Changes of topological attributes of brain structural network in patients with postpartum depression
10.3760/cma.j.cn113903-20240102-00003
- VernacularTitle:产后抑郁症患者脑结构网络拓扑属性的特征性改变
- Author:
Kai XIE
1
;
Yang LI
;
Xiaolan ZHU
;
Yujiao CAI
;
Yifeng LUO
;
Zhihong CAO
;
Yuefeng LI
;
Jiajia SHI
Author Information
1. 江苏大学附属宜兴医院医学影像科，宜兴 214200
- Keywords:
Postpartum depression;
Diffusion tensor imaging;
Brain structural network;
Topological attributes
- From:
Chinese Journal of Perinatal Medicine
2024;27(6):468-476
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the features of the brain structural network in patients with postpartum depression (PPD).Methods:This cross-sectional study included PPD patients who visited the mental health counseling clinic after delivery at the Jiangsu University Affiliated Yixing Hospital from June 2013 to September 2022 (PPD group). Matched non-PPD postpartum women based on age, years of education, and body mass index who came for postpartum follow-up (non-PPD postpartum group), and non-pregnant women who visited the hospital or underwent physical examinations during the same period (non-pregnant group) were also included. Demographic data and diffusion tensor imaging (DTI) data were collected for all three groups. The brain was partitioned into 90 regions using an anatomical template to construct the brain structural network. Network-based statistics (NBS) were applied to further screen and construct subnetworks. The efficacy of the subnetworks in identifying PPD was evaluated through multivariable logistics regression models and receiver operating characteristic curves. A comparison of the connectivity strength of white matter tracts and topological attributes of brain structural network parameters was conducted using independent samples t-tests, and the results were corrected using the false discovery rate (FDR) method. Results:(1) A total of 116 subjects were included, with 40 in the non-pregnant group, 40 in the non-PPD postpartum group, and 36 in the PPD group. PPD group had higher Edinburgh Postnatal Depression Scale (EPDS) scores than the non-pregnant and non-PPD postpartum groups [(18.0±4.1) scores vs. (2.5±1.2) and (6.1±2.1) scores, F=340.40; t=24.65,10.60 and 16.16 in pairwise comparison; all P<0.001]. (2) Compared to the non-pregnant group, there was a decrease in the connectivity strength of nine white matter tracts within the brain structural network of the postpartum group (including left dorsolateral superior frontal gyrus-left anterior cingulate and paracingulate gyrus, left dorsolateral superior frontal gyrus-right amygdala, left dorsolateral superior frontal gyrus-left insula, left insula-left lentiform nucleus, left insula-left hippocampus, left hippocampus-right amygdala, left hippocampus-left precuneus, left anterior cingulate and paracingulate gyrus-right amygdala, and right amygdala-right hippocampus) (all P<0.05, FDR corrected). No increased connection strengths were observed. There were no significant differences in the connection strengths of these nine tracts between the non-PPD and PPD groups. (3) A characteristic subnetwork for the maternal group was successfully constructed based on the nine tracts, which exhibited typical small-world properties (σ>1). Compared to the non-PPD maternal group, the characteristic path length in the PPD group was increased [(3.904±0.328) vs. (4.130±0.433), t=－2.58], and global efficiency was decreased [(0.361±0.036) vs. (0.331±0.053), t=2.91] (both P<0.05). Local property comparisons showed that the node efficiency values for the left dorsolateral superior frontal gyrus, left insula, left anterior cingulate and paracingulate gyrus, left hippocampus, right hippocampus, right amygdala, left precuneus and left putamen in the PPD group were significantly reduced [(0.273±0.023) vs. (0.267±0.030), t=0.98; (0.299±0.035) vs. (0.276±0.041), t=2.64; (0.265±0.019) vs. (0.258±0.025), t=1.38; (0.318±0.028) vs. (0.305±0.031), t=1.92; (0.312±0.027) vs. (0.302±0.031), t=1.50; (0.322±0.030) vs. (0.298±0.026), t=3.71; (0.356±0.040) vs. (0.338±0.056), t=1.62; (0.346±0.028) vs. (0.331±0.036), t=1.74; all P<0.05]. However, only the differences in node efficiency values for the left insula and right amygdala remained significant after FDR correction (corrected P=0.041 and 0.003). (4) Global efficiency, as well as node efficiency for the left insula and right amygdala, demonstrated good value for identifying PPD [areas under the curve (AUC) and their 95% CI were 0.827 (0.732-0.922), 0.741 (0.628-0.854), and 0.761 (0.653-0.867), respectively], with even better performance when combined [0.897 (0.828-0.969)]. (5) In the PPD group, global efficiency ( r=－0.43, P=0.008), node efficiency for the left insula ( r=－0.39, P=0.019), and node efficiency for the right amygdala ( r=－0.42, P=0.011) were all negatively correlated with EPDS scores. Conclusion:Aberrations in global efficiency, node efficiency for the left insula, and node efficiency for the right amygdala may serve as characteristic neuroimaging biomarkers for PPD.
