# Disrupted topological organization of the default mode network in mild cognitive impairment with subsyndromal depression: A graph theoretical analysis

## TL;DR

This study finds that mild cognitive impairment with subsyndromal depression disrupts brain network organization, which could help identify and understand different types of cognitive decline.

## Contribution

The study introduces disrupted topological metrics of the default mode network as potential biomarkers for mild cognitive impairment with subsyndromal depression.

## Key findings

- MCID showed significantly lower global and nodal efficiency in the left aMPFC compared to MCIND.
- Network metrics correlated with depressive symptoms and cognitive function in MCID.
- An SVM model achieved 83% accuracy in distinguishing MCID from MCIND using topological metrics.

## Abstract

Subsyndromal depression (SSD) is common in mild cognitive impairment (MCI). However, the neural mechanisms underlying MCI with SSD (MCID) are unclear. The default mode network (DMN) is associated with cognitive processes and depressive symptoms. Therefore, we aimed to explore the topological organization of the DMN in patients with MCID.

Forty‐two MCID patients, 34 MCI patients without SSD (MCIND), and 36 matched healthy controls (HCs) were enrolled. The resting‐state functional connectivity of the DMN of the participants was analyzed using a graph theoretical approach. Correlation analyses of network topological metrics, depressive symptoms, and cognitive function were conducted. Moreover, support vector machine (SVM) models were constructed based on topological metrics to distinguish MCID from MCIND. Finally, we used 10 repeats of 5‐fold cross‐validation for performance verification.

We found that the global efficiency and nodal efficiency of the left anterior medial prefrontal cortex (aMPFC) of the MCID group were significantly lower than the MCIND group. Moreover, small‐worldness and global efficiency were negatively correlated with depressive symptoms in MCID, and the nodal efficiency of the left lateral temporal cortex and left aMPFC was positively correlated with cognitive function in MCID. In cross‐validation, the SVM model had an accuracy of 0.83 [95% CI 0.79–0.87], a sensitivity of 0.88 [95% CI 0.86–0.90], a specificity of 0.75 [95% CI 0.72–0.78] and an area under the curve of 0.88 [95% CI 0.85–0.91].

The coexistence of MCI and SSD was associated with the greatest disrupted topological organization of the DMN. The network topological metrics could identify MCID and serve as biomarkers of different clinical phenotypic presentations of MCI.

The global efficiency and nodal efficiency were significantly lower in mild cognitive impairment with SSD (MCID), and the accuracy of the support vector machine model was 0.83 for distinguishing MCID from MCIND. MCID is correlated to the greatest disrupted topological organization, which may serve as biomarkers of different MCI subgroups.