for all electrodes. The surface electrodes included six EEG (two central electrodes [C3, C4], two frontal EEG electrodes [F3, F4], and two occipital EEG electrode [O1, O2)]), two electro-oculogram (E1, E2), submental electromyogram (EMG: Chin1-Chin2), electrocardiogram (ECG), and two reference electrodes (A1, A2). In addition, tibialis EMG and respiration were used to exclude periodic limb movements (a PLMSI > 15) and sleep apnea (an apnea–hypopnea index > 5), respectively. Participants were asked to sleep at their usual time (before 0:00 am) and wake up at 7:00 am. The sampling rate of EEG was 500 Hz and the filter settings were as follows: notch frequency at 60 Hz; low pass filter at 35 Hz; high pass filter at 0.3 Hz.

Sleep records were reviewed and scored by a registered PSG technician according to the revised AASM 2.5 sleeping scoring criteria [27]. The sleeping continuity parameters, including TST, SPT, SE (ratio of TST to time in bed × 100%), and SOL, and sleeping architecture parameters, including the number of awakenings, the number of arousals, arousal index, percentage of NREM stage 1 and 2, slow wave sleep (SWS) or NREM stage 3, and REM sleep of TST were analyzed.

Normal sleep time is 6.0 to 8.0 h, and therefore we analyzed the first 6 h of the PSG recordings. The data from the central and frontal EEG electrode (averaged C3-A2 and C4-A1 channels, averaged F3-A2 and F4-A1 channels) were generated using software of Nicolet EEG band width tools.

Most of the common artifacts were due to improper click placements (such as electrode popping, ECG or pulse artifact), body movement (muscle artifact, eye movement artifact or major body movement) or environmental factors (overheated which lead to slow-frequency artifacts). We optimized the mastoid electrodes so that ECG and pulse artifacts could be minimized. Secondly, we kept impedance below 5 kΩ to avoid electrode popping. At the same time, we maintained a temperature of 20 °C in the sleep laboratory which is the standard setting to ensure that the subjects completed the test in a comfortable environment, and avoided the influence of slow-frequency artifacts from sweat. A notch filter at 50 Hz was applied to avoid power line contamination of the electrical signals. Then, we set a high frequency filter to 35 Hz to reduce most of the interference from EMG. We chose this cutoff values as the frequency of EMG activity signal is generally contained in higher frequency bands and since the AASM recommend that EMG low frequency and high frequency filter cutoffs should be at 10 Hz and 100 Hz, respectively, to capture the muscle activity. Finally, the data fragments that were displaced or cut off due to movements or that were obviously different from the background were to excluded by visual inspection (e.g. due to the excessive loss of occipital EEG electrode signal, these data were not included in this study). Therefore, artifacts in each recording were visually inspected and removed accordingly.

The beta (16–32 Hz), sigma (12–16 Hz), alpha (8–12 Hz), delta (0.5–4 Hz), and theta (4–8 Hz) band activity was extracted for PSA analysis. The values of relative spectral power were calculated by dividing the absolute power of each frequency band by the power of the total power spectrum.

Statistical analysis was performed using the SPSS software (ver. 24.0) and with an unpaired two-tailed test of significance. A normality test and Levene’s test were used to check whether the data followed a normal distribution. A Chi-square test was used for demographic characteristics except for age. Normally distributed data with homogeneous variance were compared using a one-way ANOVA, while others were compared using a non-parametric analysis (Kruskal–wallis) with post-hoc analysis. The statistical value H represents the use of non-parametric analysis, while the statistical value F represents the use of a one-way ANOVA. In addition, we used pairwise least significant difference post-hoc tests after a one-way ANOVAs and a Bonferroni correction for multiple comparison after a Kruskal–Wallis test. Spearman’s or Pearson’s correlation analysis was used to determine the correlation between the EEG spectral power (absolute and relative) and the SOD of TST (after data normality was confirmed). A P-value < 0.05 was considered statistically significant.