Granger Causality Changes during Rt-fMRI Neurofeedback Training

of Emotion Regulation for Insomnia Patients

Zhiyuan Feng

, Hui Gao

, Zhonglin Li

, Linyuan Wang

, Chi Zhang

, Yan Bin

Yongli Li

and Li Tong

Henan Key Laboratory of Imaging and Intelligent Processing, PLA Strategic Support, Force Information Engineering

University, Zhengzhou, Henan, China

Department of Radiology, Henan Provincial People’s Hospital & People’s Hospital of Zhengzhou University, Zhengzhou,

Henan, China

Keywords: Rt-Fmri, Granger Causality Model, Emotion Regulation.

Abstract: The amygdala is a key brain region in the emotional network. Studies have shown that emotion regulation

neurofeedback training targeting amygdala based on real-time functional magnetic resonance can effectively

improve the symptoms of insomnia patients. However, the brain mechanism for this improvement remains

unclear. In this paper, Granger causality model was constructed to analyze the difference of causality

between different brain regions before and after neurofeedback training. Firstly, the brain regions related to

emotion regulation with significant differences in ReHo before and after neurofeedback training were

selected as regions of interest. Secondly, the time series of the regions of interest were extracted to establish

the Granger causality model. Finally, through group level analysis, the difference of effective connection

before and after neurofeedback training was used as a biomarker to evaluate the effect of emotion

regulation. The results have shown that rt-fMRI neurofeedback training targeting the amygdala significantly

regulated the activity of brain regions related to emotion regulation in insomnia patients. And the effective

connections from the right triangle inferior frontal gyrus to the left amygdala, the left precuneus to the left

middle frontal gyrus, and the right middle cingulate gyrus to the left middle frontal gyrus were significantly

enhanced. While the effective connection from the left middle frontal gyrus to the left precuneus was

significantly reduced. Moreover, these changes were consistent with the scale evaluation results and clinical

psychiatric studies, which further demonstrated that real-time fMRI neurofeedback training can change the

effective connectivity of brain regions related to emotion regulation, and these changes could be used as a

potential biomarker to evaluate the effect of neurofeedback training.

1 INTRODUCTION

Insomnia is a common clinical disorder

characterized by difficulty falling asleep or

maintaining sleep for more than three months. With

the development of brain science, non-invasive

methods have become a hot topic in the field of

insomnia research (Christopher and Decharms

https://orcid.org/0000-0001-7829-7244

https://orcid.org/0000-0001-5023-3222

https://orcid.org/0000-0001-5072-0354

https://orcid.org/0000-0001-6751-3087

https://orcid.org/0000-0003-4559-4925

https://orcid.org/0000-0001-5934-8181

https://orcid.org/0000-0002-4654-456X

2008). Among them, the neurofeedback technology

based on real-time functional magnetic resonance

imaging(rt-fMRI) detects the changes in blood

oxygen caused by the enhancement of neuronal

activity in a specific area of the brain to measure the

neuronal activity of the brain indirectly, and

feedback relevant information to the subjects in real

time. It has been used in the diagnosis and

intervention of various psychiatric diseases because

it can target the brain regions related to patients'

neurological defects and has high spatial and

temporal resolution (Brühl 2015). Recent studies

have shown that rt-fMRI neurofeedback emotion

regulation training can effectively improve the

symptoms of insomnia patients (CHEN 2021, Zhang

and Gao 2021).

Feng, Z., Gao, H., Li, Z., Wang, L., Zhang, C., Bin, Y., Li, Y. and Tong, L.

Granger Causality Changes during Rt-fMRI Neurofeedback Training of Emotion Regulation for Insomnia Patients.

DOI: 10.5220/0011387300003438

In Proceedings of the 1st International Conference on Health Big Data and Intelligent Healthcare (ICHIH 2022), pages 577-583

ISBN: 978-989-758-596-8

577

At present, neurofeedback emotion regulation

technology based on rt-fMRI has been successfully

applied as an adjunctive treatment for major

depression (Bodurka and Jerzy 2017; Mehler D and

Sokunbi 2018, Bruce and Doré 2018), anxiety

disorder (Abraham and Kaufmann 2013),

schizophrenia (Zweerings and Hummel 2019) and

other diseases related to emotion regulation

disorders. A number of studies have also shown that

sleep supports continuous changes in neuronal

representation of emotional experiences, and

insomnia disorder may be related to the inability of

patients to eliminate emotional distress (Bonnet and

Arand 2010). There are also data suggesting that

insomnia and depression may share a common

pathology, that is, the diagnosis of insomnia and the

severity of sleep disorders are both related to high

cortisol secretion (Young and Korszun 2010). These

studies have proved pathologically that insomnia

patients can be treated by emotion regulation

training.

How to analysis the effectiveness of rt-fMRI

neurofeedback is focus of attention. In addition to

the evaluation method of clinical scale, the

evaluation of training effect by using brain image

features has become one of the focuses of scientific

research and clinical attention. The brain

connectivity analysis of fMRI functional data is

mainly divided into two types: functional

connectivity and effective connectivity. From the

direction of connection, functional connection only

emphasizes whether there is a connection between

brain regions, and does not study the direction of the

connection, that is, undirected connection. In order

to explain the coupling relationship between neural

activities in different brain regions more accurately,

scholars put forward the concept of effective

connection, which refers to the influence exerted by

the activity of one brain region on the activity of

another brain region, revealing the causal effect of

non-adjacent brain regions in finger spac2 (Friston

and Frith 2010). For the analysis of effective

connections, model-driven methods are generally

adopted to reflect the changes of neural activities by

establishing the causal relationship between neurons.

The commonly used methods are as follows:

psychophysiological interaction (PPI), structural

equation model (SEM), dynamical causal model

(DCM) and Granger causality analysis (GCA)

(Liang X and Wang J H, 2010). GCA is one of the

most widely used effective connection analysis

methods in current research.

Many researchers have analyzed brain

connectivity abnormalities in insomniacs using

different functional connectivity methods. For

example, in 2018, Li C et al. (Li and Dong 2018)

used Granger causality model to analyze the

effective brain connectivity between patients with

primary insomnia and healthy subjects, and found

that patients with insomnia had reduced effective

connectivity from the right anterior insula to

bilateral precuneus, left posterior central gyrus and

bilateral posterior cerebellum, and decreased

effective connectivity from bilateral orbitofrontal

cortex to the right anterior insula. Dai X J et al. (Dai

and Wang 2021) used Granger causality analysis and

mediating causality analysis to study the relationship

between chronic insomnia and the seeking system

and value-driven attention network, and found that

the value-driven attention network reduced the

mediating effect of sleep regulation and the seeking

system reduced the mediating effect of negative

emotions after insomnia. However, the interaction

between these brain regions with abnormal

functional connectivity and the change of the

interaction relationship after treatment remains

unclear. In this study, a GCA model based on resting

state fMRI data of insomnia patients was established

to compare and analyze the differences of effective

brain connections of insomnia patients before and

after neurofeedback training. This study provides a

new perspective for understanding the pathological

mechanism of insomnia patients, thus promoting the

development of brain network imaging markers for

early diagnosis and treatment evaluation of

insomnia.

2 MATERIALS AND METHODS

2.1 Participants

This experimental design is described in detail in

Zhang's research (Zhang and Gao 2021). The

original study recruited 32 healthy subjects for the

experimental group and 36 healthy subjects for the

control group. Among them, the experimental group

received neurofeedback signals from the amygdala

during training, while the healthy control group

received only one baseline scan but didn’t

participate in neurofeedback training. The two

studies have been approved by the Ethics Committee

of Henan Provincial People's Hospital, and all

volunteers signed the informed consent to participate

in this study.

ICHIH 2022 - International Conference on Health Big Data and Intelligent Healthcare

578

2.2 Experimental Paradigm

The experimental design is described in detail in

Zhang's research paper (Zhang and Gao 2021). In

this experiment, the experimental group completed 6

stages of experiment, while the control group

completed 2 stages of experiment. The experiment

was conducted once a week. During stage 1, general

demographic characteristics of the experimental

group were collected. During stage 2, baseline scans

were performed for both experimental and control

subjects. Stage 3 to 5 were three neurofeedback

sessions for insomnia patients, with each session

lasting 50 minutes. Stage 6 was the follow-up

period.

The specific steps of the three neurofeedback

training sessions are shown in Figure 1. Each run

consisted of 7 "rest" blocks and 6 "happy" blocks

alternately, lasting for 6min30s. In the "happy"

block, participants upregulated the height of the

thermometer on the screen by recalling the specific

positive autobiographical memory they had written

down, and the feedback signal was updated every

repetition time (TR=2s). In the "resting" block, the

"+" appears on the screen and the subject looks at

the "+" calmly to return their brain activity to

baseline levels.

Figure 1: Experimental paradigm of rt-fMRI

neurofeedback training for amygdala.

2.3 Data Acquisition

Behavioral data and related experimental results

from the same sample have been published in an

earlier study. The data analyzed in this paper are

resting state functional images of insomnia patients

collected before Session3 and after Session5 during

neurofeedback training.

The restplus toolbox based on MATLAB was

used to preprocess data (Jia and Wang 2019). The

data preprocessing mainly includes: dicom to nifti,

remove first 10 time points, slice timing,

realign/head motion correction, normalization and

smooth (FWHM: 6mm). Subjects with head

movements greater than 2.5mm and 2.5° were

excluded in this experiment.

2.4 Analysis of fMRI Data

2.4.1 Choice of ROI

Firstly, Regional Homogeneity (ReHo) analysis

were performed on resting state data before and after

neurofeedback training to study the difference in

effective connection of emotion-related brain

regions in insomnia patients before and after

neurofeedback training. Through the paired T-test of

ReHo results, we found that 10 brain regions related

to emotion regulation were significantly changed in

insomnia patients after neurofeedback training. The

results are shown in Table 1, where: Right precuneus

(10, -47, 18), right posterior cingulate gyrus (12, -47,

30), left precuneus (-5, -47, 15), right middle

cingulate gyrus (-4, -12, 32), left insula (-42, 15, 9),

right triangle inferior frontal gyrus (52, 25, 19),

opercular part of inferior frontal gyrus (48, 18, 9),

left middle frontal gyrus (-51, 24, 33), right middle

frontal gyrus (36,39,39), right dorsolateral prefrontal

cortex (24, 6, 57). Therefore, these brain regions

were selected as ROI with a radius of 7 mm for the

next GCA analysis. In addition, since BOLD signal

of the patient's left amygdala was provided to the

subjects during the experiment, the left and right

amygdala were used as ROI, where: left amygdala (-

24, -1, -17) and right amygdala (26, 1, -18), and the

radius was 7 mm.

Table 1: Significant differences in brain regions analyzed

by Reho before and after neurofeedback training.

Brain regions Coordinates

Cluster

size

Peak

intensity

Right

precuneus

(10, -47, 18) 115 4.795

Right

posterior

cingulate

gyrus

(12, -47, 30) 32 4.394

3 Left precuneus (-5, -47, 15) 24 4.916

Right middle

cingulate

gyrus

(-4, -12, 32) 56 3.165

5 Left insula (-42, 15, 9) 36 -5.1656

Right triangle

inferior frontal

gyrus

(52, 25, 19) 51 -3.184

Granger Causality Changes during Rt-fMRI Neurofeedback Training of Emotion Regulation for Insomnia Patients

579

Opercular part

of inferior

frontal gyrus

(48, 18, 9) 45 -4.4361

Left middle

frontal gyrus

(-51, 24, 33) 28 -4.4819

Right middle

frontal gyrus

(36, 39, 39) 21 -3.8115

Right

dorsolateral

prefrontal

cortex

(24, 6, 57) 27 -4.0626

2.4.2 GCA Analysis

The GCA method was first applied to fMRI data for

effective connection analysis by Goebel et al. in

2003(Goebel and Roebroeck 2013). Granger

causality tends to be multi-variable, that is, the

causal relationship between two time series under

the influence of multiple additional time series.

GCA tests whether the current and past values of the

first sequence can better predict the results of the

second sequence by building a linear model. If so,

then these two sequences constitute a causal

relationship.

Specifically, In the Granger causality model, we

take and as generalized prediction errors. Then we

predict two time series and by linear projections of

their respective past values respectively. According

to Geweke's feedback model (Jiao and Zou 2014),

the time dynamics of two time series and with length

T can be described as,

()

11,1 1

() ()

Xt aX t k te

=-+

(1)

()

22,2 2

() ()

Xt aX t k te

=-+

(2)

Where p is the maximum number of lagged

observations included in the model (model orderp<

k ). a

,

and a

,

are autoregressive coefficients.



(t) and ε



(t) are the autoregressive residuals of

each time series respectively.

That is, if the inclusion of X



's past observations

can reduce the prediction error in the linear

regression model of X



andX



compared to a model

containing only previous observations of X



then X



cause toX



. For time series X



(t) and X



(t),

we assume that both X



(t) and can be described as

binary autoregressive models (Jiao Z-Q and Zou L,

2014),

() ()()

111,1 12,21

()

Xt AXtk AXtk et

=-+-+

åå

(3)

() ()()

221,1 22,22

()

tAXtkAXtket

=-+-+

åå

(4)

Where p is the model order. A

 ,

and A

 ,

are

signed path coefficients. and are autoregressive

coefficients respectively. e



(t) and e



(t) are

residuals of and X



(t), respectively.

GCA is a data-driven method, which overcomes

the limitations of SEM and DCM. The GCA method

does not require a prior hypothesis model, and the

model is relatively simple with low computational

complexity, which can be used to analyze the

effective connection of large brain networks

(Roebroeck and Formisano 2005).

In this study, the time series of the above 12

ROIs related to emotion regulation was extracted to

establish a GCA model based on fMRI. Then the

GCA results of subjects before and after

neurofeedback training were statistically analyzed at

group level, so as to detect the effect of

neurofeedback training on the effective connection

between emotion-related brain regions of subjects.

3 RESULT

The results of GCA-paired T test before and after

neurofeedback training for insomnia patients were

shown in Table 2 and Figure 2. It was found that

after neurofeedback training, the effective

connections from the right triangle inferior frontal

gyrus to the left amygdala, the left precuneus to the

left middle frontal gyrus, and the right middle

cingulate gyrus to the left middle frontal gyrus were

significantly enhanced. The effective connection

from the left middle frontal gyrus to the left

precuneus was significantly reduced. In figure2, Red

‘→’ represents a significant increase of the

effective connection, and blue ‘→’ represents a

significant decrease of the effective connection.

Render visualized using BrainNet Viewer (Xia and

Wang 2013).

Table 2: Significant differences in brain effective

connections before and after neurofeedbacktraining

through Granger causality analysis.

ROI A→ROI B

Right triangle inferior

frontal gyrus→Left

amygdala

2.470863 0.019598

Left precuneus→Left

middle frontal gyrus

2.587577 0.014944

Left middle frontal

gyrus→Left precuneus

-2.685914 0.011842

Right middle cingulate

gyrus→Left middle frontal

gyrus

2.413076 0.022365

ICHIH 2022 - International Conference on Health Big Data and Intelligent Healthcare

580

Figure 2: The changes of effective connections between

emotional brain regions in insomnia patients before and

after neurofeedback training.

4 DISCUSSION

The prefrontal cortex can respond to positive

emotional faces, pleasant stimulating images and

happy memories, which is an atypical reward

mechanism (Vijayakumari and Menon 2020). These

responses can be used to guide emotion and may

play an important role in emotion regulation. The

results of GCA showed that after neurofeedback

training, the effective connection of the right triangle

inferior frontal gyrus to the left amygdala was

enhanced, suggesting that the activity of the

prefrontal lobe directly affects and controlls the

amygdala. This is similar to the previous results of

rt-fMRI neurofeedback emotion regulation by Paret

et al. (Paret and Ruf 2016) in healthy subjects in

2016. Similar results have been obtained in the drug

treatment for patients with depression and other

psychiatric diseases, that is, antidepressants enhance

the activities of the prefrontal lobe in cognitive

control and other tasks, thus promoting the top-down

control of emotions. In 2018, Guo et al. (Guo and

Liu 2018) also found that functional connectivity

between amygdala, medial prefrontal cortex and

inferior frontal gyrus decreased in Alzheimer's

disease patients with depression compared with

healthy controls, proving that the functional

connectivity between amygdala and prefrontal lobe

may be an important feature of AD patients with

depression.

The precuneus, located in the posterior medial

parietal lobe, is an important structure of the

posterior default mode network and is associated

with cognitive functions (self-related information

processing, consciousness, episodic memory

extraction and visuo-spatial imagination) (Cavanna

and Trimble 2006). Research by Quinten Van Geest

et al. (Van Geest and Westerik 2017) has shown that

people with insomnia have reduced FC in the

thalamus, anterior cingulate cortex and precuneus. A

study of 18 healthy women by K Helmbold et al.

(Helmbold and Zvyagintsev 2016) showed that acute

consumption of tryptophan (a precursor of 5-HT

synthesis) led to reduced FC in precuneus and DMN.

Therefore, the decrease of FC in precuneus and

DMN was related to the decrease of 5-HT level. The

results of this study showed that after neurofeedback

training, the sending ability of left precuneus to left

middle frontal gyrus was enhanced while the

receiving ability was decreased, suggesting that the

improvement of effective connection between left

precuneus and left middle frontal gyrus was related

to the improvement of insomnia symptoms.

The cingulate gyrus is located between the

corpus callosum sulcus and cingulate sulcus, and is

an important part of the limbic system. It was first

defined and named by Professor Broca (Broca

1878), who divided the cingulate gyrus into three

parts by making vertical lines for the center. The

anterior cingulate gyrus (ACC), the middle cingulate

gyrus (MCC) and the posterior cingulate gyrus

(PCC), among which the anterior cingulate gyrus

and the middle cingulate gyrus have similar

functions, so most experiments have studied them as

a functional region. Moreover, studies have found

that the cumulative and negative memories of

objective objects and experiences can activate the

anterior cingulate gyrus and the middle cingulate

gyrus. Li et al. (Li and Yan 2018) used Dartel-VBM

technology to evaluate the changes of gray matter in

53 healthy subjects and 60 patients with primary

insomnia, and found that the gray matter volume in

the lateral prefrontal cortex and anterior cingulate

cortex in patients with insomnia was lower than that

in healthy subjects, and was negatively correlated

with the scores of anxiety scales such as SAS and

SDS, which leads to higher levels of negative

emotions such as anxiety and depression. Wang et

al. (Shi and Wang 2021) observed under fMRI plain

scanning that brain activities in bilateral brainstem,

left parahippocampal gyrus, frontal lobe and right

anterior cingulate gyrus of 15 insomnia patients

were significantly more active than before

acupuncture at Shenmen, Sanyinjiajie and Baihui

points 5 weeks after acupuncture. The results of this

study showed that the effective connection between

the left middle cingulate gyrus and the left middle

frontal gyrus was enhanced after neurofeedback

training in insomnia patients, suggesting that the

enhancement of the effective connection between

Granger Causality Changes during Rt-fMRI Neurofeedback Training of Emotion Regulation for Insomnia Patients

581

the left middle cingulate gyrus and the left middle

frontal gyrus was correlated with the improvement

of anxiety, depression and other symptoms, and thus

improved their sleep quality.

5 CONCLUSIONS

In this paper, Granger causality model was

constructed to analyze the difference of causality

between different brain regions before and after

neurofeedback training. The results showed that rt-

fMRI neurofeedback training significantly regulated

the activity of brain regions related to emotion

regulation in insomnia patients. The effective

connections between the prefrontal lobe to the

amygdala, the prefrontal lobe and the precuneus, and

the cingulate gyrus to the prefrontal lobe also

changed. These conclusions were consistent with the

results of the scale before and after the experiment

and were consistent with the results of clinical

psychiatric studies. These results further indicated

that rt-fMRI neurofeedback training can alter the

effective connectivity of brain regions related to

emotion regulation, and this change could be used as

a potential biological marker to evaluate the effect of

neurofeedback training.

Although rt-fMRI neurofeedback training of

amygdala emotion regulation has been studied from

the perspective of effective connection of brain

regions related to emotion regulation, and

preliminary results have been achieved, there are

still some limitations. Due to time constraints, the

control group experiment of neurofeedback training

for insomnia patients is still in progress. The data of

the experimental group and the control group can be

compared and analyzed in the future to further

explore the feasibility of this method in disease

treatment.

ACKNOWLEDGEMENTS

This research was supported by the National Natural

Science Foundation of China under grant 82071884

and the National Natural Youth Foundation of China

under grant 62106285.

REFERENCES

A research on resting-state functional network

connectivity after rt-fMRI neurofeedback in

insomnia[J]. Journal of Physics: Conference Series,

2021, 1907(1): 012013 (7pp).

Abraham A, Kaufmann C, Redlich R, et al. Self-referential

and anxiety-relevant information processing in

subclinical social anxiety: an fMRI study[J]. Brain

Imaging & Behavior, 2013, 7(1): 35-48.

Bodurka, Jerzy, Siegle, et al. Randomized Clinical Trial of

Real-Time fMRI Amygdala Neurofeedback for Major

Depressive Disorder: Effectson Symptoms and

Autobiographical Memory Recall[J]. American

journal of psychiatry, 2017, 174(8): 748-755.

Bonnet M H, Arand D L. Hyperarousal and insomnia:

State of the science[J]. Sleep Medicine Reviews, 2010,

14(1): 9-15.

Broca P. Le grand lobe limbique et la scissure limbique

dans la série des mammifères[J]. Revue

d'Anthropologie, 1878, 2: 385-498.

Bruce P, Doré, Odile, et al. Negative Autobiographical

Memory in Depression Reflects Elevated Amygdala-

Hippocampal Reactivity and Hippocampally

Associated Emotion Regulation[J]. Biological

psychiatry. Cognitive neuroscience and neuroimaging,

2018.

Brühl A. Making Sense of Real-Time Functional Magnetic

Resonance Imaging (rtfMRI) and rtfMRI

Neurofeedback[J]. International Journal of

Neuropsychopharmacology, 2015, 18(6).

Cavanna A E, Trimble M R. The precuneus: a review of

its functional anatomy and behavioural correlates[J].

Brain, 2006, 129(3): 564-583.

CHEN Bai-ru.Real-time functional magnetic resonance

neurofeedback technique in intervention of insomnia

patients disorder with depression

state.2021,35(04):401-404.

Christopher, Decharms. Applications of real-time fMRI[J].

Nature Reviews Neuroscience, 2008.

Dai X J, Wang N, Ai S Z, et al. Decreased modulation of

segregated SEEKING and selective attention systems

in chronic insomnia[J]. Brain Imaging and Behavior,

2021, 15(1): 430-443.

Friston K J, Frith C D, Frackowiak R S J. Time ‐

dependent changes in effective connectivity measured

with PET[J]. Human Brain Mapping, 2010, 1(1): 69-

79.

Goebel R, Roebroeck A, Kim D S, et al. Investigating

directed cortical interactions in time-resolved fMRI

data using vector autoregressive modeling and

Granger causality mapping[J]. Magnetic Resonance

Imaging, 2003, 21(10): 1251-1261.

Guo Z, Liu X, Xu S, et al. Abnormal changes in functional

connectivity between the amygdala and frontal regions

are associated with depression in Alzheimer’s

disease[J]. Neuroradiology, 2018, 60(12): 1315-1322.

Helmbold K, Zvyagintsev M, Dahmen B, et al.

Serotonergic modulation of resting state default mode

network connectivity in healthy women[J]. Amino

Acids, 2016, 48(4): 1109-1120.

Jia X Z, Wang J, Sun H Y, et al. RESTplus: an improved

toolkit for resting-state functional magnetic resonance

imaging data processing[J]. Science Bulletin, 2019.

ICHIH 2022 - International Conference on Health Big Data and Intelligent Healthcare

582

Jiao Z-Q, Zou L, Cao Y, et al. Effective connectivity

analysis of fMRI data based on network motifs[J]. The

Journal of Supercomputing, 2014, 67(3): 806-819.

Li C, Dong M, Yi Y, et al. Aberrant Effective

Connectivity of the Right Anterior Insula in Primary

Insomnia[J]. Frontiers in Neurology, 2018, 9: 317-.

Li M, Yan J, Li S, et al. Altered gray matter volume in

primary insomnia patients: a DARTEL-VBM study[J].

Brain imaging and behavior, 2018, 12(6): 1759-1767.

Liang X, Wang J H, He Y. Human connectome: Structural

and functional brain networks (in Chinese). Chinese

Sci Bull (Chinese Ver), 2010, 55:

Mehler D, Sokunbi M O, Isabelle H, et al. Targeting the

affective brain—a randomized controlled trial of real-

time fMRI neurofeedback in patients with

depression[J]. Neuropsychopharmacology, 2018.

Paret C, Ruf M, Gerchen M F, et al. FMRI neurofeedback

of amygdala response to aversive stimuli enhances

prefrontal-limbic brain connectivity[J]. Neuroimage,

2016, 125: 182-188.

Roebroeck A, Formisano E, Goebel R. Mapping directed

influence over the brain using Granger causality and

fMRI[J]. Neuroimage, 2005, 25(1): 230-242.

Shi X H, Wang Y K, Li T, et al. Gender ‐ related

difference in altered fractional amplitude of low ‐

frequency fluctuations after electroacupuncture on

primary insomnia patients: A resting ‐ state fMRI

study[J]. Brain and Behavior, 2021, 11(1): e01927.

Van Geest Q, Westerik B, Van Der Werf Y, et al. The role

of sleep on cognition and functional connectivity in

patients with multiple sclerosis[J]. Journal of

neurology, 2017, 264(1): 72-80.

Vijayakumari A A, Menon R N, Thomas B, et al.

Glutamatergic response to a low load working

memory paradigm in the left dorsolateral prefrontal

cortex in patients with mild cognitive impairment: a

functional magnetic resonance spectroscopy study[J].

Brain imaging and behavior, 2020, 14(2): 451-459.

Xia M, Wang J, He Y. BrainNet Viewer: a network

visualization tool for human brain connectomics[J].

PloS one, 2013, 8(7): e68910.

Young E, Korszun A. Sex, trauma, stress hormones and

depression[J]. Molecular psychiatry, 2010, 15(1): 23-

28.

Zweerings J, Hummel B, Keller M, et al. Neurofeedback

of core language network nodes modulates

connectivity with the default-mode network: a double-

blind fMRI neurofeedback study on auditory verbal

hallucinations[J]. NeuroImage, 2019, 189: 533-542.

Granger Causality Changes during Rt-fMRI Neurofeedback Training of Emotion Regulation for Insomnia Patients

583