This study recruited female nurses from Yancheng Clinical Medical College of Yangzhou University as subjects, with data collection conducted from September 2024 to December 2024. Inclusion criteria: the occupational burnout group: (1) female, aged 20–40 years; (2) right-handed; (3) according to the burnout norms of Chinese nurses (19), the critical value of burnout was determined as all three-dimensional scores exceeding critical values (emotional exhaustion (EE) ≥ 27 points, depersonalization (DP) ≥ 8 points, and personal accomplishment (PA) ≤ 24 points) and the healthy control group: (1) female, aged 20–40 years; (2) right-handed; (3) all three-dimension scores below critical values (EE < 27 points, DP < 8 points, PA > 24 points). Exclusion criteria: all subjects with any of the following conditions were excluded: (1) endocrine, neurological, or psychiatric disorders or other primary diseases; (2) pregnant or lactating women; (3) history of drug dependence, smoking, or alcohol consumption; (4) adverse reactions during scanning leading to termination of the experiment or contraindications to MRI scanning; (5) data collection failure during scanning or unclear images; (6) MRI images showing organic brain lesions; (7) other serious physical illnesses. Based on the inclusion and exclusion criteria, 80 subjects were ultimately selected, with 40 in the occupational burnout group and 40 in the heathy control group. The two groups were matched in terms of age and years of education. This study strictly adhered to the ethical principles of the Declaration of Helsinki and has received approval from the Ethics Committee of the Yancheng Clinical Medical College of Yangzhou University (2024–82) and obtained informed consent from all subjects involved.

Prior to MRI scanning, general information and clinical data were collected, including age, years of education, body mass index (BMI), Beck Anxiety Inventory (BAI) (20), Beck Depression Inventory-II (BDI-II) (21), and Maslach Burnout Inventory-Human Services Survey (MBI-HSS) scale. MBI-HSS scale assesses burnout across three dimensions in service industries: EE (9 items): emotional exhaustion from work; DP (5 items): depersonalized responses to care recipients; PA (8 items, reverse-scored): feelings of competence and achievement. Items are rated on a 7-point Likert scale (0: never to 6: daily) (22). Within our sample, the Cronbach’s α coefficients for the dimensions of EE, DP, and PA in the MBI-HSS were 0.970, 0.965, and 0.960, respectively.

Rs-fMRI and structural 3D-T1-weighted images were acquired using a 3.0 T MRI scanner with a 24-channel head coil (Discovery 750w, GE, United States) at Yancheng Clinical Medical College of Yangzhou University. Parameters included: rs-fMRI: repetition time (TR)/ echo time (TE) = 3,000/35 ms, 128 volumes, field of view (FOV) = 24 cm × 24 cm, Slice thickness = 5.0 mm, and voxel size = 3.75 × 3.75 × 4 mm; structural 3D-T1: TR = 750 ms, TE = 2.8 ms, FOV = 24 cm × 24 cm, Slice thickness = 1.0 mm, number of slices = 152, flip angle = 15°, and voxel size = 0.5 × 0.5 × 1 mm.

Preprocessing was performed using the DPABI 8.2 software, including: (1) removal of the first 10 time points; (2) slice timing correction; (3) realignment for head motion; (4) exclusion of participants with maximum displacement > 3 mm or rotation > 3°; (5) spatial normalization to the standard Montreal Neurological Institute (MNI) space achieved through the Dartel alignment method; (6) linear regression to reduce errors; (7) regression of nuisance covariates; and (8) band-pass filtering (0.01–0.1 Hz). All participants included in the final analysis met these head motion criteria.

Brain structural 3D-T1-weighted images were preprocessed using SPM12 and CAT12, including bias field correction, skull stripping, alignment to MNI template, and segmentation into gray matter, white matter, and cerebrospinal fluid. TIV was extracted for all participants.

DC was calculated using DPABI 8.2, computing the DC value for each voxel in the brain. Pearson correlation coefficients were used to estimate functional connectivity between all pairs of gray matter voxels. A threshold of r > 0.25 was used to derive the adjacency matrix, followed by conversion of individual voxelwise DC values into a z-score map. Subsequently, the DC maps obtained were smoothed spatially using a 6-mm full width at half-maximum (FWHM) Gaussian kernel.

Using the AAL 90 template, brain regions with significant DC differences between the burnout and control groups were selected as ROIs for whole-brain FC analysis. The average time series for each ROI was calculated, and Fisher’s z-transformation was applied to obtain normally distributed z-score maps.

SPSS 27.0 software was used for statistical analysis. First, normality tests were conducted for age, years of education, BMI, TIV, BAI, BDI-II, and MBI-HSS. For normally distributed measurement data, independent sample t-tests were used, while non-parametric tests were applied for skewed distribution data. A p-value < 0.05 was considered to indicate significant between-group differences.

Two-sample t-tests were conducted to compare differences in FC and DC between the burnout group and the control group, with age, years of education, TIV, BAI, and BDI-II as covariates using SPM12 in MATLAB (R2020b). Multiple comparison corrections used cluster-level False Discovery Rate (FDR), with voxel-level p < 0.001 and cluster-level p < 0.05. DC and FC values from ROIs (showing significant group differences) were extracted for subsequent correlation and machine learning analyses.

SPSS 27.0 software was used to conduct Spearman correlation analysis between the extracted brain DC and FC signal values with significant between-group differences and burnout-related scales (EE, DP, PA). A p-value < 0.05 was considered to indicate significant correlation.

This study used DC values and FC values from brain regions with significant differences obtained through two-sample t-tests and cluster-level FDR correction, as well as combined DC and FC values, as input features for Linear discriminant analysis (LDA) to construct classification models for burnout and healthy populations. The predictive performance of the models was evaluated using receiver operating characteristic (ROC) curves, calculating the area under curve (AUC), sensitivity, specificity, and accuracy, and comparing the performance of these three models in the validation set. Leave-one-out cross-validation (LOOCV) was used to validate the predictive ability and stability of the classification models. Subsequently, permutation tests were conducted to evaluate the significance of classification accuracy and AUC. During the permutation tests, group labels were randomly shuffled 5,000 times, and classification was performed on the newly generated datasets, calculating the classification accuracy and AUC each time. The p-value for classification accuracy and AUC was the number of times the classification accuracy or AUC in the 5,000 random cases exceeded the true value, divided by 5,000. Results were considered significant when the p-values for both accuracy and AUC were less than 0.05.

The study included 40 participants in occupational burnout group and healthy control group. Demographic and clinical characteristics of both groups are summarized in Table 1. No statistically significant differences were observed between the occupational burnout and healthy control groups in age (median [interquartile range, IQR]: 33 [27–37] vs. 33.5 [28–36] years; p = 0.973), years of education (median [IQR]: 16 [16–16] vs. 16 [16–16]; p = 0.724), or body mass index (BMI; median [IQR]: 21.89 [19.64–23.81] vs. 20.81 [19.55–23.63]; p = 0.544). The occupational group exhibited markedly higher anxiety levels on the BAI (median [IQR]: 27 [23.25–33] vs. 24 [22–26]; p = 0.009) and greater depressive symptoms on the BDI-II (median [IQR]: 9.5 [5.5–13.75] vs. 4 [0–9.5]; p = 0.002). Scores on MBI-HSS subscales further distinguished the groups. The occupational group reported significantly higher EE (median [IQR]: 36.0 [30.5–40.00] vs. 13.5 [9–17]; p < 0.001) and DP (median [IQR]: 13.0 [10.0–16.75] vs. 3 [0–5]; p < 0.001), alongside significantly lower PA (median [IQR]: 20.0 [17.25–22.0] vs. 38.5 [30–43.75]; p < 0.001) compared to the healthy control group.

As shown in Table 2 and Figure 1, significant differences in DC were observed between the occupational burnout group and the healthy control group. After applying cluster-level FDR correction (voxel threshold: p < 0.001 and cluster threshold: p < 0.05), the occupational burnout group exhibited significantly reduced DC values in the left precuneus (t = −4.37, p < 0.001) and the right precuneus (t = −4.55, p < 0.001) compared to the healthy control group.

Seed-based whole-brain FC analysis was conducted using the DC-differentiated brain regions as seeds. As summarized in Table 3 and illustrated in Figure 2, the occupational burnout exhibited significantly reduced FC values between the left precuneus and the right medial orbitofrontal cortex (mOFC, t = −4.57, p < 0.001) compared to the healthy control group.

As illustrated in Figures 3A,B, EE was negatively correlated with DC values in the left precuneus (r = −0.46, p = 0.003) and right precuneus (r = −0.36, p = 0.023). Additionally, DP was negatively correlated with DC values in the left precuneus (Figure 3C) (r = −0.348, p = 0.028). As shown in Figure 3D, PA was positively correlated with FC values between the left precuneus and the right mOFC (r = 0.378, p = 0.016).

To evaluate the classification performance, we employed LOOCV LDA model to build models distinguishing occupational burnout individuals from healthy controls. Input features for these models were DC and FC values extracted from brain regions exhibiting significant group differences. As shown in Table 4 and Figure 4, the LDA model incorporating both DC and FC values achieved the highest performance (sensitivity = 0.800, specificity = 0.900, accuracy = 0.850, and AUC = 0.902), significantly outperforming models using only DC values (sensitivity = 0.800, specificity = 0.675, accuracy = 0.738, and AUC = 0.851) or only FC values (sensitivity = 0.575, specificity = 0.675, accuracy = 0.625, and AUC = 0.657).