Schizophrenia—Brain network hub and working memory performance in patients with schizophrenia

Editor—Summer Leaf

Working memory (WM) disorder is a cognitive symptom
of schizophrenia.
Cerebral cortex damage has been demonstrated in different sensory patterns and at different levels of task requirements, often manifested as decreased capacity and slower response in the cerebral cortex [1].

Although WM deficits in schizophrenia have been comprehensively studied using imaging and electrophysiology [2-4], few studies have investigated WM deficits in the context of brain network topology
。 Revealing topological features associated with cognitive symptoms is important because it can guide targeted treatment
.

General functional link abnormalities in schizophrenia have been extensively studied
.
In schizophrenia, structural connections and other structural indicators are weakened [5-7], however, both enhanced and decreased functional connections have been reported, and abnormalities occur in many different regions/networks of the brain [8, 9].

The diffuse distribution of this abnormal functional connection may reflect heterogeneity in study populations and methods, as well as the complexity
of the pathophysiology of schizophrenia.
Graph theory provides a robust framework
for studying functional brain topologies and hypothetical dysfunctional network hubs in schizophrenia.

Recently, the Department of Psychiatry at Massachusetts General Hospital and the Hamdi Eryilmaz Research Group at Harvard Medical School in Charleston published a report entitled "Network hub centrality and working memory" at Schizophrenia performancein schizophrenia"
.
The study used weighted degrees to quantify participants' cortical centers to identify key nodes in
working memory (WM) performance.
The results showed that an increase in weighting of the default mode network (DMN) was generally associated with poorer WM performance (accuracy and reaction time) in both patient and control groups.

The higher the weighting of the ventral attention (VAN) node in the right temporal epithelial layer, the better
the patient's WM performance (accuracy).
These results suggest that schizophrenia is associated with
dysfunctional centers of the cortical system that support internal and external cognition.

The study included 29 patients and 29 control participants
matched by age and sex.
The Sternberg Item Recognition Paradigm (an oral WM task) is used for assessment4 WM function under different loads (1, 3, 5 and 7 letters), and collect MRI data
during the mission.

Preprocessing and functional connectivity analysis
of acquired resting fMRI data.
Use Gordon et al.
's template [10] to define regions of interest (ROIs), which include surface-based resting state functional connection boundary mapping generation 333 blocks
.
These blocks are assigned to 12 different networks: DMN, visual, frontoparietal, Dorsal attention, ventral attention, salience, cingulo opercular), somatomotor, somatomotor latera, auditory, cingulate gyrus parietal lobe ( Cingulo Parietal) and retrosplenial
.
Functional connections are calculated using Pearson correlations, the correlation
between the average signal time series for each ROI and the average time series for each other ROI.
Use the Fisher z-transform to convert the 333 × 333 correlation matrix to z-map to enhance the normality
of the correlation distribution.
The weighting is further calculated by the formula k i = _{∑n j = 1}a_ij, where a i _j is the i The correlation between block and jth block time series, n is the total number of blocks and k i is the _i The weighting of
blocks.
The weighted differences between groups were detected using a permutation test, and FDR was used to correct
for significance p-values.

Figure 1 results show an average weighting plot
for each group.
Among them, Figure 1A is the average weighting plot of the normal person group, Figure 1B is the average weighting plot of patients, and Figure 1C is the block used in the study and the corresponding network color coding
.

Fig.
1 Weighting of
each group.

(Source: Eryilmaz, H.
et al.
, Schizophr, 20 2 2).

The average performance of both groups at each WM load level was analyzed (Figure 2).

The results of the between-group two-sample t-test showed that the patient's performance was not as accurate as the control group when the WM task load was 3, 5 and 7 (p = 0.
001) (Figure 2A).

The analysis also revealed a significant difference in response between the two groups to moderate WM loads (3 and 5), manifested by slower patient responses (p = 0.
025).
(Figure 2B).

Figure 2 Behavior results

(Source: Eryilmaz, H.
et al.
, Schizophr, 20 2 2).

The relationship between weighting and the behavioral outcomes of WM accuracy (average of all WM loads) and reaction time (average of all trials and loads) was evaluated using correlation analysis.

The results showed that the weighting of the left pre-wedge lobe and the right parahippocampal gyrus (PHG) was strongly inversely correlated with WM accuracy in patients, while ventral attention was multiple The weighting of superior temporal sulcus (STS) nodes is positively correlated with WM accuracy (Fig 3）
。 These include the inferior parietal lobule (IPL) and the dorsolateral prefrontal cortex The default mode network, including dlPFC) and the frontal eyefield (FEF), and higher weighting of the dorsal attention area were associated
with longer response times.
Posterior auxiliary motor area (SMA), superior temporal sulcus (STS).
and fusiform gyrus (FG) weighting, the faster the patient response (Figure 3).

Figure 3 Weighting and behavioral results

(Source: Eryilmaz, H.
et al.
, Schizophr, 20 2 2).

In order to test the predictive value of weighting on the behavior of new individuals, a weighting-based prediction model
was used for the entire sample (N=58).
Retention-one-method cross-validation (LOOCV) is used to select weighted characteristics
related to WM performance.
The results show that the main nodes selected by the model under the 7-letter condition - DMN includes the back of the cingulate gyrus (PCC), Ventral attention nodes in the medial frontal cortex (mPFC), left dorsolateral prefrontal cortex (dlPFC), and superior temporal sulcus (STS) (Fig 4）
。

Figure 4 Prediction model

(Source: Eryilmaz, H.
et al.
, Schizophr, 20 2 2).