Network Anticorrelation and Attention in Glioma Patients
Laura van Loon
Department of Psychology: Clinical Neuropsychology, University of Amsterdam
Master thesis Clinical Neuropsychology Student number: 10000873
Date: 03-10-2016 Supervisor: H. Feenstra Second Supervisor: J. Murre
2
AbstractIntroduction
This study investigated the negative correlation between the default mode network (DMN) and the frontoparietal network (FPN), and its association with attention in glioma patients. We predicted that a negative correlation exists between the DMN and FPN in this patient group, and that this correlation is negatively correlated with attention.
Methods
We conducted a preoperative language functional magnetic resonance imaging (fMRI) along with three standardized neuropsychological tests measuring attention and memory (Stroop Color Word test, Word Learning Test and Fluency test) in 59 glioma patients. Functional connectivity between the DMN and the FPN was computed based on eight ROIs (regions of interest) for each network using an Automated Anatomical Labeling (AAL) atlas. We calculated the correlation between the DMN and FPN, and referred to it as internetwork connectivity. Confounding effects of age, education, Karnofsky score, tumour volume, and tumour type on the association between internetwork connectivity and cognition were also investigated.
Results
Glioma patients displayed a negative correlation between the DMN and FPN. In this study the negative correlation between the DMN and the FPN is referred to as internetwork connectivity. We only found a significant negative association between attention and the internetwork connectivity when gender was taken into account. When patients with and without attention impairments were compared, no difference in the relationship between attention and internetwork connectivity was found. Men did show a more recognizable trend towards a significant correlation between internetwork connectivity and attention than women.
Discussion
A negative internetwork connectivity was found between the DMN and FPN, but this study found no association between internetwork connectivity of the DMN and FPN, and attention in glioma patients. Only when we took gender into account a association was seen.
3
There was also no difference of internetwork connectivity between patients with and without attention impairments. As this study investigated newly diagnosed patients whose tumour is not as profound as in patients later on in the disease, an association between internetwork connectivity and attention cannot be excluded.4
GlossaryGlioma Tumour that originates from glial cells in the
brain.
Graph Theory Theory behind the mathematical
characterization of a network, encompassing numerous nodes and edges. A graph can be seen as a set of nodes, which represents specific brain regions; edges are the connections between the nodes, and together they give an abstract representation of a real-world system with its corresponding interactions.
Structural connectivity The anatomical white matter links between different brain regions.
Functional connectivity Represents the connectivity between different and sometimes spatially distributed brain regions that share certain functional properties. Functional connectivity is then computed by calculating the statistical interdependencies—temporal correlations of co-activity between regions—of neural activity or related metabolic measures between these regions.
Internetwork connectivity The negative correlation between the default mode network and the frontoparietal network.
fMRI Functional magnetic resonance imaging; The
detection of differences in brain regions activity through their effects on blood flow and oxygenation.
BOLD signal Reflects oxygen flow to different brain areas that are more active, is most commonly used when computing functional connectivity with fMRI.
Resting state fMRI Resting state is when a participant is awake and alert during an fMRI but no cognitive or behavioural activities need to be performed. Task based fMRI Short cognitive or behavioural activities are
performed during an fMRI.
Degree distribution This parameter is the probabilistic distribution of all the degrees—number of edges connected to nodes—in the network
Hubs Highly connected brain regions with a high
degree distribution compared to other regions.
DMN; Default mode network A brain network, which is involved in computations necessary for self-referential thought.
FPN; Frontoparietal network A brain network, which is more involved in cognitive control processes and goal-directed behaviour such as attention, memory and executive functioning.
KPS; Karnofsky performance score A performance scale to determine general well being for tumour patients.
5
IntroductionCognitive deficits in glioma patients
Gliomas, primary brain tumours that originate from the glial cells in the brain, make up almost 60% of the neoplasms that are found in the central nervous system (Markert, 2005; Bondy et al., 2008). Based on the distinction of the World Health Organization (WHO), malignant gliomas can be divided into low grade (LGG; grade II) and high grade gliomas (HGG; grade III & IV).The average overall survival for glioma patients ranges from several months to several years (Young et al., 2015; Stupp, Mason, 2005). Besides the short overall survival time, neurological and cognitive deficits make that a glioma is a difficult disease to treat (Markert, 2005).
When a patient is suspected of having a glioma, radiologic, clinical, and pathological assessments are needed for diagnosis. Medical doctors use MRI and biopsy to locate and diagnose the glioma (Osborn et al., 2015). Also a clinical examination is conducted, because the symptoms often do not correspond with the discovered local radiological lesions. Along with seizures, which occur in 30 to 90 % of the cases in both LGG and HGG, glioma patients present a wide variety of symptoms (Heimans & Reijneveld, 2012; Taphoorn & Klein, 2004). Researchers have reported that most patients have symptoms such as, inter alia, headaches, dizziness, nausea, slowness, hemiparesis, visual loss, stroke-like symptoms, or personality changes (Young et al., 2015; Douw et al., 2010; Giovagnoli, 2012; Taphoorn & Klein, 2004). Aside from these symptoms there are often cognitive deficits. Studies have shown that impairment of cognitive functioning occurs in several domains such as language, memory, visuospatial perception, time-space orientation, attention, and executive function (Tucha et al., 2000; Taphoorn & Klein, 2004; Douw et al., 2010; Derks, Reijneveld & Douw, 2014). Taphoorn and Klein (2004) argue that cognitive deficits may be more profound in low-grade gliomas than in high grade gliomas due to the overshadowing of cognitive deficits by neurological deficits in high grade gliomas. Furthermore, research has shown that in many cases the cognitive deficits present in glioma patients have a greater global character than
6
expected, based solely on the location and size of the tumour (Heimans & Reijneveld, 2012). As a result of the more profound global character of these cognitive deficits, the question has arisen as to whether gliomas could be the cause of wider brain dysfunction, such as global cognitive deficits, in addition to focal disruption.Networks and functional connectivity
Our brain can be regarded as a complex network containing a myriad of links between diverse brain regions. Tucha et al. (2000) reasoned, “Cognitive functions that are associated, via neural networks, with a variety of brain regions and that involve multiple systems, such as memory or executive functions, may be especially vulnerable” (p. 332). This can be interpreted, as cognitive functions that are distributed in different brain regions are vulnerable if the tumour is present in one of these regions. Following this reasoning, cognitive deficits that arise in glioma patients may be due to a disruption within a part of the complex network in their brain. This complex network can be determined using graph theory (Bullmore & Sporns, 2009; Mears & Pollard, 2016). A graph can be seen as a set of nodes, which represents specific brain regions; edges are the connections between the nodes, and together they give an abstract representation of a real-world system with its corresponding interactions (Stam & Reijneveld, 2007; Rubinov & Sporns, 2010; Bullmore & Sporns, 2009).
Two kinds of networks can be computed based on graph theory. The first network that can be defined is based on structural connectivity; these are the anatomical white matter links between different brain regions (Bullmore & Sporns, 2009; Wang et al., 2015). Here the nodes can represent the individual neurons, and the edges the synaptic links between these neurons (Mears & Pollard, 2016). A second network that can be computed is based on functional connectivity. Functional connectivity represents the connectivity between different and sometimes spatially distributed brain regions that share certain functional properties (Bullmore & Sporns, 2009).
To determine functional connectivity, functional magnetic resonance imaging (fMRI) or magnetoencephalography (MEG) is used. With these non-invasive recording techniques,
7
similar patterns of activity in differing brain regions are recorded. Functional connectivity is then computed by calculating the statistical interdependencies—temporal correlations of co-activity between regions—of neural co-activity or related metabolic measures between these regions (Bullmore & Sporns, 2009; Mears & Pollard, 2016; Sporns, Chialvo, Kaiser, & Hilgetag, 2004; Stam & Reijneveld, 2007). Both techniques can be used, although MEG has high temporal resolution and fMRI has high spatial resolution (Derks et al., 2014). The blood-oxygenation level dependent (BOLD) signal, which reflects oxygen flow to different brain areas that are more active, is most commonly used when computing functional connectivity with fMRI (Bressler & Menon, 2010; Derks et al., 2014). Within a functional connectivity network, the nodes represent regions of interest (ROI) and the edges exemplify the functional interdependencies between these nodes (Bullmore, 2009; Rubinov & Sporns, 2010).Previous studies have reported that age, gender, and education have an effect on functional connectivity in healthy participants (Arenaza-Urquijo et al., 2013; Douw et al., 2011; Wu et al., 2011; Tian, Wang, Yan & He, 2010). Cahill (2006) also reported a great effect of gender on cognitive functions such as memory and attention, based on structural brain differences in cognitive regions. In fMRI research resting state is often used. Resting state is when a participant is awake and alert but no cognitive or behavioural activities need to be performed. The detected activity represents spontaneous fluctuations of BOLD signal. On the contrary, Krienen, Yeo and Buckner (2014) found that there was a similar stable spontaneous activity within task-based fMRI. They found a strong correlation between these states and concluded that this stable spontaneous activity represents the resting state fMRI. This means that a task based fMRI can be used equally well in research as a resting state fMRI.
Functional connectivity in the DMN and FPN
To compute a functional connectivity network different parameters can be used. Degree distribution is the most important parameter, and the one other measures are based on. This parameter is the probabilistic distribution of all the degrees—number of edges connected to
8
nodes—in the network (Rubinov & Sporns, 2010; van den Heuvel & Sporns, 2011; van Straaten & Stam, 2013). Sporns, Honey and Kötter (2007) made the assumption that some brain regions are highly connected, because they consist of a high degree distribution compared to other regions. They investigated these highly connected regions and defined them as brain hubs. Two readily investigated networks, which include brain hubs, are the default mode network (DMN) and the frontoparietal network (FPN) (Derks et al., 2014). The DMN is involved in self-referential thought. The FPN, which consists of more frontal and parietal brain regions, is more involved in cognitive control processes and goal-directed behaviour such as attention, memory, and executive functioning (Anticevic et al., 2012).Greicius, Krasnow, Reiss & Menon (2003), who conducted research on functional connectivity between different brain regions involved in the DMN, found that the most involved region is the posterior cingulate cortex (PCC). The ventromedial prefrontal cortex (vmPFC) is also involved in this network. They reported that the DMN was most active during resting state, but that there was also activity visible during passive sensory processing which was replicated by later studies (Greicius et al., 2003; Buckner, Andrews-Hanna, & Schacter, 2008). The role of the DMN in disease and cognition is important when investigating functional connectivity and cognition in glioma patients. Based on its self-referential and musing nature, the DMN can conflict with goal-directed activities such as attention and/or working memory.
A number of studies have reported an anticorrelation or negative correlation between the DMN and other networks, including attention based networks, such as the FPN (Wu et al., 2011, Fornito, Harrison, Zalesky, & Simons, 2012, Uddin et al., 2009). Research suggests that lower DMN activity during a task causes better cognitive performance Cole et al. (2012) indicate that the ability of more frontal networks to suppress the activity of the DMN is correlated with intelligence and goal-directed behaviour. Furthermore, they suggest the negative correlation between the DMN and FPN is the result of a stronger suppression of the DMN and may lead to higher intelligence and better goal-directed behaviour.
9
Weismann, Roberts, Visscher and Woldorff (2006) discovered increased default mode activity in healthy participants that was linked to lapses of attention. In addition, Anticevic et al. (2012) found that schizophrenia patients exhibited a less negative correlation between the DMN and frontal regions, which may be the basis of the cognitive deficits in this patient group. They also argued that the actual lack of suppression of the DMN by the frontal regions is the cause of these deficits. Based on these assumptions, Harris et al. (2013) investigated the functional connectivity of the DMN in diffuse brain tumour patients, and found the integrity of the DMN to be impaired.This body of research supports the idea that the global character of memory and attention deficits in glioma patients is caused by the global brain dysfunction of the tumour in different functional networks. Further studies have reported correlations between changes in functional connectivity in glioma patients and their cognitive dysfunctions (Bartolomei et al., 2006a; Bartolomei et al., 2006b; Bosma et al., 2009). These studies, using MEG, show global changes in connectivity. These global changes can account for the impaired integrity of the networks in glioma patients and the global character of the cognitive deficits. Previous research found that tumour location has a profound effect on cognitive deficits but not on functional connectivity. In contrast, tumour grade does have an effect on the functional connectivity (Bartolomei et al., 2006; Harris et al., 2013; van Dellen et al., 2012).
Current research
In spite of the fact that there is an abundant body of research on functional connectivity networks and cognition in glioma patients, there is still little scientific understanding of the global character of their cognitive deficits, and the specific changes that might occur in the negative correlation between the DMN and the FPN. Therefore, this research attempts to address the lacuna in this field by answering the following question: What does the negative correlation between the DMN and the FPN, hereinafter referred to as internetwork
connectivity, look like in these patients, and what is the association of internetwork
10
In the current study, we aim to investigate the association between internetwork connectivity and attention in glioma patients. Overall, we hypothesize that the internetwork connectivity is negatively associated with attention. In consequence, glioma patients who have a more negative internetwork connectivity are more likely to have less attention impairments. To investigate the relationship between the internetwork connectivity and attention, four key questions are asked. 1) We aim to investigate the presence of a negative internetwork connectivity within the patient group. It is then hypothesized that a negative internetwork connectivity exists in this sample of glioma patients. 2) We investigate the association between attention and internetwork connectivity. It is hypothesized these are negatively correlated. 3) We aim to investigate the difference between patients with and without attention impairments, and the association with internetwork connectivity. It is hypothesized that patients who have worse attention impairments have a less negative internetwork connectivity than patients with no attention impairments. 4) We examine confounding effects of age, education and Karnofsky performance score (KPS) (Karnofsky et al., 1948)—a performance scale to determine general well being—on the association, as well as tumour specific features. We predict that the confounding effects do have an influence on the relationship between internetwork connectivity and attention. Subsequently we predict that features specific for brain tumour, such as volume, grade and location, do not have a significant effect on the relationship between internetwork connectivity and attention.This study will provide an important opportunity to advance our understanding of functional network topology and cognitive deficits in glioma patients. The investigated relationship between different negatively correlated networks, such as the DMN and FPN, and the connection with attention in glioma patients will provide a more comprehensive understanding of complex networks within the brain, and contribute to the growing interest in functional connectivity.
11
MethodsParticipants
Newly diagnosed glioma patients at the VUmc were eligible if 1) a glioma WHO grade II-IV was histologically confirmed; 2) a preoperative language fMRI—word completion task—was assessed; 3) a structural 3D MRI was available for co-registration; and 4) a neuropsychological examination was conducted before surgery. Seventy-one patients were initially selected for this study. Of these 71 patients, 12 patients were later excluded based on previous craniotomy, previous chemo- or radiotherapy and neurological or psychiatric comorbidity. The fMRI data and neuropsychological examination data of each patient were then combined into one dataset. Leaving a dataset of 59 patients.
A medical doctor used the KPS to indicate daily functional impairment in glioma patients. The KPS score can range from 0, denoting death, to a 100, meaning perfect health. No patient who participated in this research scored below a KPS score of 70. The KPS score for this dataset was then summarized categorically as either 70-80 or 90-100.
Before the scan and the neuropsychological examination took place, all participants gave written informed consent.
MRI acquisition
All patients underwent an MRI scan on a 1.5T scanner (Siemens Sonata), including an anatomical 3D T1-weighted scan. An fMRI was performed using a standard echo-planar imaging (EPI) sequence (TR = 2850 ms, TE = 60 ms, 144 volumes, ~7 minute acquisition). The patients also underwent a language fMRI scan that consisted of a word completion task.
MRI analysis
For this study, the FSL 5 was used for the imaging processing. The Brain Extraction Tool (Smith, 2002) was used to delete non-brain tissue, and grey and white matter were segmented using FAST (Zhang et al., 2001). An Automated Anatomical Labeling (AAL) atlas was used to define the 78 cortical regions in order to construct a functional brain for every patient
12
individually (Tzourio-Mazoyer et al., 2002). This atlas was transformed from standard space to native space, which is individual for patient, and masked with each subject’s native grey matter mask. By drawing in the tumour manually, tumour masks were created for every patient [LD]. If the AAL regions in a patient were fully covered by the tumour mask these regions were subsequently excluded; because tumour tissue may theoretically alter the BOLD response locally, thereby confounding our network analysis, it was necessary to remove this region from further analyses.fMRI analysis
When pre-processing the fMRI data, this study used the standard FSL procedures (Smith et al., 2004) included in Melodic (Beckmann et al., 2005). These standard procedures included removing the first 5 volumes, motion correction, spatial smoothing (6mm full with half maximum Gaussian kernel), and high-pass filtering (100 second cut-off). The functional images were co-registered to the anatomical scans using linear and non-linear co-registration methods (Jenkinson et al., 2002).
In order to ascertain that the connectivity results would not be caused by motion during fMRI (Van Dijk et al., 2010), strict exclusion criteria for motion were applied: (1) average relative motion ≥4mm, and (2) more than five frame-to-frame movements ≥5mm.
DMN and FPN functional connectivity calculation
Connectivity analysis was performed using in-house scripts, and the Brain Connectivity Toolbox (Rubinov and Sporns, 2010) in Matlab R2012a (Mathworks, Natick, MA, US). Of the 78 cortical regions, ten had poor signal due to the scanner settings. Therefore 68 regions were included in our analysis. Calculating Pearson correlation coefficients between time series from all 68 regions created a 68x68 connectivity matrix per patient. The absolute values were used as a weighted indication of connectivity between all region pairs. The AAL atlas was used to define the ROIs of the DMN and the FPN. For each network, the eight ROIs, representing the hubs of these networks, were selected based on the earlier research of van
13
Dellen et al. (2012). The brain regions selected for the DMN were the medial superior frontal gyrus, posterior cingulate, angular gyrus and precuneus; the brain regions selected for the FPN were the middle frontal gyrus, superior frontal gyrus, inferior parietal gyrus, angular gyrus and precuneus.Analysing the different time series—series of data points in time order—between the 68 ROIs of both networks, allowed us to calculate the correlations between these networks for each patient. Indicating that for each patient the correlations between all ROIs of both networks were calculated. Pearson correlation coefficients were used to calculate connectivity. Subsequently, they were converted into Fisher z-scores for normalization. The average correlation between all FPN regions and DMN regions were calculated as a measure of internetwork connectivity. Due to the fact that the average correlation averages out the strongest negative correlation, it was necessary to determine both the minimum correlation (highest negative correlation) and the average of the minimum correlation. In the end we had three variables for internetwork connectivity: average correlation, minimum correlation and average minimum correlation.
Neuropsychological evaluation
Each patient underwent a neuropsychological examination. The addressed cognitive domains were attention and (working) memory. Three tests were used for these domains: Word Learning Test, Animal Fluency Test and Stroop Colour Word Test. Individual subjects’ test scores were converted into t-scores, using the norms of the tests (Schmand, Houx & Koning, 2012).
Attention was assessed by means of the Stroop Colour Word test (Hammes, 1971). This test is used to investigate mental speed and executive attention. The test consists of three cards. For the first card the patient reads the colour names—blue, red, yellow and green—out loud and as fast as possible without making mistakes. The second card consists of squares in the same four colours and the patient needs to name the colours as fast as possible without making mistakes. The last card consists of the four colour names printed in different colours
14
(e.g. blue is written in red, green in yellow etc.). The patient needs to name the colour of the ink as fast as possible and without making mistakes. The reliability and validity of this test are rated as good and sufficient, respectively. Test-retest correlation coefficients (Pearsons r) for the three cards are 0.73, 0.78, and 0.85 (Bouma et al., 2012).Memory was determined by means of the Word Learning Test (Kalverboer & Deelman, 1964). This test is used to investigate verbal learning and short-term memory by having the patients memorize fifteen words (Bouma et al., 2012). The words are read out loud to the patient five times and the patient is asked to immediately recall as many words as possible. This process is repeated five times, after which there is a pause of 20 minutes. This is then followed by a delayed recall and recognition phase. The reliability and validity of this test are both rated as good (Bouma et al., 2012). The internal consistency is a Cronbach’s α of 0.91.
For this test, working memory was determined by means of the Animal Fluency Test (Mulder et al., 2006). Patients need to fabricate as many unique words within a certain category in a certain period of time. The test-retest reliability of this test is rated as sufficient (Pearsons r is 0.82) (Bouma et al., 2012).
Statistical analysis
All statistical analyses were performed using IBM SPSS Statistics for Windows, Version 22.0 (IBM Corp, Armonk, NY, US). Probabilities of p < .05 (two-tailed) were considered statistically significant.
Because there was a broad variance in internetwork connectivity in this patient group, a negative correlation needed to be ascertained between the DMN and the FPN. One-sample t-tests were used to examine the three variables for internetwork connectivity; average correlation, minimum correlation (highest negative correlation) and average minimum correlation. The three variables are compared with a criterion set at zero. Normality was checked with Kolmogorov-Smirnov’s test. All internetwork connectivity variables were measured at scale level. A boxplot was examined to identify outliers. Due to the clinical aspect of this sample and the size, omitting outliers could have caused deviating results. Since
15
the outcome of the internetwork connectivity variables did not change when the analysis was run without outliers, the outliers were then included in the dataset.Multiple regression analyses were used to determine the relationship between the internetwork connectivity of the DMN and the FPN, and attention. Education, KPS, age, tumour location, tumour volume, and tumour type were used as covariates because they may have an effect on the association between attention and internetwork connectivity. Based on previous research, this study predicted that there would be an effect of gender on internetwork connectivity. Therefore, gender was used as an independent variable in the first step of the regression analysis, with internetwork connectivity as the dependent variable. In the second step, the attention variable was added as independent variable together with gender. The three Stroop scores were tested separately for a variety of reasons; first, to maintain statistical power, second, because the three cards all test different kinds of attention, and third, because the three cards may be differently associated with internetwork connectivity. The remaining covariates were also added separately. The variables were then put into the analysis with the ENTER method. The assumptions of normality, linearity, multicollinearity, and homogeneity of variances—checked by producing plots, and with the Durbin-Watson test and Kolmogorov-Smirnov test—were maintainable.
An ANCOVA was used to determine if internetwork connectivity increased due to attention impairments, and whether education, KPS, tumour location, tumour volume and tumour type have an effect on this increase. The patients were dichotomized based on their score of the attention test. Those who scores below a t-score of 35 on the attention test were labelled as having an indication for attention impairment. The dependent variable was internetwork connectivity, and the covariates were education, KPS, tumour location, tumour volume and tumour type.
16
ResultsPatient characteristics
This study included 59 patients that had a preoperative language fMRI as well as a neuropsychological examination. Values were missing due to the fact that only 57 patients completed the Stroop test. The final dataset included 36 men and 23 women. The patients were on average 38 years old (age ranges from 17 to 62 years). The education level of the patients was labelled according to the Verhage classification, which runs from 1 to 7 (Verhage, 1964). This scale is commonly used in the Netherlands. Of the 59 patients, two (3.4%) had an education of 3. Nine (15.3%) had an education of 4, eighteen (30.5%) had an education of 5, 20 (33.9%) patients had an education of 6, and ten (16.9%) had an education of 7. Additional demographic and clinical characteristics are presented in Table 1. The average connectivity for both the DMN and the FPN, and the average scores for the neuropsychological tests are presented in Table 2.
There was a difference between present (N=15) and non-present (N=42) attention impairment patients regarding age (F (1, 56) = 4.95, p = .03) and gender (X² = 4.69, df = 1, p = .03). There was no difference between these patients regarding tumour type (X² = 1.584, df = 1, p = .208), KPS (X² = 2.946, df = 1, p = .086), frontal location (X² = 1.629, df = 1, p = .202), tumour volume (t(55) = -.87, p = .39) and education (X² = 5.82, df = 4, p = .21).
17
Table 1Note.aAverage; bStandard Deviation; cGlioblastoma Multiforme; dLow Grade Glioma; eHigh
Grade Glioma; fKarnofsky Performance Score
Demographic and Clinical Characteristics Demographic Characteristics Sex (n [%]) Male Female Age in years Ma (SDb) Education Verhage (n [%]) Scale 3 Scale 4 Scale 5 Scale 6 Scale 7 Clinical Characteristics Histology (n[%]) Astrocytoma Oligodendroglioma Oligoastrocytoma GBMc Side (n [%]) Left Right Type (n [%]) LGGd HGGe Location (n [%]) Occipital Frontal Temporal Parietal Frontoparietal Frontotemporal Tumour volume (mm3) Ma (SDb) KPSf (n [%]) 70-80 90-100 36 (61) 23 (39) 38 (11) 2 (3.4) 9 (15.3) 18 (30.5) 20 (33.9) 10 (16.9) 21 (35.6) 19 (32.2) 11 (18.6) 8 (13.6) 31 (52.5) 28 (47.5) 36 (61) 23 (39) 1 (1.7) 24 (4) 13 (22) 8 (14) 5 (8) 8 (14) 60801.07 (45484.53) 31 (52.5) 28 (47.5)
18
Table 2Average of DMNa connectivity, FPNb connectivity and neuropsychological tests
Mc SDd Maximum Minimum
DMN .33 .17 1.03 .04
FPN .33 .18 .99 .09
WLTe 45.8 12.3 67 14
Animal Fluency Test 42,5 12.9 82 18
Stroop Test 41.9 13.8 98 10
Note. aDefault Mode Network; bFrontoparietal Network; c Average; dStanderd Deviation;
eWord Learning Test
Negative internetwork connectivity in glioma patients
To establish if an internetwork connectivity existed in this patient sample, different group averages were computed. The group average of the correlation between the DMN and the FPN was r = .4 (M = .41, SD = .23, Max = 1.18, Min = -.03). The group average of the minimum correlation was -.7 (M = -.77, SD = .56, Max = .87, Min = -2.15). The group average of the mean minimum correlation was -.3 (M = -.33, SD = .18, Max = .00, Min = -.76).
Due to deviations within the patient group for internetwork connectivity, which may cause a total value that is less negative than expected, one-sample t-tests were used to investigate the internetwork connectivity variables. The one-sample t-tests showed a statistically significant negative internetwork connectivity lower than zero (M = 0) for both minimum correlation (t(58) = -10.67, p < .001) and for mean minimum correlation (M = 0;
t(58) = -14.06, p < .001). Average correlation was significant higher than zero (t(58) = 13.76, p < .001). Because the minimum correlation had the lowest average (-.7) it was used in the
subsequent analyses. This ensured that our analyses were not affected by the aspecificity of the AAL ROIs and the variance between subjects for the exact networks.
Negative internetwork connectivity and attention
Results from a Pearson’s correlations matrix showed that the strongest negative correlation between internetwork connectivity and attention was between the variables minimum correlation and Stroop card 3, which was not significant (r = -.15, n = 57, p = .14). For further analyses Stroop card 3 was used as predictor.
19
A multiple linear regression was performed to predict internetwork connectivity based on gender and attention. The first step of the multiple regression analysis showed that gender significantly predicts internetwork connectivity (β = -.27, t(55) = -2.10, p = .04). The explained variance is R2 = .07 (F(1,55) = 4.42, p = .04). Adding attention score, measuredwith Stroop Card 3, in step two of the analysis (β = -.31, t(54) = -2.41, p = .02 for gender and β = -.12, t(54) = -1.61, p = .11 for attention, with an R2 = .12 (F(1,54) = 3.57, p = .03)) did not
result in a significant increase of the explained variance of internetwork connectivity (ΔR² = .04, Fchange (1,54) = 2.6, p = .11). The other measurements of attention and covariates did not show a relationship with internetwork connectivity.
Attention impairments and internetwork connectivity
After dichotomizing the patient sample a one-way ANCOVA showed that no significant association between attention impairments and internetwork connectivity was found after controlling for education (F(1,54) = 0.22, p = .64), KPS (F(1,54) = 0.01, p = .91), location (F(1,54) = 0.017, p = .89), tumour volume (F(1,54) = 0.06, p = .80) and tumour type (F(1,54) = 0.02, p = .88). The dependent variable was normally distributed in both groups and the assumption of homogeneity of variances was met for all covariates except for KPS and tumour type. Because there was a difference between present and non-present attention impairment patients regarding age and gender, these variables were not used in this analysis.
Exploratory analyses
Exploratory analyses were carried out to investigate the relationship between the remaining neuropsychological tests and internetwork connectivity, and to further explore the found difference between men and women.
First, the relationship between internetwork connectivity and the WLT and Animal Fluency Test were examined. This analysis was executed to check if these neuropsychological tests, which measure (working) memory, show a different relationship with internetwork connectivity compared to the Stroop Test. Regression analysis showed no
20
significant relationship between WLT and internetwork connectivity (β = -.07, t(57) = -.54, p = .59) and no significant relationship between Animal Fluency Test and internetwork connectivity (β = .14, t(55) = 1.04, p = .30)Second, to further explore the differences between men and women in the relationship between internetwork connectivity and attention, gender differences were investigated. An independent t-test was used to investigate the differences between men and women concerning internetwork connectivity. The average internetwork connectivity for men was -.67 (SD = .57), and for women -.93 (SD = .52). The independent t-test showed no significant difference between men and women regarding internetwork connectivity, even though the average internetwork connectivity of women was higher compared to men (t(57) = 1.72, p = .09). Another independent t-test was used to investigate the differences between men and women concerning attention scores. The average attention score for men was 43.97 (SD = 14.75), and for women 38.38 (SD = 11.51). The independent t-test showed no significant difference between men and women regarding attention score (t(55) = 1.49, p = .14). A correlation was used to investigate the association between attention scores and internetwork connectivity for men and women. For women, no significant correlation was found between internetwork connectivity and attention (r (19) = -.114, p > .05). However, men did show a trend towards a significant correlation between internetwork connectivity and attention (r (34) = -.25, p = .07).
21
DiscussionThis study aimed to investigate the negative correlation between the default mode network (DMN) and the frontopartiental network (FPN), referred to as internetwork connectivity, and attention in glioma patients. The DMN and FPN were computed with functional connectivity based on fMRI in 59 glioma patients. We hypothesized that there is a negative relationship between internetwork connectivity and attention in glioma patients. This is not confirmed.
We address four research questions in this study. First, we show that there is a significant negative correlation of the internetwork connectivity in a sample of glioma patients. Second, we find that attention score is not related to the internetwork connectivity between the DMN and FPN. Only when gender was taken into account there is an association. A more negative internetwork connectivity is then associated with a better attention score. Third, after dichotomizing the patient sample we show that the presence or absence of attention impairments in glioma patients is not related to internetwork connectivity. There is no difference between patients with attention impairments regarding internetwork connectivity and patients without attention impairments. Finally, the confounding effects such as age, KPS, and other tumour characteristics have no effect on the association.
The effect of gender in this study is further explored and we find that women actually have a more negative internetwork connectivity on average compared to men even though this is not significant. This would mean women suppress their DMN more sufficiently when performing goal-directed behaviour, and are expected to have higher attention scores on average. Unfortunately, in our findings only men show a trend towards a significant correlation between internetwork connectivity and attention scores. For women this association was not found, which was contradicting with our expectations based on their attention scores and internetwork connectivity. The meaning of this is not as substantial as hoped, because the sample size of this study is rather small. This is why we cannot expand further on this finding with these results.
22
There is a number of possible explanations for the findings of the research questions in this study. First, an association between attention scores and internetwork connectivity has been found before in a study investigating schizophrenia patients. This study assessed the lack of DMN suppression when patients were performing a demanding cognitive task (Anticevic et al., 2012). No other studies are known wherein the correlation between the DMN and FPN and attention in glioma patients is investigated. A previous study found an altered integrity of the DMN in glioma patients (Harris et al., 2013), which could be the explanation for the found negative correlation between the networks in our study. A potential explanation regarding the missing association between attention and internetwork connectivity in this study might be the difference in pathology compared with the Anitcevic et al. (2012) study. Anticevic et al. studied schizophrenia patients and their lack of suppression of the DMN. This lack of suppression is more profound compared to our cohort of newly diagnosed glioma patients.Second, the found effect of gender on the associations is a remarkable finding. We find that gender predicts internetwork connectivity and that women have on average a higher internetwork connectivity than men. In fMRI research, there is a wide variety of findings concerning as to whether gender differences are present in functional connectivity. Many fMRI studies found remarkable differences between men and women in functional connectivity and network topology (Lui et al., 2006; Biswall et al., 2010; Gong et al., 2010). In addition, the effect of gender was stated in previous research with MEG (Douw et al., 2011). That study found a significant correlation between local connectivity and cognition in men, but not in women. In contrast, Tian et al. (2010) did not find gender differences in network topology in their fMRI research, which is contradicted by our results.
Third, with regard to differences between patients with and without attention impairments, a possible alternative explanation might be the limited number of patients with attention impairments in this study. Furthermore, the unevenly distributed categories might be the cause for the found results, for example the fact that there were more LGG patients compared to HGG patients.
23
Some limitations of the current study should be considered. First, because only patients who scored between 70 and 100 on the KPS participated in this study, our sample might be biased towards patients who perform relatively high. In addition, glioma patients who participate in research generally do not have highly aggressive tumours and other profound cognitive deficits. This means that the current sample might not be representative of glioma patients in general. Second, due to this group consisting of newly diagnosed patients, the overall deficits are not yet as profound as expected in a later stage of the disease. Future studies should aim to include a control group in order to examine if this patient group actually has cognitive deficits compared to healthy controls. Also by including patients with more diverse clinical characteristics it is possible to clearer distinct between patients with and without attention impairments. Third, due to the fact that boundaries of gliomas are hard to define, there is no guarantee of exclusion of all tumour-covered AAL atlas regions. This might cause differences between patients in the number of ROIs used in this research. Fourth, as is stated before by Uddin et al. (2009), not all FPN regions may be negatively correlated with DMN regions. They found differences between negatively correlated regions of the vmPFC and PCC. Therefore, using an improved selection of ROIs may produce stronger results in future studies. As an example the independent component analysis (ICA) could be used. The ICA separates the data into independent components, which are used as brain regions in the analysis (Demoiseaux et al., 2006). Final, in general a resting state fMRI is preferred in fMRI research (Lee, Smyser, & Shimony, 2013), but due to the fact that the imaging has taken place for clinical purposes, no resting state was available. Previous studies do show that functional connectivity is consistent across states (Krienen et al., 2014); this is why a task based fMRI was used in this study.To conclude, our current results show a negative internetwork connectivity between the default mode network and the frontoparietal network in glioma patients, but no association between this internetwork connectivity and attention is shown. This study shows that
attention is only associated with internetwork connectivity when gender is taken into account, suggesting that differences between men and women regarding functional connectivity are
24
substantial. An interesting next step would be to even further explore the relationship between gender and anticorrelation network changes in glioma patients, and the implications of these differences for cognitive functioning.25
ReferencesAnticevic, A., Cole, M. W., Murray, J. D., Corlett, P. R., Wang, X. J., & Krystal, J. H. (2012). The role of default network deactivation in cognition and disease. Trends in Cognitive
Sciences, 16(12), 584–592.
Arenaza-Urquijo, E. M., Landeau, B., La Joie, R., Mevel, K., Mézenge, F., Perrotin, A., … Chételat, G. (2013). Relationships between years of education and gray matter volume, metabolism and functional connectivity in healthy elders. NeuroImage, 83, 450–457.
Bartolomei, F., Bosma, I., Klein, M., Baayen, J. C., Reijneveld, J. C., Postma, T. J., … Stam, C. J. (2006). Disturbed functional connectivity in brain tumour patients : Evaluation by graph analysis of synchronization matrices. Clinical Neurophysiology, 117(18), 2039–2049.
Bartolomei, F., Bosma, I., Klein, M., Baayen, J. C., Reijneveld, J. C., Postma, T. J., … Stam, C. J. (2006). How do brain tumours alter functional connectivity? A
magnetoencephalography study. Annals of Neurology, 59(1), 128–138.
Beckmann, C. F., DeLuca, M., Devlin, J. T., & Smith, S. M. (2005). Investigations into resting-state connectivity using independent component analysis. Philosophical Transactions
of the Royal Society of London B: Biological Sciences, 360(1457), 1001-1013.
Biswal, B. B., Mennes, M., Zuo, X. N., Gohel, S., Kelly, C., Smith, S. M., ... &
Dogonowski, A. M. (2010). Toward discovery science of human brain function.Proceedings
of the National Academy of Sciences, 107(10), 4734-4739.
Bondy, M. L., Scheurer, M. E., Malmer, B., Barnholtz-sloan, J. S., Davis, F. G., Il, D., & Buffler, P. A. (2008). Brain Tumour Epidemiology : Consensus From the Brain Tumour Epidemiology Consortium. Cancer, 113(7 supp(September), 1953–1968.
26
Bosma, I., Reijneveld, J. C., Klein, M., Douw, L., van Dijk, B. W., Heimans, J. J., & Stam, C. J. (2009). Disturbed functional brain networks and neurocognitive function in low-grade glioma patients: a graph theoretical analysis of resting-state MEG. NonlinearBiomedical Physics, 3, 9.
Bouma, A., Mulder, J., Lindeboom, J., & Schmand, B. (2012). Handboek
neuropsychologische diagnostiek.-2e herz. dr. Pearson.
Bressler, S. L., & Menon, V. (2010). Large-scale brain networks in cognition: emerging methods and principles. Trends in Cognitive Sciences, 14(6), 277–290.
Buckner, R. L., Andrews-Hanna, J. R., & Schacter, D. L. (2008). The brain’s default network: Anatomy, function, and relevance to disease. Annals of the New York Academy of
Sciences, 1124, 1–38.
Bullmore, E. and O. S. (2009). Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci., 10(3), 186–198.
Cahill, L. (2006). Why sex matters for neuroscience. Nature Reviews Neuroscience, 7(6), 477-484.
Cole, M. W., Yarkoni, T., Repovs, G., Anticevic, A., & Braver, T. S. (2012). Global connectivity of prefrontal cortex predicts cognitive control and intelligence. Journal of
Neuroscience, 32(26), 8988–8999.
Damoiseaux, J. S., Rombouts, S. A. R. B., Barkhof, F., Scheltens, P., Stam, C. J., Smith, S. M., & Beckmann, C. F. (2006). Consistent resting-state networks across healthy subjects. Proceedings of the national academy of sciences, 103(37), 13848-13853.
Derks, J., Reijneveld, J. C., & Douw, L. (2014). Neural network alterations underlie cognitive deficits in brain tumour patients. Current Opinion in Oncology, 26(6), 627–633.
27
Douw, L., van Dellen, E., de Groot, M., Heimans, J. J., Klein, M., Stam, C. J., &Reijneveld, J. C. (2010). Epilepsy is related to theta band brain connectivity and network topology in brain tumour patients. BMC Neuroscience, 11, 103.
Douw, L., Schoonheim, M. M., Landi, D., van der Meer, M. L., Geurts, J. J. G., Reijneveld, J. C., … Stam, C. J. (2011). Cognition is related to resting-state small-world network topology: An magnetoencephalographic study. Neuroscience, 175, 169–177.
Fornito, A., Harrison, B. J., Zalesky, A., & Simons, J. S. (2012). Competitive and cooperative dynamics of large-scale brain functional networks supporting recollection.
Proceedings of the National Academy of Sciences, 109(31), 12788–12793.
Giovagnoli, A. R. (2012). Investigation of cognitive impairments in people with brain tumours. Journal of Neuro-Oncology, 108(2), 277–283.
Gong, G., He, Y., & Evans, A. C. (2011). Brain connectivity gender makes a difference. The Neuroscientist, 17(5), 575-591.
Greicius, M. D., Krasnow, B., Reiss, A. L., & Menon, V. (2003). Functional connectivity in the resting brain: a network analysis of the default mode hypothesis. Proc Natl Acad Sci U
S A, 100(1), 253–258.
Hammes, J. (1971). De Stroop Kleur-Woord Test: Handleiding. Lisse: Swets & Zeitlinger.
Harris, R. J., Bookheimer, S. Y., Cloughesy, T. F., Kim, H. J., Pope, W. B., Lai, A., … Ellingson, B. M. (2013). Altered functional connectivity of the default mode network in diffuse gliomas measured with pseudo-resting state fMRI. Journal of Neuro-Oncology,
28
Heimans, J. J., & Reijneveld, J. C. (2012). Factors affecting the cerebral network in brain tumour patients. Journal of Neuro-Oncology, 108(2), 231–237.Jenkinson, M., Bannister, P., Brady, M., & Smith, S. (2002). Improved optimization for the robust and accurate linear registration and motion correction of brain
images. Neuroimage, 17(2), 825-841.
Kalverboer, A. F., & Deelman, B. G. (1964). De 15-woorden test [The 15 words test]. Groningen, the Netherlands: Academisch Ziekenhuis Afdeling Neuropsychologie.
Karnofsky, D. A., Abelmann, W. H., Craver, L. F., & Burchenal, J. H. (1948). The use of the nitrogen mustards in the palliative treatment of carcinoma. With particular reference to bronchogenic carcinoma. Cancer, 1(4), 634-656.
Krienen, F. M., Yeo, B. T., & Buckner, R. L. (2014). Reconfigurable task-dependent functional coupling modes cluster around a core functional architecture. Phil. Trans. R. Soc.
B, 369(1653), 20130526.
Liu, H., Stufflebeam, S. M., Sepulcre, J., Hedden, T., & Buckner, R. L. (2009). Evidence from intrinsic activity that asymmetry of the human brain is controlled by multiple
factors. Proceedings of the National Academy of Sciences, 106(48), 20499-20503.
Lee, M. H., Smyser, C. D., & Shimony, J. S. (2013). Resting-state fMRI: a review of methods and clinical applications. AJNR. American Journal of Neuroradiology, 34(10), 1866–72.
Markert, J. (2005). Glioblastoma multiforme. Jones & Bartlett Learning.
Mears, D., & Pollard, H. B. (2016). Network science and the human brain: Using graph theory to understand the brain and one of its hubs, the amygdala, in health and disease.
29
Mulder, J.L., Dekker, P.H. & Dekker, R. (2006). Woord-Fluency Test/ Figuur-FluencyTest. Handleiding. Leiden: PITS;
Osborn, A. G., Salzman, K. L., Jhaveri, M. D., & Barkovich, A. J. (2015). Diagnostic
imaging: brain. Elsevier Health Sciences.
Rubinov, M., & Sporns, O. (2010). Complex network measures of brain connectivity: Uses and interpretations. NeuroImage, 52(3), 1059–1069.
Schmand, B., Houx, P., & Koning, I. D. (2012). Normen neuropsychologische tests 2012. Gepubliceerd op de website van de sectie Neuropsychologie: Nederlands Instituut van
Psychologen.
Smith, S. M. (2002). Fast robust automated brain extraction. Human brain
mapping, 17(3), 143-155.
Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E., Johansen-Berg, H., ... & Niazy, R. K. (2004). Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage, 23, S208-S219.
Sporns, O., Honey, C. J., & Kötter, R. (2007). Identification and Classificatino of Hubs in Brain Networks. PLoS ONE, 2(10), 1–14.
Sporns, O., Chialvo, D. R., Kaiser, M., & Hilgetag, C. C. (2004). Organization , development and function of complex brain networks. Trend in Cognitive Sciences, 8(9), 418–425.
Stam, C. J., & Reijneveld, J. C. (2007). Graph theoretical analysis of complex networks in the brain. Biomed, 1(3), 1–19.
30
Stupp, R., Mason, W. P., Van Den Bent, M. J., Weller, M., Fisher, B., Taphoorn, M. J., ... & Curschmann, J. (2005). Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. New England Journal of Medicine, 352(10), 987-996.Taphoorn, M. J. B., & Klein, M. (2004). Cognitive deficits in adult patients with brain tumours. Lancet Neurology, 3(3), 159–168.
Tian, L., Wang, J., Yan, C., & He, Y. (2011). Hemisphere-and gender-related differences in small-world brain networks: a resting-state functional MRI study.Neuroimage, 54(1), 191-202.
Tucha, O., Smely, C., Preier, M., & Lange, K. W. (2000). Cognitive deficits before treatment among patients with brain tumours. Neurosurgery, 47(2), 324-334.
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., ... & Joliot, M. (2002). Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject
brain. Neuroimage, 15(1), 273-289.
Uddin, L. Q., Kelly, A. M. C., Biswal, B. B., Castellanos, F. X., & Milham, M. P. (2009). Functional Connectivity of Default Mode Network Components: Correlation, Anticorrelation, and Causality. Human Brain Mapping, 30(2), 625–637.
van Dellen, E., Douw, L., Hillebrand, A., Ris-Hilgersom, I. H. M., Schoonheim, M. M., Baayen, J. C., … Reijneveld, J. C. (2012). MEG Network Differences between Low- and High-Grade Glioma Related to Epilepsy and Cognition. PLoS ONE, 7(11).
van den Heuvel, M. P., & Sporns, O. (2011). Rich-Club Organization of the Human Connectome. Journal of Neuroscience, 31(44), 15775–15786.
31
Van Dijk, K. R., Hedden, T., Venkataraman, A., Evans, K. C., Lazar, S. W., & Buckner, R. L. (2010). Intrinsic functional connectivity as a tool for human connectomics: theory, properties, and optimization. Journal of neurophysiology,103(1), 297-321.van Straaten, E. C. W., & Stam, C. J. (2013). Structure out of chaos: Functional brain network analysis with EEG, MEG, and functional MRI. European
Neuropsychopharmacology, 23(1), 7–18.
Verhage, F. (1964). Intelligentie en leeftijd: Onderzoek bij Nederlanders van twaalf tot
zevenenzeventig jaar. Assen: Van Gorcum.
Wang, Z., Dai, Z., Gong, G., Zhou, C., & He, Y. (2015). Understanding structural-functional relationships in the human brain: a large-scale network perspective. Neuroscientist,
21(3), 290–305.
Weissman, D. H., Roberts, K. C., Visscher, K. M., & Woldorff, M. G. (2006). The neural bases of momentary lapses in attention. Nature Neuroscience, 9(7), 971–8.
Wu, J. T., Wu, H. Z., Yan, C. G., Chen, W. X., Zhang, H. Y., He, Y., & Yang, H. S. (2011). Aging-related changes in the default mode network and its anti-correlated networks: A resting-state fMRI study. Neuroscience Letters, 504(1), 62–67.
Young, R. M., Jamshidi, A., Davis, G., & Sherman, J. H. (2015). Current trends in the surgical management and treatment of adult glioblastoma. Annals of Translational Medicine,
3(9), 121.
Zhang, Y., Brady, M., & Smith, S. (2001). Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE
