A total of 201 postpartum women who underwent a physical examination 42 days after delivery at the Beijing Haidian Maternal and Child Health Hospital from June 2021 to January 2022 were initially included in the study. Their ages ranged from 22 to 43 years, with an average age of 30.06 ± 3.86 years.

Inclusion criteria: (1) Participants were included if they were 42 days post-delivery. The defined “postpartum” as the period starting immediately after childbirth and extending to 6 weeks after delivery. This period is consistent with the clinical definition of early postpartum. (2) Only women who had a singleton, live infant delivery were included in the study. (3) Participants were required to be permanent residents of Beijing, having lived in the city throughout pregnancy and within 1 year after childbirth. (4) All participants had to have clear consciousness and be capable of independently and accurately completing the questionnaires.

Exclusion criteria: (1) Participants with a history of psychiatric disorders (e.g., depression, anxiety, and bipolar disorder) were excluded to avoid confounding effects from pre-existing mental health conditions. (2) Women with any neurological diseases were excluded to ensure that the study focused on the mental health effects of postpartum conditions. (3) Complicated Pregnancies or Medical Conditions: Those with medical conditions that could contribute to postpartum depression or affect sleep patterns (e.g., pre-eclampsia, hypothyroidism) were excluded from the study. (4) Drug Abuse or Other Substance Use Disorders: Participants with a history of drug or alcohol abuse were excluded due to their potential influence on mental health outcomes. (5) Women who had experienced multiple pregnancies (e.g., twins, triplets) were excluded to control for the possible additional physical and emotional strain of caring for more than one infant. (6) Participants who failed to provide complete questionnaire responses were excluded from the final analysis.

All participants signed informed consent voluntarily before the study began, and the research was approved by the Ethics Committee of the Department of Psychological and Cognitive Sciences at Tsinghua University (Protocol Number: 60101).

Insomnia is diagnosed by clinicians according to the criteria in the Chinese Guidelines for the Diagnosis and Treatment of Adult Insomnia (23), including: (1) The latency period of falling asleep is prolonged, and the sleep time is >30 min; (2) Sleep maintenance disorder, wake up more than twice during the night or wake up early; (3) Decreased sleep quality; (4) Reduced total sleep time, usually <6 h.

The Chinese version of the Edinburgh Postnatal Depression Scale (EPDS) was used to screen for postpartum depressive status (24), which is a self-report questionnaire consisting of 10 items with a total score of 30 points. It has been proven to have sufficient reliability and validity (25) and is widely used in clinical and research settings within Chinese-speaking populations (26, 27). In this study, a total score of 10 points was chosen as the critical value for screening postpartum depression, with a sensitivity and specificity of 85 and 84% (28). The survey was conducted using the Wenjuanxing platform (wjx.cn) on mobile phones, ensuring the authenticity and effectiveness of the questionnaire results.

The verbal fluency task (VFT) was applied to assess vocabulary, lexical access speed, and executive function (29). In this study, the Chinese version of VFT was used with three parts. During the first part (pre-task), participants started saying “one,” “two,” “three,” “four,” and “five” continuously for 30 s in Chinese. During the second parts (task period), participants were asked to list as many items as possible under the categories of the coloer or fruits in 60 s (Figure 1B). The total runtime was 145 s. Participants were introduced to the VFT to confirm their understanding of the instructions. It was ensured that participants could listen to voice prompts, could form Chinese words, and could articulate the results. During the assessments, participants were seated in a comfortable and natural position, facing away from the screen. Their hands were placed naturally on their knees or the chair's armrest.

Hemoglobin concentration was measured using a 48-channel near-infrared optical imaging system (NirScan, Danyang Huichuang Medical Equipment Co., Ltd., China). The placement of channels was focused on the anterior part of the brain, particularly the frontal and temporal regions. This choice was based on previous research that highlights the significance of these areas in understanding the neural mechanisms underlying postpartum depression and sleep disorder (30, 31). The sampling frequency was 11 Hz, and the wavelengths were 730, 808, and 850–730 and 850 nm were the major wavelengths and 808 nm was used as the isotopic wavelength for correction. The FPz channel in the 10/20 international system was used as the central midline probe. A total of 31 source-detector (SD) probes were placed, including 15 sources and 16 detectors. The probes were fixed at 3 cm intervals, covering the PFC of each subject, with the lowest probe placed along the Fp1-Fp2 line (Figure 1A). The channels and their corresponding brain regions are detailed in Supplementary Table 1.

The HOMER2 toolbox was employed, a graphical user interface program based on MATLAB 2013b (MathWorks, Inc.; Natick, US), for preprocessing the near-infrared spectroscopy data (32). Initially, a method based on motion standard deviation and spline interpolation was used to eliminate motion artifacts (33, 34). Motion artifacts were distinguished by identifying sliding window standard deviations that exceeded a certain threshold. Any signal changes in the data channels that exceeded the threshold (standard deviation, STD, and amplitude, AMP) were labeled as motion artifacts (hmr Motion Artifact By Channel; input parameters: tMotion = 0.5, tMask = 1, STD thresh = 30, AMPthresh = 0.5). Spline interpolation was applied to remove the artifacts (utilizing the hmr Motion Correct Spline function with input parameter P = 0.99). Subsequently, a second-order Butterworth bandpass filter with a cutoff frequency of 0.01–0.1 Hz was employed to remove physiological noise caused by heartbeats (1 Hz), respiration (~0.2–0.5 Hz), and high-frequency noise (35, 36). Finally, the filtered optical data was transformed into HbO and HbR concentrations using the modified Beer-Lambert law (37). This study focused on the HbO signal due to its ability to better reflect cortical activity (38). A 60-s task period was utilized as the time window for analyzing changes in oxygenated hemoglobin.

In this study, our objective is to analyze the brain functional network properties by extracting the time series of interest from all channels using photodetectors. Subsequently, Pearson correlation analysis was performed on these time series to obtain the correlation coefficient r between each pair of brain regions, generating the corresponding correlation coefficient matrix R_ij. A specific threshold was then applied to the correlation matrix to binarize it, yielding the binarized connectivity matrix R_ij of the brain functional network. Since network parameters vary with the level of network sparsity and there is currently no standardized threshold for selecting the optimal sparsity level, we selected a set of sparsity thresholds with a step size of 0.01, ranging from a minimum of 0.1 to a maximum of 0.40 (0.1 < S < 0.40), based on previous experience (39). This resulted in a series of brain functional network sets with gradually increasing sparsity levels for each subject.

The brain functional properties calculated include the clustering coefficient, characteristic path length, global efficiency, and local efficiency, which were quantitatively measured using graph theory methods (40). The specific calculation formulas are as follows:

If a real network's normalized topological indices meet the criteria γ=Cpnorm =Cp-real /Cp-rand >1 and λ=Lpnorm =Lp-real /Lp-rand ≈1, it is considered to exhibit small-world topological organization. Additionally, a comprehensive measure of small-world network characteristics, denoted as σ = γ/λ > 1, can be defined.

Statistical analysis was conducted using SPSS 20.0 (IBM Corp., NY, USA). Graphs were generated using the NirsKit package and GraphPad Prism 8 software. The Shapiro-Wilk test was applied to assess the normality of the data. Continuous variables were expressed as means ± standard deviations, and categorical variables were expressed as numbers (percentages). Demographic data were analyzed using independent sample t-tests. Functional connectivity was examined by Pearson correlation analysis of the time series of each channel pair. Brain network topological properties, including small-world attributes, clustering coefficient, global efficiency, local efficiency, and characteristic path length, were compared between groups using independent sample t-tests or Mann-Whitney tests. The relationship between postpartum women's sleep status and EPDS scores was evaluated using Spearman's correlation. Statistical significance was set at p < 0.05; all p-values were two-tailed. False discovery rate FDR correction was applied to results from multiple comparisons across channels.

Out of the initial screening, a total of 58 participants were excluded for the following reasons: six due to pre-existing medical conditions, seven due to pre-existing psychiatric conditions, 14 due to multiple pregnancies, and 31 due to incomplete data. As a result, 143 postpartum women were included in the final analysis. The sleep status of 143 postpartum women was assessed by clinicians. Forty of them were included in the insomnia group with a positivity rate of 27.97 % (mean age = 33.63 years; SD = 3.43), while 103 without insomnia were included in the control group (mean age = 31.20 years; SD = 3.40). The depression status of 143 postpartum women was also assessed using EPDS ≥10 points as a threshold. The EPDS scores were significantly higher in the insomnia group than in the control group (Mann-Whitney test: Z = 2.71, p = 0.007). Of the 40 postpartum women with insomnia, eight had an EPDS score ≥ 10. Among the 103 postpartum women without insomnia, six had an EPDS score ≥ 10.

We examined the functional connectivity strength of the whole PFC based on HbO signals. Two 48 × 48 connectivity coefficient matrix of the functional brain network for insomnia and control group were calculated (Figure 2). The results showed that the 22–33 [HbO: (0.53 ± 0.31 vs. 0.28 ± 0.32), t₍₁₄₁₎ = −4.20, p < 0.001, 95% CL (0.13–0.37)] and the 40–45 [HbO: (0.70 ± 0.18 vs. 0.53 ± 0.25), t₍₁₄₁₎ = −3.97, p < 0.001, 95% CL (0.09–0.26)] channel pairs connection were significantly higher in the insomnia group than in the control group, which passed the FDR correction. The differences in brain network connectivity are shown in Figure 2. Channel 22 represents pars triangularis Broca's area, and 33 represents Middle and Superior Temporal gyrus; Channels 40 and 45 represent the left and right dorsolateral prefrontal cortex (dlPFC), respectively (Supplementary Table 1).

As depicted in Figure 3, small-world topology properties, quantified by the σ metric, were >1 in both groups across a wide range of sparsity thresholds (0.1 < S < 0.4). However, lower small-world topology metrics were observed in the insomnia group compared to the healthy group. Significant differences were observed in the sparsity thresholds of 0.17, 0.18, 0.19, and 0.20. Consequently, small-world topology metrics were analyzed within the sparsity range with significant differences between the insomnia group and the control group.

Compared with the control group, the insomnia group had a significantly lower clustering coefficient (p = 0.009, mean difference = 0.030) and local efficiency (p = 0.020, mean difference = 0.025) of brain functional networks, whereas no significant difference was found in global efficiency (p = 0.13, mean difference = 0.017) and characteristic path length (p = 0.12, mean difference = 0.125). Next, the network topology properties are computed for each node. Results revealed that only node 12 (left temporal pole) exhibited significantly reduced local efficiency (p < 0.000, mean difference = 0.259) and clustering coefficient (p < 0.000, mean difference = 0.225) compared to the control group. Differential p-values for brain regions are presented in Figure 3.

To further confirm the association between changes in network topology properties of the temporal pole area induced by insomnia and postpartum depression, the Spearman correlation analysis was conducted to analysis between the local efficiency and clustering coefficient of the left temporal pole and EPDS score. The results revealed a negative correlation between the local efficiency of the temporal pole area and EPDS scores in the postpartum woman (r = −0.19, p = 0.03). Also, the characteristic path length of the temporal pole area exhibited a negative correlation with EPDS scores (r = −0.2, p = 0.02; Figure 4).

Our study results reveal the EPDS scores of insomnia group were significantly higher than those in the control group, indicating a close association between insomnia and postpartum depressive states. This finding is consistent with previous research, supporting the link between sleep disorders and postpartum depression (41, 42).

The brain is considered to be a dynamic and interconnected functional network. Sleep is crucial in maintaining normal brain function, and global or local dysfunction may lead to insomnia. The PFC is involved in maintaining the quiescent state of the brain, integrating internal and external environmental information, emotion regulation, and episodic memory retrieval (43). Researchers have used different brain imaging techniques, such as fMRI, EEG, and fNIRS, to explore the PFC function in patients with chronic insomnia disorder. The fMRI study by Altena et al. (44) revealed low activation of medial and inferior frontal cortex regions in insomnia patients, highlighting the importance of the PFC in sleep. An EEG study by Perrier et al. (45) found a lower power spectrum of β1 in the prefrontal cortex in patients with primary insomnia. The fNIRS study by Gong et al. (46) showed a significant reduction in prefrontal mean channel functional connectivity in patients with insomnia.

The connectivity between the left and right dlPFC was enhanced in the insomnia group compared with the control group. The dlPFC is a region closely associated with cognitive functions such as emotion regulation, cognitive control, and attention (47). The enhanced connectivity between the dlPFC may reflect an increased information transfer and functional coordination between these brain regions in postpartum women with insomnia. Women in the postpartum period experience dramatic changes in their lives, including the care of a newborn and changes in life patterns. These factors may have led to decreased sleep quality and triggered enhanced prefrontal activity in emotional and cognitive control in response to these challenges (41, 48). This alteration of the functional brain network may be intended to stabilize cognitive functioning during sleep-disordered states.

The female brain undergoes dynamic neural plasticity during pregnancy and the postpartum period, with notable gray matter volume reductions in various brain regions, such as the hippocampus, cingulate cortex, medial orbital frontal cortex, and insula (49, 50). These changes have been shown to play critical roles in social processes (51), emotion regulation (52), and their relevance to the development of depression (53). The increased connectivity in the dorsolateral prefrontal cortex, a region crucial for emotion regulation and cognitive control, may signify adaptive alterations in postpartum women with sleep disorders in handling emotional and attention aspects. While these adaptations are believed to prepare new mothers for their roles, they cannot rule out the potential contributions to the occurrence of psychiatric disorders (54), given the concurrent nature of structural brain changes and postpartum mental health issues. Significantly higher EPDS scores in the sleep-disordered group further support this point.

The results also revealed enhanced connectivity between the right Broca's area and the left superior temporal gyrus, regions closely associated with cognitive functions such as language processing and semantic comprehension. This study involved activating brain areas related to language processing during the VFT. Thus, the increased connectivity in postpartum women with insomnia may reflect intensified information transmission and functional coordination between different brain hemispheres. Borragán et al. (55) found that in normal individuals staying awake for a short period, the right PFC exhibited increased activity, while the left PFC showed reduced activity. Furthermore, Honma et al. (56) demonstrated that in alertness tests for healthy subjects, increased activation in the right prefrontal cortex was positively correlated with alertness. This suggests that activating the right prefrontal cortex may assist healthy individuals or short-term insomniacs overcome drowsiness and provide sufficient activity to meet the demands of more cognitively challenging tasks.

In short, insomnia have a notable impact on the PFC in postpartum women. Postpartum women may adapt to effects by altering connectivity between brain regions to maintain stable cognitive and emotional functions. This additional enhancement of functional connectivity might serve as a compensatory mechanism to address potential deficiencies in other brain regions. The intensified connectivity could be a natural physiological response, but it may also have implications for sleep problems and the risk of depression. However, further research is needed to gain a deeper understanding of these intricate relationships, determining whether enhanced prefrontal cortex activity plays a positive or negative role and how it influences sleep and depressive states.

The topological properties of prefrontal brain networks were analyzed based on graph theory to understand the mechanisms underlying the adjustment of functional brain networks between the insomnia group and the control group. Small-world attributes are crucial indicators describing network connectivity and information transmission efficiency in brain networks. Small-world networks exhibit a balance of high clustering and short path lengths, meaning that nodes in the network tend to form tight clusters and are interconnected through relatively short pathways (57). In the postpartum insomnia group, the brain functional network maintains a small-world topology similar to healthy participants, suggesting a balance between local specialization and global integration in information processing remains (58). However, the small-world topology indices in postpartum women with sleep disorders were lower than in the healthy control group. The reduced small-world attributes might reflect abnormalities in functional connectivity and integration in the prefrontal cortex region (58). This implies that information transmission and integration in the prefrontal cortex region could be disrupted in the context of sleep disorders, showing significant differences in the network's sparsity range from 0.17 to 0.20.

Further analysis revealed that in postpartum women with sleep disorders, local efficiency and clustering coefficient were significantly reduced, while network global efficiency and characteristic path length did not exhibit significant reductions. Local efficiency is an indicator describing the efficiency of information transmission between nodes in a network, with higher local efficiency indicating that nodes in the network are more efficient in information transmission and integration (59). The clustering coefficient measures the degree to which nodes in the network cluster together, representing the tendency of nodes in the network to form tightly-knit clusters (59). Local efficiency and clustering coefficient reflect the local information processing capability of nodes in the network and the degree of close connections between nodes.

The reduction in local efficiency and clustering coefficient may suggest that in postpartum women with sleep disorders, there is a decrease in the brain's ability to transmit and process local information efficiently. Further analysis of the local efficiency and clustering coefficient for each node showed a significant decrease in these measures in the temporal pole region of the insomnia group. The temporal pole is closely associated with various functions, including memory, emotions, language, and spatial cognition (60). The reduction in local efficiency and clustering coefficient may imply a functional decline in the temporal pole region of postpartum women with sleep disorders, which could decrease their ability to process and regulate information, affecting aspects such as memory, emotions, and language capabilities. The reduction in local efficiency and clustering coefficient may also reflect a decrease in the connectivity and integration capabilities between the temporal pole region and other brain areas. Impaired information transmission and coordination between the temporal pole region and other areas of the brain could lead to isolation and disrupted information transfer in certain functional areas of the network, thereby affecting overall information processing and functional integration. The reduced local efficiency and clustering coefficient of the temporal pole nodes were negatively correlated with depressive symptoms, and previous research has already shown the crucial role of the temporal pole in emotion regulation and cognitive function (61). Functional abnormalities in the temporal pole region may lead to difficulties in emotion regulation and cognitive processing in postpartum women, associated with the onset and persistence of depressive symptoms. While our results suggest a potential link between functional brain network alterations in the temporal pole area and EPDS scores, these findings remain exploratory due to the relatively small sample size. Future research with larger cohorts is essential to validate these associations and establish stronger evidence for the role of functional connectivity changes in postpartum depression and sleep disorders.

Future studies should address the limitations identified in our current analysis by employing a more granular approach to patient classification. Specifically, increasing sample sizes and conducting detailed subgroup analyses—differentiating between patients with sleep disorders and depression, those with sleep disorders without depression, patients with depression without sleep disorders, and those with neither condition—will allow researchers to better isolate the independent and combined effects of sleep disturbances and depression on brain connectivity. This refined methodology is expected to enhance the robustness of findings and lead to more specific conclusions, ultimately improving the generalizability of results to a broader population of postpartum patients. Moreover, incorporating longitudinal designs could provide valuable insights into how these relationships evolve over time, thereby deepening our understanding of the interplay between sleep and mental health in postpartum contexts. To further strengthen future research, we recommend including a detailed assessment of support systems, such as caregiving duties, the presence of family or spouses, and the use of external childcare. This will help to elucidate additional factors that may influence the mental health and sleep quality of postpartum individuals. This study did not include NIRS measurements both pre- and post-partum. While the inclusion of such measurements could provide valuable insights into the neural changes associated with postpartum depression, the constraints of this design limited our ability to implement them. Future research should consider incorporating pre- and post-partum comparisons to enhance understanding of brain activity changes during this critical period. Future studies could incorporate neuroimaging techniques like MRI to examine potential volumetric differences alongside functional assessments, offering a more comprehensive view of the neurobiological mechanisms involved in postpartum sleep disorders.