Phase-Amplitude Coupling in Theta and Beta Bands: A Potential Electrophysiological Marker for Obstructive Sleep Apnea

Introduction

Obstructive sleep apnea (OSA) a prevalent sleep-related respiratory disorder, affects a significant portion of the adult population.^1,2 Characterized by recurrent upper airway obstruction (apnea) during sleep, OSA lead to sleep fragmentation,^3,4 intermittent hypoxia, snoring, and excessive daytime somnolence.⁵ Long-term chronic hypoxia associated with OSA can induce central nervous system dysfunction, manifesting as cognitive impairment and behavioral abnormalities.^6–9 Increasing evidence implicates OSA in the etiology of neuropsychiatric disorders and cardiovascular disease, underscoring its far-reaching health implications.^10–13 Sleep electroencephalography (EEG) has been utilized as a valuable tool for monitoring cerebral electrical activity in OSA patients,^14–16 Our preliminary study has identified abnormal sleep EEG spectral patterns in these individuals, characterized by reduced power in low-frequency theta and alpha bands, and high-frequency beta oscillations, correlating with indicators of chronic hypoxia.¹⁴ These results shed light on the effects of OSA on cortical activity from an electrophysiological perspective. However, EEG researches of OSA have predominantly focused on spectral analysis, offering scant insights into the modulation of global brain network dynamics. The modulation of neural oscillations across high and low frequency EEG bands in OSA patients and their potential as electrophysiological markers of OSA severity, remains to be elucidated.

This study investigates the characteristics of neural oscillatory coupling in the frontal and occipital cortices of OSA patients across severity levels. We hypothesize that phase-amplitude coupling (PAC) values will vary among OSA patients, potentially correlating with disease severity and chronic hypoxia. By analyzing the modulation index of various frequency combinations (1–45 Hz) using three distinct calculation methods, we aim to elucidate the long-term consequences of OSA-related oxygen deprivation on central nervous system function.

Sleep is characterized by dynamic fluctuations in information exchange and neural oscillation connectivity across distinct cerebral regions. These oscillations exhibit intricate mutual influences and regulatory functions. Cross-frequency coupling, particularly phase-amplitude coupling (PAC),¹⁷ exemplifies this phenomenon, reflecting the synchronization between low-frequency phases and high-frequency amplitudes. This coupling is intimately linked to neural plasticity, cognitive learning, memory formation, and sensory mechanisms.^18–20 Research has elucidated that the modulation index (MI) provides a more robust assessment of signal-to-noise ratio, amplitude independence, multimodal sensitivity, and modulation width.²¹ Sleep EEG analyses in healthy adults have demonstrated that PAC exhibits stage-dependent flow directions, with low-frequency delta-theta/alpha bands PAC progressively intensifies concomitant with deepening of sleep.²² Consequently, PAC is regarded as an effective biomarker for the modulation of low-frequency activity and the synchronization within high-frequency bands. Exploring neurophysiological connectivity patterns and oscillations dynamics could enhance our understanding of the pathophysiological mechanisms underlying OSA.²³

Cross-frequency coupling may elucidate the network patterns of different frequency bands in OSA patients. Studies have shown that sleep disturbances may affect cortical information transmission based on cross-frequency PAC.²⁴ EEG studies have identified distinct characteristics of low and high-frequency oscillations in OSA, indicating engagement of cortical activity in chronic hypoxia. PAC features have shown remarkable sensitivity to various sleep-related disorders.²⁵ Additionally, the combination of PAC with Random Forest classification models has demonstrated efficacy in distinguishing between seven sleep disorders.²¹ In OSA, reduced MI values for theta-gamma oscillations have been observed in the somatosensory-motor cortex across all sleep stages, while delta-alpha coupling demonstrates opposite changes during NREM and REM periods.²⁶

Our study aims to investigate the characteristics of neural oscillatory coupling in the frontal and occipital cortices of OSA patients with different severity levels, and to determine whether the inter-frequency coupling of neural oscillatory could serve as a sensitive indicator of OSA severity. We hypothesized that PAC values will vary with OSA severity, potentially reflecting a correlation with OSA severity and chronic hypoxia. By analyzing the modulation index of frequency combinations from 1 to 45 Hz using three distinct calculation methods (klmi, mvlmi, and plv), we seek to uncover novel electrophysiological biomarkers for OSA. This investigation promises to enhance our understanding of OSA’s underlying pathophysiology and potentially revolutionize the assessment of disease severity and progression. By exploring the intricate relationships between neural oscillations and chronic hypoxia, we aim to pave the way for more targeted interventions and improved patient outcomes in OSA management.

Materials and Methods

Study Population

This retrospective study was conducted at the Department of Neurology at the First Affiliated Hospital of Zhengzhou University in China. It included 521 patients with snoring symptoms and clinical suspicion of OSA from 2020 to 2023. To ensure the accuracy of the study, certain exclusion criteria were applied. Patients with other types of sleep disorders, such as central sleep apnea, rapid eye movement sleep behavior disorder, restless legs syndrome, and periodic limb movement disorder were excluded. Additionally, individuals who had taken medications known to affect sleep status (eg, hypnotics, benzodiazepines) within the past month were also excluded. Patients with significant artefacts in their EEG signals, as well as those with severe pulmonary, renal, hepatic, and cerebral disorders were excluded as well (Figure 1). Finally, clinical data and PSG studies of 103 subjects were selected for analysis. We considered apnea-hypopnea index (AHI) and oxygenation desaturation index (ODI) and the percentage of oxygen saturation less than 90% (SaO2<90%) in order to evaluate the severity of nighttime hypoxia. The subjects were divided into three groups based on previous study: mild OSA group (5 ≤ AHI < 15 events/hour, N=27), moderate OSA group (15 ≤ AH I< 30 events/hour, N=30), and severe OSA group (30≤AHI events/hour, N=46) (Figure 1 and Table 1).¹⁴ The study was conducted in accordance with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study protocol was approved by the Ethics Committee of the First Affiliated Hospital of Zhengzhou University, and the requirement for informed consent was waived (Ethics Approval No. 2022 [101]). We pledge to strictly adhere to patient data confidentiality and take all necessary measures to protect patient privacy.

Table 1 Characteristics of the Population (N = 103) Categorized by OSA Severity

Figure 1 Flow chart of patient screening and grouping, clinical data collection, sleep EEG processing and data analysis.

All Night Polysomnography

Sleep apnoea was confirmed by nocturnal polysomnography (PSG, Embla 4500S, Ambulatory Monitoring Inc., Ardsley, NY, U.S.A).²⁷ A total of 6 EEG signals (F3-M2, F4-M1, C1-M2, C2-M1, O1-M2, and O2-M1), 2 electro-ocular signals (E1-M2, E2-M2), and chin EMG (EMG1-EMG2, EMG EMG3), electrocardiogram (ECG), respiration (nasal pressure, airflow), oxygen saturation (SpO2), abdominal-thoracic movements and leg movements were recorded.²⁸ Due to limitations in the collection equipment and techniques, the central channel signals (C1/C2) were not included in the analysis. Sleep stages and respiratory events were analyzed based on the guidelines recommended by the AASM scoring criteria.²⁹ Sleep stages were categorized as non-rapid eye movement (NREM) sleep, which includes stages N1, N2, and N3, rapid eye movement (REM) sleep and wakefulness. Sleep-related parameters such as including bedtime, total sleep time (TST), sleep latency (SL), sleep efficiency (SE), wakefulness after sleep onset (WASO), and the ratio of each sleep stage (N1/TST, N2/TST, and N3/TST) to the total sleep time were calculated and reported in the PSG study (Table 1).

Data Analysis (Pre-Processing)

We recorded the scalp EEG signals of the sleep participants using non-invasive electrodes, as described in the overnight PSG method. EEG signals were recorded at a sampling rate of at 200 Hz and stored in EDF format. We selected EEG data from the first two sleep cycles for analysis by reviewing the entire night’s PSG data. Initially, we preprocessed 3–12 minutes of raw data from each stage of the sleep cycle (NREM, REM, and wakefulness). The EEG signals were preprocessed using MATLAB software (MATLAB 2013a). The raw signals were subjected to a bandpass filter of 0.5–45.0 Hz using the EEGLab software (eeglab14_0_0b Yueying Technology). The data underwent preprocessing through artifact removal, eliminating data with an amplitude of ± 80 mV, and utilized Independent Component Analysis (ICA) to remove interference from eye movements and electromyographic noise.^11,30 Finally, we chose 3 minutes of the preprocessed data from four channels (F3-M2, F4-M1, O1-M2 and O2-M1) for further analysis.

Phase-Amplitude Coupling Analysis

We analyzed modulation indices (MI) for various frequencies ranging from 1 to 45 Hz using three different computation methods (the Kullback–Leibler based modulation index, klmi; the mean vector length modulation index, mvlmi; the Phase-Locking Value, plv) with the FieldTrip toolbox. In our initial step, we utilized the EEGfilt function within EEGLAB to extract low- and high-frequency signals. Subsequently, by employing a Hilbert transform with a 2 Hz increment, we acquired the instantaneous amplitudes of signals in the high-frequency range (7 to 45 Hz) and the instantaneous phases of signals in the low-frequency range (1 to 29 Hz). The phase angles of the low-frequency signals were segmented into eighteen bins, each spanning 20 degrees from -π to π, and the corresponding mean amplitudes of the high-frequency signals for each bin were calculated. To normalize the mean amplitude within each bin, we divided the value of each bin by the sum of the values across all 18 bins, yielding a phase-amplitude distribution. MI values were calculated for each frequency pair.³¹ The MI values, ranging from 0 to 1,³² with an MI of 1 indicating maximal phase-amplitude coupling, and an MI of 0 indicating no coupling. To visualize represent our findings, we plotted co-modulation plots of MI values for each channel across different sleep stages in OSA patients. In this study, we performed statistical analysis on the MI values for the low-frequency theta (5–7 Hz) and high-frequency beta (25–29 Hz) band of interest, and further analyzed the correlation between theta-beta MI values and hypoxia indices in PSG metrics (see flowchart in Figure 1, Step 4).

Statistical Analysis

GraphPad Prism (version 8.3.0) and SPSS (version 26.0) software were used for statistical analysis and plotting. Data were expressed as mean ± standard deviation, unless otherwise as number (percentages in parentheses). MI statistical variables of the three OSA groups were compared using one-way ANOVA. Correlation analyses were performed using Pearson’s two-tailed t-test. A significance level of p < 0.05 was considered statistically significant.

Results

Demographic and PSG Characteristics of OSA Patients

An initial cohort of 521 patients with clinical suspicion of OSA was identified. Ultimately, 103 patients who were clinically diagnosed with OSA were included in the study. For details of the selection process, please refer to Figure 1. The demographic and PSG characteristics of the three severity groups of OSA were summarized in Table 1. The median age of the patients was 55 years. In terms of demographics, we found a significant difference in body mass index (BMI) between the severe OSA group and the other two groups (F (2,81) = 5.508, P=0.0057), which is consistent with previous studies. For PSG-related indices, patients with severe OSA exhibited increased apnea-hypopnea index AHI during both NREM sleep (F (2,100) = 128.0, P<0.0001) and REM sleep (F (2,100) = 36.67, P<0.0001). Moreover, apnea and hypoventilation during NREM (F (2,100) = 48.26, P<0.0001) and REM sleep (F (2,100) =17.66, P<0.0001), microarousal indices during NREM (F (2,100) = 9.760, P<0.0001) and REM sleep (F (2,100) =6.285, P=0.0027), ODI (F (2,100) = 42.57, P<0.0001) and Sa02<90% (F (2,100) = 18.38, P<0.0001) were significantly higher in patients with severe OSA (Table 1). However, parameters such as bedtime, total sleep time, REM sleep latency, sleep efficiency and wake after sleep onset (WASO) did not differ significantly among the three groups (Table 1). Additionally, except for a decrease in N2 sleep time in severe OSA patients (F (2,100) = 4.399, P = 0.0148), sleep time in other stages did not change.

Phase-Amplitude Coupling Analysis

The results of modulation index (MI) analysis for frequencies ranging from 1 to 45 Hz using three different computational methods (klmi, mvlmi, and plv) were found to be similar. In this study, we present the results obtained using klmi. By analyzing the modulation index of phase at 1–29 Hz and amplitude at 7–45 Hz during each period in the frontal and occipital lobes, it was found that compared to mild OSA patients, moderate and severe OSA patients showed a trend towards higher of MI in the low-frequency (5–7Hz) and high-frequency (25–29Hz) in all sleep periods (Figure 2–4). Figure 2 shown the variations of PAC-MI for the F3/F4 channels across the mild OSA group, moderate OSA group, and severe OSA group during the N1 (Figure 2A), N2 (Figure 2B), N3 (Figure 2C) sleep stages. The variations of PAC-MI for the O1/O3 channels across the N1-N3 stages were shown in Figure 3. Figure 4 illustrates the increase in PAC-MI for the F3/F4 and O1/O2 channels in OSA patients between the REM and wakefulness states. The MI values for the F3/F4 and O2 channels demonstrate a statistically significant increase only in the N1 (F3: F(2, 97) =2.131, P=0.028; F4: F(2, 97)=4.437, P=0.022) and wakefulness periods (F3: F(2, 97)=1.073, P=0.467; F4: F(2, 97)=3.633, P=0.0470; O2: F(2, 99) =3.110, P=0.029; F(2,99) =3.521, P=0.035), as shown in Figure 5.

Figure 2 Comodulograms of signals with PAC-MI values in the frontal cortex in OSA patients during the NREM sleep. The modulation index (MI) of amplitudes in the high-frequency range (7–45 Hz) and the phases of signals in the low-frequency range (1–29 Hz) of F3/F4 channels across the mild OSA, moderate OSA, and severe OSA groups during NREM stages N1 (A), N2 (B), and N3 (C). The PAC-MI of the low-frequency (5–7 Hz) and high-frequency (25–29 Hz) was significantly increase in the N1 stage.

Figure 3 Comodulograms of signals with PAC-MI values in the occipital cortex in OSA patients during the NREM. The modulation index (MI) of amplitudes in the high-frequency range (7–45 Hz) and the phases of signals in the low-frequency range (1–29 Hz) of O1/O2 channels across the mild OSA, moderate OSA, and severe OSA groups during NREM stages N1 (A), N2 (B), and N3 (C). The PAC-MI of the low-frequency (5–7 Hz) and high-frequency (25–29 Hz) was also significantly increase in the N1 stage.

Figure 4 Comodulograms of signals with PAC-MI values in the frontal and occipital cortex in OSA patients during the REM sleep and wakefulness. The modulation index (MI) of high-frequency amplitudes (7–45 Hz) and low-frequency phases (1–29 Hz) for the F3/F4 and O1/O2 channels across the mild OSA, moderate OSA, and severe OSA groups showed no statistically significant differences during REM sleep (A); however, it increased significantly during wakefulness (B).

Figure 5 Histogram of the theta-beta PAC-MI in the whole sleep stages in OSA patients. (A and B) Comprise with mild OSA and moderate OSA group, the theta-beta PAC-MI of F3/F4 channel in the severe OSA group was significantly increase in NRME N1 and wakefulness. (C) The theta-beta PAC-MI of O1 channel in the whole sleep states had no statistically significant differences in the three OSA groups. (D) In O2 channel, the theta-beta PAC-MI in the severe OSA group was significantly increase only in the wakefulness period. Star symbols represent statistical significance levels: * represents p < 0.05, no stars represent p > 0.05.

Correlation Analysis

We further analyzed the correlation between the theta-beta bands MI values during N1 and wakefulness periods and clinical and PSG indicators (Figure 6). Statistical analysis revealed no correlation between gender and MI values. BMI showed a positive correlation with the N1 sleep of F4 channel (r=0.514 P<0.001), the wake sleep of F4 channel (r=0.453 P<0.001) and the wakefulness period of O2 channel (r=0.346, P=0.001). The AHI exhibited a positive correlation with MI value during N1 sleep in the F4 channel (r=0.230, P=0.022), wake period in the F4 channel (r=0.249, P=0.013) and wakefulness in the O2 channel (r=0.235, P=0.019), indicating a relationship between PAC and the severity of sleep-disordered breathing, with higher PAC-MI values associated with more severe conditions. ODI was positively correlated with MI during N1 sleep in the F3 channel (r=0.216, P=0.031), while SaO2<90% was positively correlated with MI values during wake period in the O2 channel (r=0.222, P=0.025), suggesting that PAC could reflect the degree of chronic hypoxia in OSA patients. We categorized the three groups based on AHI while focusing on the correlation between AHI and PAC-MI. The results showed that there was a positive correlation between the MI value of F3 channel and wakefulness only in severe OSA patients (r=0.321, P=0.031), and there was no statistical significance in other analyses. The correlation analysis was shown as individual scatter plots in supplementary Figure 1.

Figure 6 Matrix of Pearson’s correlation coefficients among demographics, PAC-MI, OSA symptom severity. BMI showed a positive correlation with the N1 sleep and wakefulness of F4/O2 channel. The AHI exhibited a positive correlation with MI value during N1 sleep and wakefulness of F4/O2 channel. ODI was positively correlated with MI during N1 sleep in the F3 channel, while SaO2<90% was positively correlated with MI values during wake period in the O2 channel. A color-coded correlation scale is presented on the right of the plot. Based upon the scale, blue ones stand for lower correlations and red ellipses stand for higher correlations, ns illustrate insignificant correlations of a given variable with itself. Star symbols represent statistical significance levels: *** represents p < 0.001, ** represents p < 0.01, * represents p < 0.05, no stars represent p > 0.05. MAI: microarousal index.

Discussion

Our current understanding of the brain oscillation dynamics associated with Obstructive Sleep Apnea (OSA) is limited and does not provide a comprehensive perspective on this phenomenon. Cortical abnormalities, such as reduced theta and beta power, have been consistently reported in association with OSA. Given that the phase of lower frequencies governs the amplitude of higher frequencies in neural oscillation, and considering the observed abnormal activities in the amplitude of certain frequency bands, we conducted a study on the coupling between the phase and amplitude of different frequency bands in instances where abnormal phase synchrony is associated with Obstructive Sleep Apnea (OSA). Our study found that patients with moderate to severe OSA exhibited an increasing trend in the modulation index (MI) of low-frequency theta phase and high-frequency beta amplitude in the frontal and occipital regions during N1 and wakefulness periods, in comparison to patients with mild OSA. The elevated phase-amplitude coupling and OSA-related hypoxia indicators were positively correlated, suggesting that the PAC value may be one of the electrophysiological indicators reflecting the severity of OSA and chronic hypoxia.

EEG-PAC and OSA

PAC is a neuronal oscillation phenomenon that reflects interactions between neural oscillations of different frequencies. It is considered a crucial mechanism for brain information processing and integration, and can be used for diagnosing and stratifying the severity of brain disorders and neurodegenerative diseases.^33,34 OSA is a common sleep disorder that can lead to intermittent hypoxia and sleep fragmentation, affecting brain function.^6–9 Studies have shown that children exhibit different patterns of slow-wave modulation of high-frequency neural activity across sleep stages. Piantoni et al (2013) found that OSA patients had reduced θ-γ coupling strength during slow-wave sleep.³⁵ Erlan et al discovered that white matter damage is associated with patterns of high neuronal synchronicity during sleep stages, suggesting a link between white matter damage and chronic fatigue symptoms.³⁶ J Rajeswari et al’s study on brain functional connectivity in OSA patients revealed that the low-frequency delta band is closely related to OSA. The nodes and edges of the delta band highlighted the connection between the brain and OSA subjects, distinguishing OSA from healthy subjects.³⁷ Recent research by Silverstein BH et al shows that sedative drugs, while reducing sleep quality and increasing wake time, also decrease the high-frequency gamma-weighted phase-lag index associated with wakefulness.³⁸ Haralampos et al proposed that OSA patients exhibit significantly reduced sleep stage-specific PAC in sensorimotor areas, providing an objective marker for quantifying daytime sleepiness and respiratory distress in OSA.²⁵ More longitudinal studies are needed to understand how OSA affects PAC over time and whether changes in PAC can predict long-term outcomes in OSA. Further exploration of the relationship between PAC and other sleep parameters (such as sleep staging and micro-arousals) may provide new insights into the pathophysiological mechanisms of OSA.

PAC-MI Value of Low Frequency Theta Phase and High Frequency Beta in Phase N1

Theta waves and beta waves are two primary waveforms observed in EEG, with the highest power observed in the frontal lobe. Theta waves are commonly associated with states of relaxation and meditation, whereas beta waves are linked to wakefulness, focus, and cognitive activity.³⁹ Both theta and beta power have been found to be associated with OSA.^40,41 However, Relatively limited attention has been given to theta-beta PAC or PAC in general in patients with OSA. PAC is considered as a fundamental neurological process that promotes synchronization between local and global networks in the brain.⁴² It also connects and integrates activity across different frequency bands.^43–45

Our study revealed that in the N1 phase of NREM sleep, which marks the initial transition from wakefulness to sleep, there was a significant increase in PAC of theta and beta frequencies in the frontal lobe region of moderate and severe OSA groups, as compared to the mild OSA group. This suggests augmented synchronization of neural oscillations within the theta and beta frequency ranges, potentially indicating a disruption in the transitional state between relaxation and wakefulness in the brains of OSA patients.⁴⁶ Consequently, this situation could lead to an increased occurrence of wakefulness periods during sleep, adversely affecting sleep quality. Moreover, the observed enhancement in synchronization may potentially indicate adaptive changes within the brains of individuals with OSA as they cope with sleep disruptions and hypoxemia-induced stress. Importantly, we also found that enhancement in theta-beta PAC was prominent not only in the frontal lobe but also in the occipital lobe, and this was closely associated with hypoxia indicators. This suggests that chronic nocturnal hypoxia may intensify modulation of neural oscillations in both the frontal and occipital cortices, aligning with findings from previous studies.^47,48 These findings emphasize the complex interplay between OSA-related physiological challenges and neurological adaptations. Further research is essential to validate these findings. To ascertain the precise significance and underlying mechanisms of this synchronization enhancement, a larger sample size and more comprehensive analysis are warranted. Additionally, it is crucial to explore the potential impact of enhancement synchronization on the daily functioning and quality of life of individuals with OSA.

PAC-MI of Low-Frequency Theta Phase and High-Frequency Beta Amplitude During Wakefulness

Our study provides preliminary evidence for the increased theta-beta PAC-MI in the frontal and occipital regions during wakefulness. Beta oscillations are known to be closely associated with wakefulness,³⁹ and previous research has shown a correlation between beta activity and sleepiness levels,^49,50 suggesting that beta oscillations may serve as an indicator of sleep need.⁵¹ Clinical studies on insomnia patients have reported increased decoupling of beta and gamma oscillation.⁵² Additionally, research has found that patients with obesity-hypoventilation syndrome exhibit enhanced power in the slow-wave theta frequency range and high-frequency beta oscillations during wakefulness, which is associated with attention deficits, sleepiness, and nocturnal hypoxemia.^40,41 Intermittent hypoxia, characteristic of sleep apnea has complex, stage-specific effects on brain health. It disrupts hippocampal neurogenesis by reducing neural progenitors early and inhibiting neuron differentiation later, potentially through mechanisms involving the activation of HIF1a signaling, which subsequently enhances neuron generation following hypoxia.^53,54

In this study, we found that moderate and severe OSA patients exhibited increased PAC of low-frequency theta and high-frequency beta during wakefulness compared to mild OSA patients, which was positively associated with hypoxemia-related measures. This suggests that PAC can serve as an electrophysiological marker reflecting the degree of hypoxemia in OSA patients. Further investigation into the synchronization of low-frequency and high-frequency brain activity in OSA patients during wakefulness, as well as its relationship with cognitive function, attention, and behavioral performance, will help us understand the physiological mechanisms underlying the perception of fatigue in OSA patients during wakefulness.

Limitations and Future Directions

There are several limitations and shortcomings in our study. Firstly, we did not include a healthy control group for comparison. Our findings reveal significant variability in certain data sets, potentially stemming from EEG interference and sensitivity limitations, coupled with an inadequate sample size. In future research, we plan to incorporate a larger sample size and include healthy individuals as controls to further validate and enhance our findings. Secondly, our analysis of the brain’s electrical signals only focused on certain areas since we only focused on PSG-related channels, which limited our examination to specific regions. Future studies should consider utilizing high-density resting-state and sleep EEG recordings to improve spatial resolution. Lastly, while our study focused on resting-state EEG data, event-related EEG signals during activities, such as the N-back task, may provide more detailed information on brain activity.

Conclusion

In this research, we analyzed the modulation index of various frequency combinations (1–45Hz) in the brain’s electrical signals, as recorded by polysomnography (PSG) in patients with varying degrees of obstructive sleep apnea (OSA). Our study revealed a significant increase in phase-amplitude coupling (PAC) in the theta and beta frequency bands between the frontal and occipital cortices during N1 and wakefulness periods in OSA patients. The higher PAC-MI was positively correlated with indices of OSA severity and hypoxia indicators. Therefore, PAC may serve as a promising biomarker for OSA severity, and offer insights into the neurobiological mechanisms underlying cortical dysfunction in OSA patients. The present study provides new insights into the investigation of the associations between neural oscillations and respiratory hypoxia.