Background: Late-life depression (LLD) is characterized by cognitive and social impairments. Determining neurobiological alterations in connectivity in LLD by means of fMRI may lead to a better understanding of the neural basis underlying this disorder and more precise diagnostic markers. The primary objective of this paper is to identify a structural model that best explains the dynamic effective connectivity (EC) of the default mode network (DMN) in LLD patients compared to controls.
Methods: Twenty-seven patients and 29 healthy controls underwent resting-state fMRI during a period of eight minutes. In both groups, jackknife correlation matrices were generated with six ROIs of the DMN that constitute the posterior DMN (pDMN). The different correlation matrices were used as input to estimate each structural equation model (SEM) for each subject in both groups incorporating dynamic effects.
Results: The results show that the proposed LLD diagnosis algorithm achieves perfect accuracy in classifying LLD patients and controls. This differentiation is based on three aspects: the importance of ROIs 4 and 6, which seem to be the most distinctive among the subnetworks; the shape that the specific connections adopt in their networks, or in other words, the directed connections that are established among the ROIs in the pDMN for each group; and the number of dynamic effects that seem to be greater throughout the six ROIs studied [t = 54.346; df = 54; p < .001; 95 % CI difference = 5.486-5.906].
Limitations: The sample size was moderate, and the participants continued their current medications.
Conclusions: The network models that we developed describe a pattern of dynamic activation in the pDMN that may be considered a possible biomarker for LLD, which may allow early diagnosis of this disorder.
Keywords: Default mode network; Dynamic effective connectivity; Late-life depression; Major depressive disorder; Structural equation modeling; fMRI.
Copyright © 2022 The Authors. Published by Elsevier B.V. All rights reserved.