Frontal, temporal, cerebellar changes link to sepsis survivors' cognitive issues: A resting state functional magnetic resonance imaging study

Abstract

BACKGROUND

Sepsis is a life-threatening condition defined by organ dysfunction, triggered by a dysregulated host response to infection. there is limited published literature combining cognitive impairment with topological property alterations in brain networks in sepsis survivors. Therefore, we employed graph theory and Granger causality analysis (GCA) methods to analyze resting-state functional magnetic resonance imaging (rs-fMRI) data, aiming to explore the topological alterations in the brain networks of intensive care unit (ICU) sepsis survivors. Using correlation analysis, the interplay between topological property alterations and cognitive impairment was also investigated.

AIM

To explore the topological alterations of the brain networks of sepsis survivors and their correlation with cognitive impairment.

METHODS

Sixteen sepsis survivors and nineteen healthy controls from the community were recruited. Within one month after discharge, neurocognitive tests were administered to assess cognitive performance. Rs-fMRI was acquired and the topological properties of brain networks were measured based on graph theory approaches. GCA was conducted to quantify effective connectivity (EC) between brain regions showing positive topological alterations and other regions in the brain. The correlations between topological properties and cognitive were analyzed.

RESULTS

Sepsis survivors exhibited significant cognitive impairment. At the global level, sepsis survivors showed lower normalized clustering coefficient (γ) and small-worldness (σ) than healthy controls. At the local level, degree centrality (DC) and nodal efficiency (NE) decreased in the right orbital part of inferior frontal gyrus (ORBinf.R), NE decreased in the left temporal pole of superior temporal gyrus (TPOsup.L) whereas DC and NE increased in the right cerebellum Crus 2 (CRBLCrus2.R). Regarding directional connection alterations, EC from left cerebellum 6 (CRBL6.L) to ORBinf.R and EC from TPOsup.L to right cerebellum 1 (CRBLCrus1.R) decreased, whereas EC from right lingual gyrus (LING.R) to TPOsup.L increased. The implementation of correlation analysis revealed a negative correlation between DC in CRBLCrus2.R and both Mini-mental state examination (r = -0.572, P = 0.041) and Montreal cognitive assessment (MoCA) scores (r = -0.629, P = 0.021) at the local level. In the CRBLCrus2.R cohort, a negative correlation was identified between NE and MoCA scores, with a statistically significant result of r = -0.633 and P = 0.020.

CONCLUSION

Frontal, temporal and cerebellar topological property alterations are possibly associated with cognitive impairment of ICU sepsis survivors and may serve as biomarkers for early diagnosis.

Key Words: Sepsis; Cognitive impairment; Functional magnetic resonance imaging; Graph theory; Granger causality analysis; Topological properties

Core Tip: This study innovatively integrates graph theory and Granger causality analysis to uncover distinct topological alterations in frontal, temporal, and cerebellar networks of intensive care unit sepsis survivors. Key findings reveal compensatory hyperconnectivity in the right cerebellum Crus 2, inversely linked to cognitive scores, alongside reduced efficiency in frontal and temporal hubs. Disrupted cerebro-cerebellar directional connectivity further highlights network reorganization post-sepsis. These region-specific network deviations, particularly cerebellar dynamics, offer novel neuroimaging biomarkers for early detection of sepsis-associated cognitive impairment, bridging neuropathological mechanisms with clinical outcomes.

INTRODUCTION

Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection[1]. According to epidemiological data of World Health Organization, 19.4 million people develop sepsis each year[2]. Although the use of antimicrobials and other appropriate management strategies has significantly reduced the incidence and mortality of sepsis, long-term sequela, including physical, cognitive, and psychological impairments, have become another pressing issue. Among these various sequela, cognitive impairment is particularly common[3], significantly decreasing the life quality for patients and increasing the burden on caregivers[4,5]. However, the possible brain alterations underlying cognitive impairment in sepsis survivors remains unclear.

Magnetic resonance imaging (MRI) is a non-invasive and non-radioactive tool widely used in the field of neurocognitive impairment. Specifically, functional MRI (fMRI) is based on the principle that the blood oxygen level-dependent (BOLD) signal fluctuations strongly reflect neuronal activity. With numerous analytical methods, resting-state fMRI (rs-fMRI) can be used to explore the intrinsic activity and connectivity of the brain networks of both healthy individuals and patients[6]. Graph theory provides a mathematical framework that is instrumental in analyzing rs-fMRI data. In the brain networks, brain regions are considered as nodes while the connections between these nodes are edges. This allows for the construction of the brain topological architecture and the quantification of its topological properties at both global and local levels[7]. As revealed by graph theoretical analysis, healthy human brain networks exhibit a “small-world” organization which enables efficient information processing at low energy costs[8]. Small-world topological abnormalities have been repeatedly implicated in cognitive impairment in numerous diseases[9-11]. In addition to functional connectivity, effective connectivity (EC) can also be applied in rs-fMRI data analysis to reveal the neural underpins of brain disorders. It identifies the intensity and direction the information flow, thus determining interaction and causal influence between brain regions[12].

From the perspectives of topological properties and cognitive function, research has demonstrated that compared to healthy controls, patients with mild cognitive impairment display compensatory changes within the frontal lobe as well as alterations in the cortex - subcortical circuitry associated with cognitive processes[13]. A study has linked unilateral cerebellar infarction to diminished network efficiency in key regions across both cerebral hemispheres, with these alterations being strongly associated with cognitive impairment[14]. For mild cognitive impairment, the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) are widely used screening tools. While both assess various cognitive domains such as orientation, attention, memory, language, and visuospatial abilities, the MoCA places a greater emphasis on executive function, language comprehension, and visuospatial skills. Both instruments are well- suited for evaluating mild cognitive impairment[15-17].

To date, however, there is limited published literature combining cognitive impairment with topological property alterations in brain networks in sepsis survivors. Therefore, we employed graph theory and Granger causality analysis (GCA) methods to analyze rs-fMRI data, aiming to explore the topological alterations in the brain networks of intensive care unit (ICU) sepsis survivors. Using correlation analysis, the interplay between topological property alterations and cognitive impairment was also investigated.

MATERIALS AND METHODS

The study was approved by ethics committee of Jiangyin People's Hospital Affiliated with Southeast University and registered in the clinical trials.gov (NCT03946839).

Participants

A total of 24 ICU sepsis survivors were initially recruited for this study and underwent neurocognitive testing and MRI scanning. Patients were recruited from the ICU of Jiangyin People's Hospital Affiliated with Southeast University. 24 age-, gender-, and education-matched healthy volunteers were selected as the healthy controls through community postings and media advertising. All participants were right-handed Chinese Han individuals. After excluding participants with incomplete MRI scans, a history of epilepsy, abnormal MRI report or excessive head motion, sixteen sepsis survivors and nineteen healthy controls were included in the final analyses. All participants provided written informed consent.

The inclusion criteria for sepsis survivors included the following: (1) Sepsis or septic shock patients (diagnosed with Sepsis-3.0)[1] aged 18 to 79 years; (2) At least 6 years of education (able to speak, read, and write); and (3) Hospitalized in the ICU for more than 2 days, with survival and subsequent discharge. Exclusion criteria for the sepsis survivor group were: (1) Death during hospitalization; (2) Leaving against medical advice or transfer to another hospital; (3) Declining to participate or withdrawing midway; (4) History of neuropsychiatric disease [e.g., cerebrovascular disease, Parkinson's disease (PD), Alzheimer's disease (AD), demyelinating disease, epilepsy, depression, schizophrenia, etc.]; (5) Severe brain injury; (6) Severe systemic disease (e.g., hepatic encephalopathy, ketoacidosis, hyperosmolar hyperglycemic state, chronic renal failure, etc.); (7) History of drug abuse or insobriety; (8) Use of psychotropic medications such as sleeping pills, selective serotonin reuptake inhibitors, etc.; and (9) MRI incompatibility. Healthy controls were required to have a MMSE score ≥ 24[18]. Exclusion criteria for the health controls were a history of neuropsychiatric disease, head injury, alcohol and drug addiction or ferrous/electronic implants.

Neurocognitive measurements

All subjects underwent a comprehensive battery of neurological examinations 2 hours before MRI scanning, including the MoCA, MMSE, complex figure test (CFT), auditory verbal learning test (AVLT), digit span test (DST), verbal fluency test (VFT), clock drawing test, symbol digit modalities test (SDMT), and trail making test (TMT). Each subject had 70-90 minutes to complete the neuropsychological examinations.

MRI data acquisition

MRI was performed at the Medical Imaging Center of Jiangyin People’s Hospital using a Discovery MR750w 3.0T scanner (GE, Boston, United States), equipped with a 24-channel head coil. Head motion was controlled using foam pads during the scans. All participants, wearing earplugs, were instructed to remain awake, motionless with their eyes closed, and not to think of anything in particular. Resting state BOLD images were acquired over a period of 7 minutes and 40 seconds by a gradient-recalled echo-planar imaging sequence with repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, flip angle (FA) = 90°, field of view (FOV) = 220 mm × 220 mm, matrix size = 64 × 64, slice thickness = 4 mm, gap = 0 mm, number of slices = 35. The slices were acquired in an interleaved order (1, 3, 5 …, 35, 2, 4, 6 …, 34). T1-weighted 3D fast spoiled gradient recalled echo images were collected over 4 minutes 25 seconds with TR = 7.2 ms, TE = 3.1 ms, FA = 8°, FOV = 256 mm × 256 mm, matrix size = 256 × 256, slice thickness = 1 mm, gap = -0.5 mm, number of slices = 312, voxel size = 1 mm × 1 mm × 1 mm. Additionally, routine axial T2-weighted images were obtained to exclude subjects with major cerebral infarction, white matter changes, or other brain lesions.

fMRI image data preprocessing

SPM12 (http://www.fil.ion.ucl.ac.uk/spm) and DPABI (http://rfmri.org/dpabi) were used for preprocessing rs-fMRI data in the MATLAB environment. The first 10 images were removed to allow steady state, leaving 220 functional volumes for each subject. After slice timing correction, images were realigned to the first volume to correct for head motion. After excluding 1 healthy participant and 4 sepsis patients with head motion > 2.0 mm maximum displacement in any direction (x, y, z) or 2.0° of angular motion, no significant differences in head motion were found between the two groups (P > 0.05). Then, 3D T1-weighted images were registered to the functional images and subdivided into white matter, gray matter, and cerebrospinal fluid (CSF) using the new segment and DARTEL technique, followed by spatial normalizing into the Montreal Neurological Institute EPI template (voxel size 3 mm × 3 mm × 3 mm). Other preprocessing steps included spatial smoothing with an isotropic Gaussian kernel of 6 mm × 6 mm × 6 mm, temporal bandpass filtering at 0.01-0.08 Hz, and nuisance signal regression (including white matter signal, CSF signal, and Friston-24 head motion parameters).

Network construction and topological analysis

Preprocessed rs-fMRI data were used to construct the whole brain functional connectivity network for each subject. We used automated anatomical labeling atlas[19] to identify 116 functional regions of interest (ROIs) throughout the brain. Each brain region was considered as a network node. The time series of all voxels in each ROI were extracted and subsequently averaged to obtain a representative time series. Pearson’s correlation coefficients between the mean time series of all possible pairs of the 116 regions were computed, which were considered as the edges of the network. To improve the distribution of data for group analysis, Pearson correlation coefficients (r) were standardized by Fisher’s z transformation, resulting in a 116 × 116 correlation weight matrix for each subject.

Topological properties of networks were analyzed using GRETNA software (http://www.nitrc.org/projects/gretna/) based on graph theory. Only positive correlations were involved in the subsequent network metrics analysis to minimize potential confounding effects of global signal regression. A network sparsity (S) was applied to produce binary undirected functional networks, and a wide range of sparsity threshold was identified, ranging from 0.05 to 0.4 with an interval of 0.01. The global and local metrics of brain functional network were estimated for each brain region at each selected sparsity threshold. Global metrics included small-world parameters [clustering coefficient (Cp), characteristic path length (Lp), normalized clustering coefficient (γ), normalized characteristic path length (λ), and small-worldness (σ) and network efficiency (global efficiency (Eg) and local efficiency (Eloc)]. Specifically, σ = γ/λ, and σ > 1 was used as an indication of a small-world organization of the network. Two nodal centrality properties were employed for regional nodal network analysis: Degree centrality (DC) and nodal efficiency (NE). For further statistical comparison, area under the curve (AUC) for each network metric was calculated, providing a summarized scalar for topological parameters, avoiding the error caused by a single threshold.

GCA

GCA is a statistical concept of causality that is based on the multiple variant auto regression model[12]. By extracting temporal dynamics from rs-fMRI signals, GCA is used to estimate EC, which characterizes directional functional connections among brain regions[20]. In this study, GCA was conducted based on those brain regions showing significant DC or NE changes between sepsis survivors and healthy controls. Those brain regions were selected and defined as ROIs for seeds using WFU-pick-atlas (https://www.nitrc.org/projects/wfu_pickatlas), then resampled to 3 mm × 3 mm × 3 mm. Based on pre-processed data, the REST_v1.8 software package (https://rfmri.org/REST) was applied to compute the causal effects of the time series x of selected seed points and the time series y of each voxel over the whole brain. In the Granger causality model, a value of 0 indicates no causal connection from x to y, a value of 1 indicates strong positive causality, and a value of -1 indicates strong negative causality. The causality analysis was performed twice on each ROI: From the seed point to whole-brain voxels (x to y) and from whole-brain voxels to the seed point (y to x). The obtained EC graph was transformed by Fisher’s z to improve distribution normality, resulting in a Z-map of GCA.

Statistical analysis

Demographic and neurocognitive data: Statistical analyses were performed using SPSS software (version 27.0.1, IBM SPSS Inc., Chicago, IL, United States). Categorical variables were compared between groups using the χ² test or Fisher’s exact test, as appropriate. Continuous variables, including demographic data and cognitive scores, were analyzed using independent t-tests. A P value less than 0.05 was considered statistically significant.

rs-fMRI data: The topological properties of brain functional networks were analyzed using GRETNA software. Between-group comparisons of the AUC for these topological properties were conducted using one-way ANCOVA in R software (version 4.3.2; https://www.r-project.org/), with age, gender, and years of education included as covariates. Multiple comparisons were corrected using the false discovery rate (FDR) method.

GCA: Granger causal influence measures (Z-EC values) obtained from healthy control subjects and sepsis survivors were compared using two-sample t-tests, with age, gender, and years of education as covariates. Statistical significance was determined using Alphasim correction with a voxel-level threshold of P < 0.01 and a cluster-level threshold of P < 0.01 (two-tailed).

Correlation analysis: In the sepsis survivor group, Pearson correlation analyses were performed to assess the relationships between cognitive test scores and topological properties, controlling for age, gender, and years of education. These analyses were conducted using SPSS software, and a P value less than 0.05 was considered statistically significant.

RESULTS

Demographic and neuropsychological data

A total of sixteen ICU sepsis survivors and nineteen healthy controls were included in the final analysis (Supplementary Figure 1). Most of the survivors suffered from severe abdominal infections (Table 1). There were no statistical differences in gender, age, education years, diabetes and hypertension between sepsis survivors and healthy controls (Table 1). Sepsis survivors had a significant lower body mass index (Table 1, t = -2.914, P = 0.006) likely due to weight loss during their ICU stay. In terms of cognitive function, sepsis survivors scored significant lower on the MoCA, MMSE, CFT-I, CFT-D, AVLT-N4, AVLT-N5, DST, VFT and SDMT. Conversely, sepsis survivors had higher completion times on the TMT-A and TMT-B tests compared to healthy controls (Figure 1).

Figure 1 Graphical representation of the comparisons of cognitive performance between groups. SS: Sepsis survivor; HC: Healthy control; MoCA: Montreal Cognitive Assessment; MMSE: Mini-Mental State Examination; CFT-I: Complex figure test-immediate; CFT-I: Complex figure test-delay; AVLT-N4: Auditory verbal learning test; AVLT-N5: Auditory verbal learning test; DST: Digit span test; VFT: Verbal f1uency test; CDT: Clock drawing test; SDMT: Symbol digit modalities test; TMT-A: Trail making test-part A; TMT-B: Trail making test-part B.

Table 1 Demographic and clinical data of sepsis survivors and healthy controls, mean ± SD.

Characteristics	SS (n = 16)	HC (n = 19)	χ2/t	P value
Sex, female/male	8/8	8/11	0.218	0.640
Age, years	62.00 ± 9.42	63.32 ± 8.85	-0.426	0.673
Education, years	9.25 ± 3.98	9.11 ± 3.76	0.111	0.913
BMI, kg/m2	21.27 ± 2.82	23.75 ± 2.23	-2.914	0.006a
Diabetes, yes/no	3/13	0/0	NA	0.086
Hypertension, yes/no	7/9	6/13	0.551	0.458

^aP < 0.05.

SS: Sepsis survivor; HC: Healthy control; NA: Not available; BMI: Body mass index.

Comparison of topological properties of functional network

Global network properties: Seven topological small-world parameters (Figure 2) were analyzed across a sparsity range of 0.05 to 0.40 with an interval of 0.01.γ, λ, σ, and Lp were negatively (Figure 2A-C, and E) whereas Cp, Eg and Eloc (Figure 2D, F, and G) were positively correlated with sparsity. The σ values for both sepsis survivors and healthy controls were greater than 1, indicating that both groups exhibited a small-world organization pattern. However, the γ value of sepsis survivors was significantly lower than that of healthy controls (Figure 2A, F = 7.807, P = 0.009, FDR). Meanwhile, the σ value for sepsis survivors was significantly lower than that of healthy controls (Figure 2C, F = 7.494, P = 0.010, FDR). No differences of λ, Cp, Lp, Eg, Eloc were observed between the two groups (Figure 2B, D, and E-G).

Figure 2 Changes of global properties of the functional brain networks with sparsity and the comparison of area under the curve of global properties. A: γ of two groups at selected sparsity and comparison the area under the curve (AUC) of γ; B: Λ of two groups at selected sparsity and comparison the AUC of λ; C: Σ of two groups at selected sparsity and comparison the AUC of σ; D: Cp of two groups at selected sparsity and comparison the AUC of Cp; E: Lp of two groups at selected sparsity and comparison the AUC of Lp; F: Eg of two groups at selected sparsity and comparison the AUC of Eg; G: Eloc of two groups at selected sparsity and comparison the AUC of Eloc. SS: Sepsis survivor (orange); HC: Healthy control (blue); AUC: Area under the curve; Cp: Clustering coefficient; Lp: Shortest path length; Eg: Global efficiency; Eloc: Local efficiency. The left column stood for the sparsity range (0.05-0.4). The right column is the area under the curve. ^aP < 0.05.

Local network properties: At the local level, significant differences between groups were found in a temporal, a frontal and a cerebellar region. In detail, in a frontal region ORBinf.R, sepsis survivors had significantly decreased DC (P = 0.003, FDR, Table 2, Figure 3A) and NE (P = 0.011, FDR, Table 2). In a temporal region TPOsup.L, sepsis survivors also had significantly decreased NE (P = 0.039, FDR, Table 2, Figure 3B). Conversely, in a cerebellar region CRBLCrus2.R, sepsis survivors had significantly increased DC (P < 0.001, FDR, Table 2, Figure 3A) and NE (P = 0.011, FDR, Table 2, Figure 3B).

Figure 3 Brain regions with positive degree centrality and nodal efficiency alterations. Blue or orange node represents a decreased or increased degree centrality/nodal efficiency value respectively. ORBinf.R, right orbital inferior frontal gyrus; CRBLCrus2.R, right cerebellum Crus 2; TPOsup.L, left temporal pole of superior temporal gyrus. A: Compared to healthy controls, sepsis survivors had decreased DC in ORBinf.R and increased DC in CRBLCrus2.R; B: Sepsis survivors had decreased NE in ORBinf.R and TPOsup.L and increased NE in CRBLCrus2.R. DC: Degree centrality; NE: Nodal efficiency.

Table 2 Brain regions with significant nodal properties differences between sepsis survivors and healthy controls.

Brain region	Lobe	AAL	Nodal parameter	SS (n = 16)	HC (n = 19)	F	P (FDR)
ORBinf.R	Right frontal gyrus	16	DC	6.407 ± 3.543	12.093 ± 3.323	21.866	0.003
ORBinf.R	Right frontal gyrus	16	NE	0.168 ± 0.030	0.208 ± 0.019	19.165	0.011
TPOsup.L	Left temporal gyrus	83	NE	0.172 ± 0.034	0.205 ± 0.020	12.492	0.039
CRBLCrus2.R	Right cerebellar lobule VII	94	DC	14.953 ± 2.512	10.312 ± 2.586	31.422	< 0.001
CRBLCrus2.R	Right cerebellar lobule VII	94	NE	0.223 ± 0.013	0.200 ± 0.019	17.986	0.011

DC: Degree centrality; NE: Nodal efficiency; SS: Sepsis survivor; HC: Healthy control; AAL: Automated anatomical labeling; FDR: False discovery rate.

GCA of brain networks: Based on the above results, we then chose ORBinf.R, TPOsup.L and CRBLCrus2.R as seed regions for further analysis of EC between these three regions and all the other regions in the whole brain. Several correlated regions were identified to have positive or negative EC between seed regions (Table 3). In sepsis survivor group, EC from CRBL6.L (Figure 4A) to ORBinf.R evidently decreased (Figure 4B). The EC from TPOsup.L (Figure 4C) to CRBLCrus1.R also decreased (Figure 4D). The EC from LING.R (Figure 4E) to TPOsup.L increased (Figure 4F). The EC between CRBLCrus2.R and other regions generated no significant results.

Figure 4 Granger causality analysis of effective connectivity between regions with abnormal local properties and other regions in the whole brain. ORBinf.R, right orbital inferior frontal gyrus; CRBL6.L, left cerebellum 6; TPOsup.L, left temporal pole of superior temporal gyrus; CRBLCrus1.R: Right cerebellum Crus 1; LING.R, right lingual gyrus. Orange dots indicated ROIs, i.e., regions with abnormal local properties which were chosen as seed regions. Blue dots indicate correlated regions. The effective connectivity (EC) between correlated regions and the seed regions shows significant alterations. Blue and gray curved arrows indicated positive and negative causality respectively. The direction of the arrows indicated the EC directions. A: Compared to healthy controls, the EC between CRBL6.L and ORBinf.R in sepsis survivors significantly changed; B: The EC from CRBL6.L to ORBinf.R in sepsis survivors decreased; C: Compared to healthy controls, the EC between TPOsup.L and CRBLCrus1.R in sepsis survivors significantly changed; D: The EC from TPOsup.L to CRBLCrus1.R in sepsis survivors significantly decreased; E: The EC from LING.R to TPOsup.L in sepsis survivors significantly changed; F: The EC from LING.R to TPOsup.L in sepsis survivors significantly increased.

Table 3 The increased or decreased effective connection from the seed regions to whole brain (x-y) or from whole brain to the seed regions (y-x).

Direction	AAL	Correlated Region	Peak MNI coordinate			Voxels	t value (Alphasim)
			X	Y	Z
y-16	99	CRBL6.L	-15	-60	-27	87	-5.25
83-y	92	CRBLCrus1.R	24	-75	-30	61	-4.27
y-83	48	LING.R	12	-36	0	43	4.89

AAL: Automated anatomical labeling; MNI: Montreal neurological institute.

Correlation analysis between cognitive performance and network properties

No significant correlations were observed between global network properties and cognitive performance. At the local level, DC in CRBLCrus2.R was negatively correlated with both MMSE (r = -0.572, P = 0.041, Figure 5A) and MoCA scores (r = -0.629, P = 0.021, Figure 5B). NE in CRBLCrus2.R was negatively correlated MoCA scores (r = -0.633, P = 0.020, Figure 5C). No significant correlations were found between EC and cognitive performance.

Figure 5 Correlation analysis between cognitive performance and topological properties. A: Degree centrality (DC) value of CRBLCrus2.R negatively correlated with Mini-Mental State Examination score; B: DC value of CRBLCrus2.R negatively correlated with Montreal Cognitive Assessment (MoCA) scores; C: Nodal efficiency value of CRBLCrus2.R negatively correlated with MoCA scores. MoCA: Montreal Cognitive Assessment; DC: Degree centrality; NE: Nodal efficiency; MMSE: Mini-Mental State Examination.

DISCUSSION

Sepsis survivors often experience long-term disabilities, including cognitive impairment, which significantly impact their quality of life and ability to return to everyday activities[21]. In the present study, we first demonstrated that sepsis survivors experienced significant impairment in various cognitive facets within one month after ICU discharge. Additionally, we observed that, at the global level, sepsis survivors had degraded small-worldness. Local properties decreased in frontal and temporal regions but increased in the cerebellum, which are involved in the cognitive impairment of sepsis survivors. Furthermore, we identified alterations in connections between the cerebrum and the cerebellum, which might also contributed to cognitive impairment in sepsis survivors. Finally, correlation analysis revealed an interplay between cognitive impairment and topological alterations in brain networks.

The majority of sepsis survivors in our study experienced cognitive impairment. Specifically, 15 and 12 out of the 16 (93.8% and 75.0%) sepsis survivors obtained MoCA and MMSE scores below the cut-off for normal performance, respectively. In a Japanese multicenter observational study, the prevalence of cognitive impairment was 37.5% in ICU survivors after 6 months[22]. A recent study of severe coronavirus disease 2019 patients who survived ICU found that 53.4% scored below the cut-off for normal performance on the MoCA, and 19% scored below the threshold for mild cognitive impairment 6 months after ICU discharge[23]. Cognitive impairment is thus common following sepsis, with its severity related to the recency of septic shock. In our study, the majority of the survivors suffered from severe abdominal infections, which are common causes of sepsis in the ICU[24]. It has also been suggested that severe abdominal infections undergoing surgical treatment tended to have poor long-term outcomes[25]. A recent meta-analysis found a significant association between severe sepsis and an increased risk of cognitive impairment, indicating that the specific cause of sepsis also influences the severity of cognitive impairment[26].

At the global level, we found that the small-worldness of sepsis survivors’ brain networks remained but decreased significantly compared to healthy controls. Among the global parameters, γ represents the local characteristics of the network, quantifying the local segregation function of the brain network. λ reflects the capability of global information integration and transmission. The ratio of γ to λ, denoted as σ, represents small-worldness. Compared to a random network, a small-world network has a smaller λ and a larger +γ, and a σ greater than 1. In our study, the σ values of both sepsis survivors and healthy controls were greater than 1, indicating that sepsis survivors retained a small-world organization pattern. However, the lower σ value in sepsis survivors suggested loss of small-worldness of the brain networks. Specifically, the γ and σ values of sepsis survivors were statistically lower than those of healthy controls, while the λ values were similar between the two groups. According to the definition of σ, it can be inferred that the decrease in σ was primarily due to the decrease in γ. Additionally, the similar λ values between the two groups demonstrated that the long-distance information transmission in the brain networks of sepsis survivors remained intact. On the other hand, the decreased γ demonstrated a reduced capacity for local information processing reflecting degraded local connections and grouping of neural units in the brain networks. It is also noteworthy that the differences in γ and σ between the groups were more pronounced when the sparsity threshold was smaller. Therefore, a small threshold would be recommended when using these small-world properties to predict network disruption. In summary, sepsis survivors exhibited degraded small-worldness, primarily due to a decreased capacity for local information processing.

We then demonstrated local information processing capability decreased primarily in a frontal and a temporal region but increased in a cerebellar region. Specifically, both DC and NE decreased in ORBinf.R, NE was decreased in TPOsup.L, and both DC and NE increased in CRBLCrus2.R. DC measures how many edges a node possesses within the entire network, with a higher DC representing a node’s central hub role for information communication[27]. A decrease in DC indicates a reduction in the number of brain regions connected to this region. NE quantifies how efficiently a node can exchange information within the network, so a decrease in NE reflects attenuated information transmission efficiency in that region. Both the ORBinf.R and TPOsup.L are implicated in cognitive function. ORBinf.R, also known as the right pars orbitalis, refers to the most rostral portion of the inferior frontal gyrus in the frontal lobe. The pars orbitalis is involved primarily in semantic processing in the dominant hemisphere. However, in the non-dominant hemisphere, it plays a role in behavioral and motor inhibition[28] and emotional regulation[29]. ORBinf.R volume loss has been significant in PD patients[30]. Furthermore, the volume of this region is associated with the diagnosis of conduct disorder, which involves various behavioral and emotional problems[31]. Besides, cortical thickness of this region has shown significant correlation with DST scores in PD patients[32], animacy scores in frontotemporal dementia and AD patients[33] and the ability to inhibit ad lib smoking during the smoking relapse analog task in individuals with nicotine dependence[34]. Interestingly, the thickness of this region could also be a predictor of suicide attempts in young major depressive disorder patients[35]. TPOsup.L, which refers to the left anterior end of temporal lobe, belongs to the anterior default mode network and is associated with semantic memory and other cognitive functions[36,37]. In a study of patients with white matter lesions (WMLs), the NE value of TPOsup.L showed significant differences across normal people, WMLs with non-dementia vascular cognitive impairment and WMLs with vascular dementia showed significant difference, indicating the role of the network properties of this region in cognitive function[38]. Thus, we speculate that the local information processing disruptions in frontal and temporal lobes may contribute to the cognitive impairment seen in sepsis survivors. In contrast, we observed that local properties increased in the cerebellum. The cerebellum has long been recognized as a crucial component involved in various cognitive functions[39]. CRBLCrus2.R belongs to lobule VII of the cerebellum. It is critical for language[40,41], visual memory[42,43], working memory[44,45] and spatial memory processing[46]. Importantly, the cerebrocerebellar circuit serves as an efficient pathway for information exchange between the cerebral cortex to the cerebellum. Therefore, we speculated that the increase in local properties in the cerebellum might compensate for the decrease in local properties in the frontal and temporal regions. Interestingly, Zhao et al[47] demonstrated that sepsis enhanced the intrinsic excitability and synaptic transmission of cerebellar Purkinje cells, which might be associated with the increase of local properties in the cerebellum. Although we did not observe significant correlations between DC or NE in ORBinf.R or TPOsup.L and cognitive performance of sepsis survivors, we did find that DC or NE in CRBLCrus2.R negatively correlated with MMSE or MoCA scores. We concluded that disruptions in frontal, temporal and cerebellar regions are involved in the cognitive impairment observed in sepsis survivors.

Using GCA, we further examined the alterations in directional connections between above mentioned regions mentioned above (ORBinf.R, TPOsup.L, and CRBLCrus2.R) and other brain regions. We found that EC from CRBL6.L to ORBinf.R and from TPOsup.L to CRBLCrus1.R decreased. EC from LING.R to TPOsup.L increased. A previous imaging study showed that gray matter atrophy in both CRBL6.L and ORBinf.R were significantly implicated in AD pathology[48]. Maesawa et al[49] investigated the connection between resting-state networks and cognitive performance of healthy individuals. Their results indicated that CRBL6.L, within the higher visual networks, exhibited within-network functional connectivity that was negatively correlated with age. Another study identified that the correlated transfer function connections between CRBLCrus1.R and left insula were the most significant markers for discriminating AD patients from healthy controls[50]. Therefore, the decrease in EC from CRBL6.L to ORBinf.R and from TPOsup.L to CRBLCrus1.R represents critical network alterations in sepsis survivors and might be related to cognitive impairment. The lingual gyrus, part of the occipital lobe, is mainly involved in processing vision, playing a role in logical analysis and encoding visual memories. Specifically, the right lingual gyrus is responsible for the perception and recognition of familiar landmarks and scenes as well as the identification of faces. Anatomically, the lingual gyrus and the temporal pole are connected by inferior longitudinal fasciculus fibers[51]. Given that our results suggested an increased EC from LING.R to TPOsup.L, it can be inferred that LING.R and TPOsup.L functionally interact in the brain network of sepsis survivors. In summary, directional connections particularly between the cerebrum and the cerebellum were disrupted and are implicated in cognitive impairment in sepsis survivors.

There are a few limitations worth noting in this study. Firstly, the small sample size of our study is a disadvantage for controlling interindividual heterogeneity, which could increase potential bias. This might explain why we did not observe significant correlations between decreased temporal and frontal topological properties and cognitive impairment. Secondly, our study only interviewed patients within one month after ICU discharge, a longitudinal follow-up will provide deeper insight into the interplay between topological alterations and cognitive function. Future research could mitigate inter-individual heterogeneity and potential bias by increasing sample size. Also, the use of more precise stratification methods is suggested to guarantee sample representativeness and boost the generalizability of the findings in subsequent studies. Such approaches will enhance the accuracy of probing into the neural basis of cognitive impairment and lay a more solid foundation for devising effective early diagnosis and treatment strategies.

CONCLUSION

Collectively, sepsis survivors suffer from cognitive impairment, which correlates with topological property alterations of their brain networks. These topological alterations may serve as biomarkers for early diagnosis of cognitive impairment in sepsis survivors.