Independent component analysis of brain network alterations associated with cognitive impairment in coal workers’ pneumoconiosis

Abstract

Objective

This study aimed to investigate alterations in brain functional networks in Coal Workers’ Pneumoconiosis (CWP) using independent component analysis (ICA), a data-driven method that separates resting-state networks from fMRI data without requiring predefined regions of interest. And we will explore their correlations with clinical indicators and cognitive functions.

Methods

A total of 91 subjects were finally included after strict inclusion and exclusion criteria, comprising 33 CWP patients with cognitive impairment (CWP-CI), 31 CWP patients without cognitive impairment (CWP-nonCI), and 27 demographically matched healthy controls (HCs). The resting-state fMRI and cognitive assessments using the Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Clock Drawing Test (CDT) were performed in all subjects. Following data preprocessing, ICA was employed to identify resting-state networks (RSNs), and both intra-network and inter-network functional connectivity (FC) analysis were conducted. The correlations were assessed between significantly altered FC, clinical indicators and cognitive scale scores.

Results

Eight RSNs were identified, including the default mode network (DMN), executive control network (ECN), salience network (SN), dorsal attention network (DAN), sensorimotor network (SMN), higher visual network (hVN), auditory network (AUN), and basal ganglia network (BG). Intra-network connectivity analysis revealed a significant difference within SMN. Inter-network connectivity analysis revealed specific alterations. Compared to the HC group, the CWP-nonCI group exhibited enhanced FC in multiple network pairs, including hVN-SN, hVN-DMN, ECN-SN, DAN-BG, DMN-BG and SMN-SN, while the CWP-CI group primarily showed enhanced FC between hVN-SN and hVN-DAN. Furthermore, compared to the CWP-nonCI group, the CWP-CI group demonstrated significantly reduced FC between BG-AUN and SN-DMN. Some of the altered inter-network FC was significantly correlated with pulmonary function, cognitive scale scores, and work duration.

Conclusions

CWP patients exhibited specific reorganization of FC between brain networks. These findings suggested that disrupted inter-network functional communication may serve as a key neural mechanism in CWP-related cognitive impairment, providing new neuroimaging evidence for its basis.

Introduction

Coal workers’ pneumoconiosis (CWP) is one of the most severe occupational respiratory diseases in China, closely linked to the development of pulmonary fibrosis caused by long-term and repeated inhalation of coal dust particles [1]. Although various preventive strategies have been implemented over time, CWP continues to pose a serious public health concern. Its persistently high rates of disability and the presence of numerous complications not only result in significant and lasting impairment of pulmonary function in affected individuals, but also lead to considerable loss of working capacity, thereby placing heavy socioeconomic and psychological burdens on patients themselves, their families, and the broader healthcare system [2]. Importantly, clinical observations have shown that even after cessation of dust exposure, the condition of CWP patients may continue to progress, with ongoing damage or deterioration [3]. Traditionally, the harm associated with CWP has been attributed primarily to its respiratory manifestations, such as progressive dyspnea and frequent pulmonary infections [4, 5]. However, accumulating recent clinical evidence indicates that cognitive impairment (CI) is emerging as another major complication in CWP patients [6]. Research has demonstrated that individuals experiencing pulmonary dysfunction are at higher risk for cognitive decline [7], potentially due to a combination of neurodegenerative changes and vascular brain injury that may underlie this association [8]. In patients with CWP, persistent chronic hypoxia, activated systemic inflammatory responses, and increased oxidative stress are commonly observed and are thought to further exacerbate disease severity [9]. Studies on chronic obstructive pulmonary disease (COPD), a disease sharing core pathological features of chronic hypoxia and systemic inflammation with CWP, have revealed that these processes are associated with hippocampal injury and brain atrophy, closely linked to progressive cognitive deterioration [10]. Furthermore, symptoms such as nocturnal coughing and recurrent episodes of dyspnea can disrupt sleep and weaken circadian rhythms, potentially exacerbating CI in CWP patients [11].

Resting-state networks (RSNs) refer to collections of brain regions exhibiting synchronized neural activity during resting-state conditions, with the default mode network (DMN), executive control network (ECN), and visual network (VN) being the most extensively studied. Abnormalities in their functional connectivity (FC) have been closely associated with CI across various diseases, including COPD [12] and Alzheimer’s disease (AD) [13]. Therefore, we hypothesized that chronic hypoxia and pulmonary dysfunction would primarily affect inter-network connectivity. However, most of these studies have employed hypothesis-driven approaches based on predefined regions of interest (ROIs), which may overlook alterations in distributed networks not captured by a priori assumptions. This methodological limitation hinders a comprehensive understanding of the neural mechanisms underlying disease-related cognitive decline and underscores the need for data-driven techniques that can unbiasedly characterize whole-brain network reorganization.

Independent component analysis (ICA) is a data-driven signal processing technique that separates mutually independent functional components—such as RSNs—from fMRI data. Compared to ROI-based and seed-based methods, ICA enables objective and unbiased identification of whole-brain latent functional networks to avoid missing non-hypothesized but pathologically relevant brain regions [14], effectively separates neuronal signals from non-neuronal noise to improve result reliability, and supports simultaneous systematic analysis of both intra-network functional integrity and inter-network functional synergy [15], which is essential for elucidating the global neural network mechanisms underlying cognitive impairment [16]. To date, ICA has been extensively utilized in the investigation of brain function across a variety of diseases, including depression [17, 18], coronary heart disease [19], traumatic brain injury [20], acute thyrotoxic myopathy [21], and Parkinson’s disease [22]. Notably, studies focusing on the relationship between changes in functional network connectivity and cognitive function in COPD patients [23] provide valuable new perspectives for understanding the potential links between CWP and cognitive impairment. However, no study has employed ICA to systematically investigate brain network alterations in CWP patients and their association with cognitive decline.

Based on the above considerations, the present study was designed to investigate alterations in brain functional networks in CWP patients using ICA and to explore their correlations with clinical indicators and cognitive functions. Clarifying this distinct pattern of network disruption could advance our understanding of the neural network mechanisms underlying cognitive impairment in CWP patients, and offer potential neuroimaging biomarkers for the early detection and intervention of CWP-related cognitive dysfunction. Specifically, the findings of this study may help to identify high-risk CWP patients prone to cognitive decline at an early stage through resting-state fMRI examination, and provide a theoretical basis for exploring strategies to delay the progression of cognitive impairment in CWP patients.

Materials and methods

subjects

This retrospective study recruited 93 patients diagnosed with CWP through relevant examinations at our hospital between January 2024 and August 2025, as well as 30 healthy subjects matched for age, gender, and education level and recruited during the same period to serve as the healthy control (HC) group. Based on the results of neuropsychological cognitive assessments, CWP patients were further categorized into a CWP with cognitive impairment (CWP-CI) group and a CWP without cognitive impairment (CWP-nonCI) group. After applying inclusion and exclusion criteria, the final cohort consisted of 33 CWP-CI patients, 31 CWP-nonCI patients, and 27 HC subjects.

Inclusion criteria for CWP patients were as follows: (1) Diagnosis of CWP according to the Chinese Diagnostic Criteria of Occupational Pneumoconiosis (GBZ 70–2015); (2) Age ≥ 45 years, conscious, and able to cooperate with neuropsychological assessments and brain functional imaging examinations; (3) No history of stroke, severe traumatic brain injury (defined as loss of consciousness exceeding 30 minutes or post-traumatic amnesia lasting more than 24 hours), or psychiatric disorders (e.g., psychotic symptoms such as delusions or persistent hallucinations). Inclusion criteria for the control group were as follows: (1) Healthy volunteers not diagnosed with CWP; (2) Age ≥ 45 years, conscious, and able to cooperate with neuropsychological assessments and brain functional imaging examinations; (3) No history of stroke, severe traumatic brain injury (defined as loss of consciousness exceeding 30 minutes or post-traumatic amnesia lasting more than 24 hours), or psychiatric disorders (e.g., psychotic symptoms such as delusions or persistent hallucinations). Exclusion criteria for all subjects were as follows: (1) Comorbid anxiety or depression, identified as a Hamilton Anxiety Scale (HAMA) score ≥ 14 or Hamilton Depression Scale (HAMD) score ≥ 17; (2) Comorbid neurological disorders that may cause cognitive impairment, such as Alzheimer’s disease or Parkinson’s disease; (3) History of head trauma, alcoholism, substance abuse, or other psychiatric disorders; (4) Severe concomitant respiratory or circulatory system diseases; (5) Contraindications for MRI examination or other reasons rendering the participant unable to cooperate. This study protocol was approved by the Medical Ethics Committee of PKU Care Lu’an hospital (approval number: 2,022,009) and was conducted in strict compliance with the principles of the Declaration of Helsinki. Prior to the MRI scan, all subjects were fully informed about the study content and methodology and provided written informed consent.

Clinical data collection and scale assessment

This study collected general clinical data from participants, including gender, age, education level, work duration, smoking history, history of cerebrovascular disease, family history of dementia, peripheral oxygen saturation (SpO2), forced expiratory volume in one second (FEV1), forced vital capacity (FVC), and forced expiratory volume to forced vital capacity ratio (FEV1/FVC).

All subjects underwent neuropsychological assessment on the same day as the MRI scan. These assessments were specifically arranged for the purposes of this study and followed standardized protocols. Neuropsychological tests included the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) to evaluate overall cognitive function; HAMD and HAMA to determine the presence and severity of anxiety or depression; and the Clock Drawing Test (CDT) to assess spatial visuospatial abilities. The MMSE and MoCA were employed only for assessing cognitive status and for conducting correlation analysis with neuroimaging data.

Imaging data acquisition

All imaging data were acquired using a Siemens Prisma 3.0T MRI scanner equipped with a 64-channel head coil. Before scanning, all participants underwent training to ensure full cooperation and safety and were asked to remove any metal objects or jewelry. During the MRI procedure, participants lay supine with their head immobilized by a foam cushion to ensure comfort and minimize head motion, and wore ear muffs to reduce the impact of scanner noise. Throughout the process, participants were instructed to keep their eyes closed, remain relaxed, breathe calmly, stay awake, and avoid any deliberate mental activity as much as possible. Scanning was immediately stopped if participants experienced dizziness, nausea, vomiting, or any other discomfort.

First, conventional T1-weighted, T2-weighted, and T2-FLAIR sequences were performed to check for obvious abnormalities. Subsequently, resting-state functional MRI (rs-fMRI) data and structural data were collected. Rs-fMRI images were acquired using an echo-planar imaging sequence with the following parameters: echo time = 30 ms; repetition time = 2190 ms; flip angle = 66°; field of view = 200 mm × 200 mm; data matrix = 94 × 94; slice thickness = 3.0 mm; number of slices = 36; iPAT = 2, total scan duration = 5 minutes 37 seconds. Structural MRI data were acquired using a T1-weighted 3D magnetization-prepared rapid acquisition gradient-echo sequence with parameters: echo time = 2.36 ms; repetition time = 1800 ms; flip angle = 8°; field of view = 250 mm × 250 mm; data matrix = 288 × 288; phase oversampling = 10%; slice thickness = 0.9 mm; number of slices = 208; iPAT = 3, total scan duration = 8 minutes 22 seconds.

Data preprocessing

Image data preprocessing was performed using the RESTplus toolkit (http://www.restfmri.net/forum/RESTplus) based on the Matlab R2018b platform, and the specific workflow was as follows: (1) DICOM images were converted to NIFTI format; (2) the first 10 time points were discarded to eliminate signal instability and magnetization equilibrium during the initial MRI scanner startup; (3) time-layer correction was performed by selecting the middle slice as the reference, with odd slices denoted as 1:2:35 and even slices as 2:2:36; (4) head motion correction was conducted to ensure positional accuracy and framewise displacement (FD) values were computed for each subject to reflect the temporal derivative of the movement parameters. Time points that exceeded a max FD of 0.5 mm were excluded from subsequent analyses; (5) image registration involved aligning each subject’s T1-weighted structural images with their BOLD images; (6) spatial normalization established both individual brain space and Montreal Neurological Institute (MNI) standard space utilizing the T1-weighted images, with resampling to a voxel size of 3 × 3 × 3 mm to minimize inter-individual structural differences; (7) spatial smoothing was applied using a 6 mm full width at half maximum (FWHM) Gaussian kernel to enhance the signal-to-noise ratio, and the resulting preprocessed data were used for subsequent analyses.

Independent component analysis

The Group ICA of fMRI Toolbox (GIFT; https://trendscenter.org/software/gift) was employed to perform independent component analysis on the preprocessed data. The analysis followed several rigorous steps to identify meaningful independent components (ICs) representing distinct brain networks. Initially, the minimum description length criterion was used to determine the optimal number of ICs, as it balances model complexity and goodness-of-fit, making it suitable for identifying stable and interpretable components in fMRI data; principal component analysis (PCA) was then implemented to further simplify the data, reduce memory consumption, and decrease computational complexity; subsequently, the Infomax algorithm, which is well-suited for fMRI data due to its ability to process non-Gaussian and time-correlated sources, was used for Group Independent Component Analysis (GICA) computation; to ensure the stability of the estimated components, the Infomax algorithm was run 100 times with ICASSO, a resampling method that repeatedly applies ICA to randomly resampled data to guarantee the reliability of component estimation; following GICA, the temporal information and spatial distribution of each component were reconstructed, after which the mean time courses for each group were converted into z-scores to standardize the data; for further refinement of IC selection, the Display GUI module in GIFT was employed to compute regression coefficients, and components with similarity values greater than 0.1-alongside judgment by visual inspection—were considered, so only relevant and meaningful ICs were identified and retained while potential artifacts and noise were discarded; additionally, the LF/HF ratio of the time course of each retained component was calculated as a supplementary criterion to further distinguish neuronal signal-related components from noise [24]. Specifically, the LF/HF ratio was defined as the ratio of the power within the low-frequency band (0.01–0.15 Hz, where the spontaneous neuronal activity of resting-state brain networks is mainly concentrated) to the power within the high-frequency band (0.15–0.25 Hz, dominated by physiological noise such as respiratory and cardiac artifacts). And, a higher ratio indicates that the time course is dominated by neuronal activity rather than non-neuronal noise [25].

Intra-network functional connectivity analysis

Among the 20 components generated by ICA, nine were identified and selected for subsequent analysis, including seven meaningful RSNs. For each RSN, one-sample t-tests were first conducted to generate z-score maps for each group, with statistical significance determined using the family-wise error (FWE) correction at a threshold of p < 0.05. Subsequently, a one-way analysis of variance (ANOVA) was performed to compare RSNs activity among the three groups, with age and education level included as covariates. Correction at the cluster level was conducted using the FWE method, applying an initial voxel-level threshold of p < 0.001, and clusters that survived a cluster-level threshold of p < 0.05 were considered statistically significant in terms of between-group differences. For brain regions demonstrating significant group effects, post-hoc two-sample t-tests were further carried out to perform pairwise comparisons among three groups, with a total of three independent comparisons performed. To control the Type I error rate, the Bonferroni correction was applied and the corrected statistical significance threshold was calculated as p < 0.0167.

Inter-network functional connectivity analysis

The Mancovan module within the GIFT software package was utilized to assess differences in functional network connectivity (FNC) between the RSNs identified by ICA. First, low-pass filtering was applied to the network time courses, retaining frequencies below 0.15 Hz. Subsequently, the time series of each network component were extracted. After regressing out the effects of age and years of education as covariates, Pearson correlation coefficients were calculated between the time series of every pair of networks. These coefficients were then transformed into Z-scores using Fisher’s Z-transform to enhance data normality, yielding the FNC matrix. Finally, a general linear model was employed to identify significant between-group differences in FNC, with a statistical significance threshold set at p < 0.05. The above procedure is illustrated in Fig. 1.

Fig. 1.

Analysis pipeline in this study

Statistical analysis

Statistical analyses of general clinical data for the CWP-CI, CWP-nonCI, and HC groups were conducted using SPSS 26.0 software. All quantitative variables were first tested for normality using the Shapiro-Wilk test. Data conforming to a normal distribution were expressed as mean ± standard deviation (±s), while non-normally distributed data were presented as median (interquartile range). For comparisons between two groups, independent samples t-tests were applied for normally distributed variables and Mann-Whitney U tests for non-normally distributed variables. For comparisons among the three groups, ANOVA was used for normally distributed data and Kruskal-Wallis nonparametric tests were used for data that did not conform to normality. A p-value < 0.05 was considered statistically significant. Correlation analyses were performed between network connectivity measures that showed significant intra- or inter-network differences and clinical variables as well as scores on neuropsychological cognitive scales. For variables meeting the assumption of normality, Pearson correlation analysis was used; otherwise, Spearman correlation analysis was applied. Statistical significance was set at p < 0.05.

Results

Demographic and clinical characteristics

No significant differences were found in age or years of education among the CWP-CI, CWP-nonCI, and HC groups (p > 0.05). Similarly, no significant differences were observed between the CWP-CI and CWP-nonCI groups in work duration, FVC, FEV1, FEV1/FVC, or SpO2 (p > 0.05). However, the CWP-CI group showed significantly lower cognitive test scores compared to the CWP-nonCI group (p < 0.05) (Table 1). No subjects had max FD > 0.5 mm on more than 30 volumes in this study. No significant difference was found in the mean and max FD values between CWP patients and health controls.

Table 1.

Demographic data and clinical scale scores

	CWP-CI (n = 33)	CWP-nonCI (n = 31)	HC (n = 27)	P-value
Age	59.3 ± 5.82	56.52 ± 4.77	58.63 ± 4.52	0.084
Education (years)	9 (9,12)	9 (6,12)	9 (6,12）	0.103
Work duration (years)	30 (24,32)	26 (21,33.5)		0.762
FVC	83.52 ± 13.56	86.55 ± 13.88		0.381
FEV1	79.91 ± 13.35	82.45 ± 17.04		0.508
FEV1/FVC	90 (84,95)	89 (83,94.5)		0.92
SpO2	97 (95,98)	97 (96,98)		0.598
MMSE	26 (24,27)	28 (27,29)		＜0.001*
MoCA	21 (19,23)	25 (22.5,26)		＜0.001*
CDT	17.24 ± 9.45	23.16 ± 6.86		0.006*

P < 0.05 indicates statistically significant differences. CWP-CI, coal workers’ pneumoconiosis with cognitive impairment group; CWP-nonCI, coal workers’ pneumoconiosis without cognitive impairment group; HC, healthy control group; FVC, forced vital capacity; FEV1, forced expiratory volume in one second; FEV1/FVC, ratio of FEV1 to FVC; SpO2, peripheral oxygen saturation; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; CDT, Clock Drawing Test

Resting-state networks

In the three-group comparison, ICA extracted 20 ICs, from which 9 were identified as RSNs. The seven RSNs were as follows (Fig. 2): Auditory Network (AUN; IC16), primarily comprising the superior and middle temporal gyri; Sensorimotor Network (SMN; IC10), centered on the bilateral precentral and postcentral gyri; Default Mode Network (DMN; IC1+15), including the posterior cingulate cortex, precuneus, inferior parietal lobule, bilateral angular gyri, and medial prefrontal cortex; Executive Control Network (ECN; IC14), involving the bilateral dorsolateral prefrontal cortex and posterior parietal cortex; Salience Network (SN; IC17), including the dorsal anterior cingulate cortex, anterior insula, and parts of the prefrontal cortex; Higher Visual Network (hVN; IC13+20), mainly comprising the lateral occipital complex and occipital extrastriate cortex; and Dorsal Attention Network (DAN; IC4), primarily covering the bilateral intraparietal sulcus and frontal eye fields.

Fig. 2.

Spatial distribution maps of the resting-state networks identified in the study across all three groups, representing the seven major RSNs common to all three groups as identified by ICA. AUN, auditory network; SMN, sensorimotor network; DMN, default mode network; ECN, executive control network; SN, salience network; hVN, higher visual network; DAN, dorsal attention network; IC, independent component

In the comparison between the CWP-nonCI group and the HC group, ICA extracted 20 ICs, among which 9 components were selected as RSNs. The eight RSNs were as follows (Fig. 3): AUN (IC3), SMN (IC9), DMN (IC2+8), ECN (IC13), SN (IC17), DAN (IC14), hVN (IC1), and Basal Ganglia Network (BG; IC10), which primarily comprised subcortical structures such as the striatum, globus pallidus, and thalamus. In the comparison between the CWP-CI group and the HC group, ICA extracted 20 ICs, among which 10 components were identified as RSNs. The seven RSNs were as follows (Fig. 4): AUN (IC2+19), SMN (IC13), DMN (IC1+15), SN (IC14), DAN (IC12), hVN (IC5+7), and BG (IC11). In the comparison between the CWP-CI group and the CWP-nonCI group, ICA extracted 20 ICs, among which 11 components were identified as RSNs. The seven RSNs were as follows (Fig. 5): AUN (IC2+4), SMN (IC6), DMN (IC1+14+15), SN (IC17+19), DAN (IC5), hVN (IC12), and BG (IC10). The detailed information of RSNs identified in each group comparison is summarized in Table 2.

Fig. 3.

Spatial distribution maps of resting-state networks identified in the study between CWP-nonCI group and HC group, representing the eight major RSNs common to both groups as identified by ICA. AUN, auditory network; SMN, sensorimotor network; DMN, default mode network; ECN, executive control network; SN, salience network; DAN, dorsal attention network; hVN, higher visual network; BG, basal ganglia network; IC, independent component

Fig. 4.

Spatial distribution maps of resting-state networks identified in the study between CWP-CI group and HC group, representing the seven major RSNs common to both groups as identified by ICA. AUN, auditory network; SMN, sensorimotor network; DMN, default mode network; SN, salience network; DAN, dorsal attention network; hVN, higher visual network; BG, basal ganglia network; IC, independent component

Fig. 5.

Spatial distribution maps of resting-state networks identified in the study between CWP-CI group and CWP-nonCI group, representing the seven major RSNs common to both groups as identified by ICA. AUN, auditory network; SMN, sensorimotor network; DMN, default mode network; SN, salience network; DAN, dorsal attention network; hVN, higher visual network; BG, basal ganglia network; IC, independent component

Table 2.

Summary of resting-state networks identified in different group comparisons

Group comparison	Identified RSNs	Corresponding IC number
Three-group comparison	AUN	IC16
	SMN	IC10
	DMN	IC1, IC15
	ECN	IC14
	SN	IC17
	hVN	IC13, IC20
	DAN	IC4
CWP-nonCI vs HC	AUN	IC3
	SMN	IC9
	DMN	IC2, IC8
	ECN	IC13
	SN	IC17
	DAN	IC14
	hVN	IC1
	BG	IC10
CWP-CI vs HC	AUN	IC2, IC19
	SMN	IC13
	DMN	IC1, IC15
	SN	IC14
	DAN	IC12
	hVN	IC5, IC7
	BG	IC11
CWP-CI vs CWP-nonCI	AUN	IC2, IC4
	SMN	IC6
	DMN	IC1, IC14, IC15
	SN	IC17, IC19
	DAN	IC5
	hVN	IC12
	BG	IC10

RSNs, resting-state networks; IC, independent component; CWP-nonCI, coal workers’ pneumoconiosis without cognitive impairment; HC, healthy controls; CWP-CI, coal workers’ pneumoconiosis with cognitive impairment; AUN, auditory network; SMN, sensorimotor network; DMN, default mode network; ECN, executive control network; SN, salience network; hVN, higher visual network; DAN, dorsal attention network; BG, basal ganglia network

Differences in intra-network functional connectivity

In the analysis of intra-network functional connectivity, one-way ANOVA demonstrated significant differences in connectivity strength within the SMN among the CWP-CI, CWP-nonCI, and HC groups (Table 3). After applying FWE correction based on clusters (initial voxel threshold p < 0.001, cluster size ≥ 5), two significant clusters were identified, with peak coordinates located in the left Rolandic operculum (ROL; x = −51, y = 0, z = 6) and right ROL (x = 54, y = 3, z = 12). However, post hoc pairwise comparisons revealed that none of the between-group differences survived Bonferroni correction for multiple comparisons (the corrected significance threshold p < 0.0167).

Table 3.

Brain regions with significant intra-network connectivity differences among the three groups

RSN	Brain regions (AAL template)	Peak MNI coordinates (x、y、z)	Peak F-value	Cluster size
SMN	R_ROL	54，3，12	10.93	8
	L_ROL	−51，0，6	11.83	6

SMN, sensorimotor network; ROL, Rolandic operculum; AAL, anatomical automatic labeling; MNI, Montreal Neurological Institute

Differences in inter-network functional connectivity

The differences in FNC values between the CWP-nonCI group and the HC group are presented in Fig. 6. Compared to the HC group, the CWP-nonCI group demonstrated significantly increased functional connectivity between the following network pairs: hVN (IC1)–SN (IC17), hVN (IC1)–DMN (IC8), ECN (IC13)–SN (IC17), DAN (IC14)–BG (IC10), DMN (IC2)–BG (IC10), and SMN (IC9)–SN (IC17).

Fig. 6.

Networks showing significant changes between the CWP-nonCI group and the HC group. (a) Group-average FC matrix at the independent component level, showing the connectivity strengths between all pairs of identified independent components. (b) Group-average FC matrix at the brain network level, derived by merging and averaging independent components belonging to the same brain network from figure (a). In (a) and (b), colors closer to red indicate stronger positive functional connectivity, while colors closer to blue indicate stronger negative functional connectivity. (c) Average FNC difference matrix between CWP-nonCI and Hc groups, with x- and y-axes representing respective RSNs. (d) Network-level FNC difference circle plot between CWP-nonCI and Hc groups. In (c) and (d), p < 0.05 indicates significant network connectivity differences. Cool colors (e.g. blue) indicate decreased connectivity, while warm colors (e.g. red) indicate increased connectivity. DAN, dorsal attention network; DMN, default mode network; SN, salience network; BG, basal ganglia network; SMN, sensorimotor network; hVN, higher visual network; ECN, executive control network

The differences in FNC values between the CWP-CI group and the HC group are shown in Fig. 7. Relative to the HC group, the CWP-CI group exhibited significantly increased functional connectivity between hVN (IC7)–SN (IC14) and hVN (IC7)–DAN (IC12).

Fig. 7.

Networks showing significant changes between the CWP-CI group and the HC group. (a) Group-average FC matrix at the independent component level, showing the connectivity strengths between all pairs of identified independent components. (b) Group-average FC matrix at the brain network level, derived by merging and averaging independent components belonging to the same brain network from figure (a). In (a) and (b), colors closer to red indicate stronger positive functional connectivity, while colors closer to blue indicate stronger negative functional connectivity. (c) Average FNC difference matrix between CWP-CI and HC groups, with x- and y-axes representing respective RSNs. (d) Network-level FNC difference circle plot between CWP-nonCI and Hc groups. In (c) and (d), p < 0.05 indicates significant network connectivity differences. Cool colors (e.g. blue) indicate decreased connectivity, while warm colors (e.g. red) indicate increased connectivity. SN, salience network; DAN, dorsal attention network; hVN, higher visual network

The differences in FNC values between the CWP-CI group and the CWP-nonCI group are shown in Fig. 8. Compared to the CWP-nonCI group, the CWP-CI group exhibited significantly reduced functional conn ectivity between BG (IC10)–AUN (IC4) and SN (IC19)–DMN (IC15).

Fig. 8.

Networks showing significant changes between the CWP-nonCI group and the CWP-CI group. (a) Group-average FC matrix at the independent component level, showing the connectivity strengths between all pairs of identified independent components. (b) Group-average FC matrix at the brain network level, derived by merging and averaging independent components belonging to the same brain network from figure (a). In (a) and (b), colors closer to red indicate stronger positive functional connectivity, while colors closer to blue indicate stronger negative functional connectivity. (c) Average FNC difference matrix between CWP-nonCI and CWP-CI groups, with x- and y-axes representing respective RSNs. (d) Network-level FNC difference circle plot between CWP-nonCI and Hc groups. In (c) and (d), p < 0.05 indicates significant network connectivity differences. Cool colors (e.g. blue) indicate decreased connectivity, while warm colors (e.g. red) indicate increased connectivity. DMN, default mode network; AUN, auditory network; SN, salience network; BG, basal ganglia network

Correlation analysis

Pearson correlation analysis was conducted between abnormal FNC values among the RSNs and clinical indicators as well as neuropsychological cognitive scale scores in both the CWP-nonCI and CWP-CI groups. In the CWP-nonCI group, SMN–SN connectivity showed significant negative correlations with FVC (p = 0.023, r = −0.407) and FEV1 (p = 0.024, r = −0.405) (Fig. 9). In the CWP-CI group, hVN–DAN connectivity exhibited significant negative correlations with FEV1 (p = 0.023, r = −0.394), FEV1/FVC (p = 0.022, r = −0.399), and SpO2 (p = 0.040, r = −0.360), as well as a significant positive correlation with MMSE scores (p = 0.020, r = 0.404) (Fig. 10). In addition, SN–DMN connectivity in the CWP-CI group showed a significant negative correlation with work duration (p = 0.029, r = −0.380) (Fig. 11).

Fig. 9.

Correlation analysis between SMN-SN connectivity and clinical indicators in the CWP-nonCI group. (a) SMN-SN connectivity shows a significant negative correlation with FVC (p = 0.023, r = −0.407). (b) SMN-SN connectivity shows a significant negative correlation with FEV1 (p = 0.024, r = −0.406). SMN, sensorimotor network; SN, salience network; FVC, forced vital capacity; FEV1, forced expiratory volume in one second

Fig. 10.

Correlation analysis between hVN-DAN connectivity and clinical indicators and cognitive scale scores in the CWP-CI group. (a) hVN-DAN connectivity shows a significant negative correlation with FEV1 (p = 0.023, r = −0.394). (b) hVN-DAN connectivity shows a significant negative correlation with FEV1/FVC (p = 0.022, r = −0.399). (c) hVN-DAN connectivity shows a significant negative correlation with SpO2 (p = 0.040, r = −0.360). (d) hVN-DAN connectivity shows a significant positive correlation with MMSE scores (p = 0.020, r = 0.404). hVN, higher visual network; DAN, dorsal attention network; FEV1, forced expiratory volume in one second; FEV1/FVC, ratio of FEV1 to FVC; SpO2, oxygen saturation; MMSE, Mini-mental state examination

Fig. 11.

Correlation analysis between SN-DMN connectivity and work duration in the CWP-CI group. SN-DMN connectivity shows a significant negative correlation with work duration (p = 0.029, r = −0.380). SN, salience network; DMN, default mode network

Discussion

This study aimed to investigate changes in RSNs associated with cognitive impairment in patients with CWP using intra-network and inter-network functional connectivity analysis based on resting-state fMRI. The results showed that, although no significant differences in intra-network FC were found among the CWP-CI, CWP-nonCI, and HC groups, notable alterations in inter-network FC were observed. Specifically, compared to the HC group, the CWP-nonCI group exhibited enhanced connectivity across multiple network pairs, including hVN-SN, hVN-DMN, and ECN-SN. In contrast, the CWP-CI group demonstrated increased connectivity only in hVN-SN and hVN-DAN compared to HC. Furthermore, relative to the CWP-nonCI group, the CWP-CI group showed significantly reduced connectivity in BG-AUN and SN-DMN. Correlation analyses revealed that, in the CWP-nonCI group, SMN-SN connectivity was negatively correlated with pulmonary function (FVC and FEV1); in the CWP-CI group, hVN-DAN connectivity displayed negative correlations with pulmonary function (FEV1, FEV1/FVC, and SpO2) but a positive correlation with MMSE scores, while SN-DMN connectivity was negatively correlated with work duration. These findings suggest that cognitive impairment in CWP patients is closely related to abnormal inter-network FC among RSNs, providing neuroimaging evidence for understanding the neural mechanisms underlying cognitive dysfunction in CWP.

In this study, a total of eight RSNs were identified, each supporting specific cognitive functions and collectively maintaining of overall cognitive balance. Among these, the DMN—the most extensively studied RSN in rs-fMRI research—is considered to be involved in intrinsic cognitive processes such as episodic memory, self-referential processing, and emotional regulation [26]. The DMN demonstrates high levels of activity when the brain is at rest and disengaged from external stimuli. The ECN primarily subserves task processing and decision-making functions [27]; as a higher-order control network, the ECN selects and implements tasks based on information input from other brain networks. The SN is chiefly responsible for the ongoing monitoring of the external environment and for determining how other networks respond to novel information or stimuli [28]. More importantly, the SN regulates the switching between internal and external processing modes across the brain’s two principal control networks, the DMN and ECN [29]. These three core networks—DMN, ECN, and SN—together constitute a fundamental cognitive control system that underpins the dynamic equilibrium of human brain function. The hVN is mainly associated with visual and perceptual processing, but it also participates in other complex cognitive tasks. The BG comprises a group of subcortical nuclei that are primarily implicated in motor control, and are also involved in motor learning, executive functioning, behavioral regulation, and emotion [30]. The DAN is consistently engaged in language and visual working memory tasks, and is linked to top-down, goal-directed attentional control [31]. The SMN encompasses the primary sensorimotor cortex, premotor cortex, and supplementary motor area, facilitating voluntary movement [32]; it detects physical inputs and translates them into electrical signals, thus initiating motor responses. The AUN is involved in both primary and higher-order processing of auditory information, not only enabling the perception of sound but also playing a critical role in language comprehension and somatosensory perception [33]. Sustained and dynamic functional connectivity among these RSNs preserves cognitive equilibrium; consequently, any pathological disturbance disrupting this inter-network balance is likely to result in cognitive impairment.

Intra-network functional connectivity alterations

First, analysis of functional connectivity within the RSN revealed significant differences in connection strength among the CWP-CI, CWP-nonCI, and HC groups within the SMN, with peak regions located in the bilateral ROL. This finding is consistent with previous research [34]. The ROL is recognized as a key region within the SMN, responsible for processing sensory-motor functions of the oral cavity and pharynx [35]. Studies have shown that patients with COPD exhibit functional reorganization of the motor cortex as a compensatory response to increased respiratory load [36]. A similar mechanism may also operate in CWP patients, where chronic respiratory insufficiency could drive functional reorganization of the SMN as an adaptive process aimed at maintaining basic motor function. Recent evidence further suggests that the ROL is not merely involved in motor execution, but also plays an important role in integrating visual and somatosensory information to generate appropriate facial motor responses, working in close cooperation with the salience and emotion processing systems of the brain [37]. This indicates that the altered SMN connectivity observed in this study may not only reflect disturbances in primary sensorimotor processing, but could also be related to dysfunction in the interaction between the sensorimotor system and the salience and emotional networks, particularly in the context of processing bodily sensations related to dyspnea in CWP patients. However, participants with clinically significant anxiety or depression were excluded. Therefore, the observed ROL alterations more likely reflect compensatory sensorimotor reorganization rather than affective dysfunction, though this remains speculative and warrants further investigation. Additionally, in post hoc pairwise comparisons, statistically significant differences were not found between any two groups. This may be attributed to the limited sample size or heterogeneity in brain functional alterations among CWP patients. This limitation reduces the statistical power of the study, and future large-sample longitudinal studies are needed to further explore the subtle changes in the intra-network connectivity.

Inter-network functional connectivity alterations

Secondly, inter-network FC analysis revealed that the CWP-nonCI group exhibited enhanced connectivity compared to the HC group, specifically between the following network pairs: hVN–SN, hVN–DMN, ECN–SN, DAN–BG, DMN–BG, and SMN–SN. The core function of the SN is to enable the dynamic switching of cognitive control. CWP patients frequently present with symptoms such as dyspnea and shortness of breath, which are closely linked to the SMN, as supported by findings in COPD respiratory function research [38]. Thus, the observed enhancement of SMN–SN connectivity may be interpreted as an adaptive mechanism by which patients strengthen monitoring of somatosensory signals and behavioral regulation to cope with symptoms of declining respiratory function. Similarly, increased ECN–SN connectivity indicates greater communication efficiency between executive control and salience detection, potentially allowing patients to better allocate attentional resources during cognitive tasks and compensate for emerging cognitive deficits associated with CWP. Enhanced connectivity between hVN and both SN and DMN may suggest the establishment of more robust pathways integrating basic sensory information into higher-order cognitive control systems in response to increasing cognitive demands.

Previous research has demonstrated that patients with COPD show decreased amplitude of low-frequency fluctuation (ALFF) in the BG, attributed to neurovascular coupling alterations and suppressed spontaneous neuronal activity under chronic hypoxic conditions [39]; such a decline in ALFF reflects reduced local efficiency in processing sensory, cognitive, and motor information. Structural neuroimaging studies have further revealed that COPD patients exhibit reduced gray matter volume in the BG and microstructural damage in connecting white matter tracts, providing morphological evidence for localized pathological changes in the BG [40]. However, the present study found that functional connectivity between the BG and both the DAN and DMN was enhanced. This pattern of increased inter-network connectivity in the context of local structural damage parallels compensatory changes observed in other conditions [41]. It is therefore plausible that the BG may strengthen its functional connections with other core cognitive networks to enhance both the brain’s responsive system to external environmental challenges and its perceptual regulation system for internal bodily states, thereby maintaining overall functional stability.

Compared to the HC group, the CWP-CI group exhibited enhanced functional connectivity between hVN–SN and hVN–DAN, indicating heightened sensitivity of visual processing pathways to pathological damage. Previous electrophysiological research has demonstrated that abnormalities in the visual network, evidenced by attenuated early event-related potentials (P1, N1), are an early and pervasive feature of cognitive decline in stages such as subjective cognitive decline and mild cognitive impairment [42]. The increased connectivity between hVN and both SN and DAN may represent a compensatory mechanism that preserves basic visual attention and environmental monitoring functions, which is consistent with findings in Alzheimer’s disease [43]. In contrast, compared with the CWP-nonCI group, the CWP-CI group showed reduced functional connectivity in BG–AUN and SN–DMN circuits. Weakened BG–AUN connectivity suggests dysfunction in subcortical-cortical circuits integrating sensory information with higher-order cognitive behaviors, supported by anatomical studies of basal ganglia–cortical pathways [44]; this likely underlies slower information processing and sensorimotor integration deficits, correlating with reduced MoCA language subscale scores in the CWP-CI group. The diminished SN–DMN connectivity disrupts the dynamic balance between cognitive control and internal mental processes [45, 46], a pattern also reported in Alzheimer’s disease patients with SN dysfunction and impaired DMN suppression [47]. Such imbalance may compromise the SN’s ability to modulate DMN activity during externally focused tasks, leading to ineffective suppression of task-irrelevant thoughts and heightened distractibility. These deficits are likely to manifest as impairments in executive function, which are captured by the attention and executive subdomains of the MoCA. Collectively, these findings suggest that SN-DMN decoupling contributes to the executive dysfunction observed in CWP-CI patients.

Correlation analysis

In correlation analysis, SMN–SN connectivity in the CWP-nonCI group showed a negative correlation with pulmonary function (FVC and FEV1). Thus, poorer lung function was associated with stronger inter-network connectivity, consistent with previous findings [48]. In the CWP-CI group, hVN–DAN connectivity was negatively correlated with FEV1, FEV1/FVC, and SpO2, but positively correlated with MMSE scores. This seemingly paradoxical finding suggests that although poorer pulmonary function is associated with stronger hVN–DAN connectivity, this enhanced connectivity may represent a compensatory mechanism that helps preserve cognitive function to a limited extent. In other words, patients with worse lung function may rely more on hVN-DAN connectivity to maintain cognitive performance, but this compensatory effect is insufficient to fully offset cognitive decline. Additionally, weakened SN–DMN connectivity in the CWP-CI group was significantly correlated with longer work duration, indicating that greater cumulative dust exposure is associated with more severe disruption of brain functional connectivity.

Limitations

This study still has several limitations. First, as a cross-sectional investigation with a limited sample size, it is difficult to directly infer causal relationships between brain functional network alterations and cognitive impairment in CWP, highlighting the need for future longitudinal studies with larger cohorts; second, although age and educational attainment were included as covariates, other confounding factors such as concomitant medication use were not controlled, potentially influencing the results; third, the analysis was limited to resting-state functional connectivity, so incorporating dynamic functional connectivity analysis or additional methods in future research may provide a more comprehensive understanding of intra- and inter-network connectivity; fourth, while ICA reveals unconstrained brain network connections, it cannot determine the directionality of interactions, so future studies employing approaches that assess influence and directionality between networks are warranted. Finally, despite rigorous head motion correction, residual group differences in motion parameters may still influence functional connectivity results, representing a potential confounding factor.

Conclusion

This study systematically examined intra- and inter-network connectivity changes within resting-state functional brain networks in CWP patients and their association with cognitive impairment using independent component analysis. While intra-network connectivity remained largely unchanged, distinct alterations in inter-network connectivity were observed: the CWP-nonCI group exhibited enhanced connectivity among multiple networks, potentially reflecting early compensatory mechanisms, whereas the CWP-CI group showed weakened connectivity, closely associated with cognitive decline. Correlation analysis demonstrated that these network-level alterations were significantly related to pulmonary function, cognitive scores, and work duration. Collectively, these findings offer new insights into the neural mechanisms underlying cognitive impairment in CWP, emphasizing the pivotal role of disrupted inter-network connectivity in CWP-related cognitive deficits.

Abbreviations

CWP

Coal workers’ pneumoconiosis

ICA

Independent component analysis

CWP-CI

Coal workers’ pneumoconiosis with cognitive impairment

CWP-nonCI

Coal workers’ pneumoconiosis without cognitive impairment

HC

Healthy controls

RSNs

Resting-state networks

FC

Functional connectivity

DMN

Default mode network

ECN

Executive control network

SN

Salience network

DAN

Dorsal attention network

SMN

Sensorimotor network

hVN

Higher visual network

AUN

Auditory network

BG

Basal ganglia network

CI

Cognitive impairment

COPD

Chronic obstructive pulmonary disease

AD

Alzheimer’s disease

ROI

Regions of interest

FVC

Forced vital capacity

FEV1

Forced expiratory volume in one second

SpO₂

Peripheral oxygen saturation

MMSE

Mini-mental state examination

MoCA

Montreal cognitive assessment

CDT

Clock drawing test

HAMA

Hamilton anxiety scale

HAMD

Hamilton depression scale

rs-fMRI

resting-state functional magnetic resonance imaging

FEW

Family-wise error

ANOVA

One-way analysis of variance

FNC

Functional network connectivity

IC

Independent component

ROL

Rolandic operculum.

Author contributions

Xuecong Lv: data collection, image processing, data analysis, visualization, manuscript writing, Gang Cao: technical support, funding application, project management, Lingling Ren: data collection, image processing, Yuxiang Zhao: data collection, image processing, Boting Xue: data collection, image processing, Zhengde Xu: data collection, Xiheng Man: data collection, Akifumi Hagiwara: supervision, resources, Yongbo Liu: project management, research support, Xiaowei Han: supervision, review and editing, and research support.

Funding

This work was supported by 2023 Shanxi Provincial Health Commission Project (2023149), 2023 Changzhi Municipal Health Commission Project (20), 2025 Shanxi Provincial Healthcare Commission Project (2024204), 2024 Lu’an Chemical Group CO., LTD Second Batch of Newly Established Projects of Science and Technology Programs (024BH706040A), Summit Advancement Disciplines of Zhejiang Province (Wenzhou Medical University-Pharmaceutics), and Quzhou Technology Project (2024 K078).

Data availability

The datasets analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Medical Ethics Committee of PKU Care Lu’an hospital (approval number: 2,022,009) and was conducted in strict compliance with the principles of the Declaration of Helsinki. All participants provided written informed consent for inclusion in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.