Connected Brains: research

The Web Connected Brains

Brain network research at the department of Clinical Neurophysiology of the VU University Medical Center

Since 2000 the research programme of the department of clinical neurophysiology and the MEG center of the VU University Medical Center is centered around the question how different brain areas communicate, and how this “functional connectivity” changes in neurological disease. This research is done to a large extent in patient cohorts of the department of neurology of the VU University Medical Center (Alzheimer’s disease and other types of dementia, Parkinson’s disease, brain tumors and multiple sclerosis [Kalkers et al., 2007; Cover et al., 2006]). In addition we collaborate with the group of prof. Boomsma and prof. de Geus of the department of biological psychology of the VU University in twin studies and we collaborate with several other groups abroad in studies on schizophrenia, depression and sleep distrubances. In the period 2000-2005 the primary focus was on developing and implementing new methods for the assessment of functional connectivity in EEG and MEG. In 2002 we developed a new method (synchronization likelihood, Stam and van Dijk, 2002) which has since been used in a large number of studies. Furthermore, a number of MEG studies in the previously mentioned patients cohorts were initiated. An overview of the research done in the period can be Found in a review paper (Stam, 2005) and a book Nonlinear Brain Dynamics(Stam, 2006).

One problem in functional connectivity, in particular those involving MEG and fMRI, is that large amount of data are obtained that are difficult to analyse and interpret due to their complexity. Around 1998 / 1999 there have been some major breakthroughs in the study of complex networks. This has resulted in a completely new theoretical approach to the study of complex networks in many fields of science. The department of clinical neurophysiology of the VU University Medical Center was the first group to apply this new approach to MEG recordings in human subjects (Stam, 2004). Since then this type of research has expanded enormously. The period 2005-2010 is therefore characterized to an important extent by the application of the graph theoretical approaches in healthy subjects and important groups of neurological and neuropsychiatric patients. An overview of the research into complex brain networks can be found in several review papers (Stam and Reijneveld, 2007; Reijneveld et al., 2007; Stam 2010a,b). An extensive description of brain network research directed at a general audience can be found in a recent Dutch book (Hersenweb, Stam, Douw and de Haan; Bert Bakker / Prometheus). Below we give a more detailed description of research results in different categories of patients.

Dementia and Alzheimer’s disease

With respect to patient care and scientific research there is a close collaboration between the department of clinical neurophysiology and the MEG center of the VU University Medical center and the Alzheimer center. MEG studies in Alzheimer’s disease have been done since 2000. In 2006-208 MEG changes in Alzheimer’s disease Were studies in detail (Stam et al., 2006; de Haan et al., 2008). In these studies it was noted that there can be increases as well as decreases of oscillatory brain activity and functional connectivity involving different brain areas and different frequency bands. Changes in functional connectivity can also be demonstrated in an early stage of dementia referred to as mild cognitive impairment (MCI). This was investigated together with neuroscientists in Italy and Spain (Babiloni et al., 2006; Gomez et al., 2009). In recent studies using network analysis it was shown that functional networks in Alzheimer’s disease loose their normal small-world structure, and regress towards a more random architecture (Stam et al., 2007; Stam et al., 2009). Using simulations it could be shown that this was due to a specific loss of so-called “hubs” (areas with a very high level of connectivity) in the brain networks of Alzheimer patients. These findings were confirmed in a recent EEG study that also revealed a completely different type of network change in frontotemporal dementia (de Haan et al., 2009). By now it has become clear that there is a direct connection between hubs and amyloid deposition. These findings will be further explored in the next years. In collaboration with dr Klaus Linkenkaer-Hansen of the NCA MEG recordings of Alzheimer patients were studied from the perspective of “self-organized criticality”; these results have been reported in a PNAS paper (Montez et al., 2009).

MEG studies were supplemented by EEG research. Patterns of visual EEG analysis in different types of dementia have been described by Roks et al. (2008) and Liedorp et al. (2009,2010). In a quantitative EEG study a relation was shown between coherency (a linear measure of functional connectivity) and the correlation dimension (Jelles et al., 2008). Gerdien Kramer demonstrated a connection between functional connectivity and APOE genotype, both in healthy subjects as well as Alzheimer patients (Kramer et al., 2008). In 2007 we applied network analysis for the first time to EEGs of Alzheimer patients (Stam et al., 2007). This study revealed abnormal changes in network structure that were further characterized in later EEG and MEG studies. The mutual interest in “resting-state” networks resulted in a collaboration with the department of radiology (prof. F. Barkhof; dr. S.A.R.B. Rombouts). This research was done to a large extent in the context of the PhD thesis of dr J. Damoiseaux (Rombouts et al., 2005; Damoiseaux et al., 2006, 2008, 2009).

Parkinson’s disease

In the period 2005-2010 systematic MEG research of a cohort of 80 non demented Parkinson patients, divided into four groups of different disease stages, and 20 healthy age controls was set up in collaboration with the Parkison group (prof. E. Wolters, dr. H.W. Berendse) of the VU University Medical Center. Spectral analysis and assessment of functional connectivity of MEG recordings of these patients was done in the PhD thesis of dr. Stoffers (2007; 2008a,b). MEG showed a characteristic pattern of hypersynchronization, also in early, untreated stages of the disease. These changes correlate to some extent with the severity of cognitive and motor changes, and respond in a characteristic way to dopaminergic treatment. MEG changes in demented Parkinson patients have been investigated in the PhD work of dr Bosboom (Bosboom et al., 2006, 2009a,b). Dementia in Parkinson’s disease is reflected in reflected by changes in different areas and frequency bands than parkinsomism without dementia. MEG changes induced by olfactory stimulation were described by dr Boesveldt (Boesveldt et al., 2009). And overview of different MEG-related studies in demented and non-demented Parkinson patients can be found in Berendse and Stam (2007) and Stam (2010a).

Brain tumours and epilepsy

During a stay at the MEG center of the VU University Medical Center dr Fabrice Barolomei, working together with the neuro oncology group, showed for the first time that localized lesions such as brain tumors can give rise to extensive global changes in brain networks (Bartolomei et al., 2006a). In a follow-up study he showed that network analysis reveals a more random topology of brain networks in brain tumor patients (Bartolomei et al., 2006b). Shortly afterwards, working together with drs Sophie Ponten, he applied for the first time network analysis to depth recordings of epileptic seizures (Ponten et al., 2007). During epileptic seizures the topology of functional networks was shown to become abnormally regular. These three studies laid the foundation for a number of succesfull grant applications with the Dutch Epilepsy Foundation (NEF) and several PhD studies in the field of brain tumors, complex brain networks and epilepsy.

Dr Bosma performed a systematic study of MEG network changes in patients with low grade glioma, and related these findings to cognitive changes (Bosma et al., 2008a,b; 2009). In 2010 she succefully defende her PhD thesis on this topic. Dr Linda Douw did a large number of studies involving MEG, EEG and electrocorticography, all aimed at the relation between tumors, cognition and epilepsy (Douw et al., 2009, 2010a,b). She successfully defended her PhD thesis on 8-11-2010. This line of research is continued in the PhD work of Edwin van Dellen (van Dellen et al., 2009). In the context of this research a fruitful collaboration with the network group of prof. Van Mieghem et the Technical University of Delft has been started. The collaboration has already lead to two joint publications (Wang et al., 2010; Stam et al. 2010). Epilepsy is not only studies in relation to brain tumours and epilepsy surgery, but als in a NEF project on seizure mechanisms in critically ill patients in the intensive care unit. These studies will be part of the PhD thesis of dr Ponten (Ponten et al., 2007,2009, 2010a,b; Ronner et al., 2008; Slooter et al., 2006).

Functional connectivity in other disorders

Within the field of functional connectivity and network research collaborations have been established with several other groups, within and outse the VU University Medical Enter, within the Netherlands and abroad. Here we limit ourselves to those collaborations that have resulted in at least one joint publications. Together with the department of pediatrics an MEG study was done in children with obesitas. Children with obesitas were shown to have different networks compared to healthy age controls (Olde Dubbelink et al., 2008). Together with the department of intern medicine MEG studies were done in a cohort of patients with diabetes mellitus type I with and without retinal damage (as a marker of vascular involvement). This study showed that with MEG extensive network changes can be shown in diabetes even at a very early stage (van Duinkerken et al., 2009). These changes might be involved in cognitive dysfunction in diabetes. In collaboration with the Alzheimer centre and the department of neuropsychology of the VU University the effects of lateralization and gender on functional brain networks was studied (Gootjes et al., 2006). In collaboration with the group of prof. Boomsma and prof. De Geus of the department of biological psychology of the VU University developmental and genetic aspects of EEG networks were studied in a large cohort of twins. Both functional connectivity as well as network properties (clustering and pathlength) turned out to be genetically determined, and subject to considerable age and gender effects (Boersma et al., 2011; Smit et al., 2008, 2010; van’t Ent et al., 2009; Posthuma et al., 2005). The development of complex brain networks shows a systematic shift from a more random to a more ordered, small-world pattern (Boersma et al., 2010; Smit et al., 2008, 2010).

In collaboration with the department of psychopharmacology of the University of Utrecht it was shown that longstanding moderate use of alcohol affects brain networks in a specific way (de Bruin et al., 2006). With the department of psychiatry of the university of Utrecht a collaboration has been started on the topic of MRI network analysis. The topology of the healthy brain was described in detail, and a remarkable relation between short pathlength and high intelligence was revealed (Van den Heuvel et al., 2008,2009).

Together with the Italian group of dr Ferri EEG changes in connectivity and graph theoretical properties were investigated (Ferri et al., 2005, 2006a,b, 2008; Zucconi et al. 2005). These studies showed tha functional brain networks display a systematic shift in the direction of more ordered topology during sleep. Changes in brain networks during sleep in depressed patients were described in collaboration with the Brussels group of prof. Linkowski (Leistedt et al., 2009). Together with Sifis Micheloyannis from Greede EEG network studies were performed in healthy subjects, patients with schizophrenia and patients with traumatic brain damage (Pachou et al, 2008 ; Micheloyannis et al., 2006a,b, 2009; Pachou et al., 2008; Tsirka et al., 2010). EEG network studies in schizophrenia were also done together with the Australian group of Michael Breakspear (Rubinov et al., 2009). Recently a collaboration has been started with the group of Mark Molnar in Budapest, Hungary (Gaal et al., 2010). In a longstanding collaboration with Claire Calmels from Parins connectivity changes related to the mirror neuron system have been studied (Calmels et al., 2006a,b, 2008, 2009, 201).

Methodological research

In addition to the application of connectivity and network analysis in different disorders we also perform research into new analysis methods and the modeling of findings in healthy subjects and patients. Here our own software is of considerable importance. Until recently we used the program DIGEEGXP. Since this application can only be used in a Windows environment, and has become very complex and difficult to manage, a new application was developed (BrainWave) programmed in the modern language Java. BrainWave can run on all major operating systems. In BrainWave all methods for connectivity and network analysis have been implicated; it serves as the basis for almost all network studies done by or in collaboration with the department of clinical neurophysiology and the MEG center of the VU Universityy Medical Center.

In the period 2005-2010 the synchronization likelihood, introduced in 2002 by Stam and van Dijk , has been further improved (Montez et al., 2006). In addition a new method has been developed, the phase lag index, that can measure connectivity without the disturbing effects of volume conduction or common reference effects (Stam et al., 2007). Furthermore, a macroscopic model of structural and functional brain networks has been implemented,f irst in DIGEEGXP, and now also in BrainWave. With this model the relation between the topology of structural and functional brain networks has been investigated (Ponten et al., 2010b). More recently this model has been used to show how complex brain networks can arise from two simple principles: random, distance dependent outgrowth of connections, and selective reinforcement or weakening of connections based upon the presence or absence of synchronous activity of brain areas (“Neurons that fire together wire together”)(Stam et al., 2010b). Finally, methods for source localization of MEG in neurosurgical patients have been studied (Willemse et al., 2007; 2011) as well as the integration of EEG and fMRI (van Houdt et al., 2010).

The brain as a “web”

An image of the brain as a complex, self-organizing network is beginning to emerge from the research that has bee done in the last years. The complexity of the healthy adult brain does not result from a preprogrammed blueprint, but it the result of a dynamical process of increasing self-organization, probably under the control of a relatively small number of genes. This development is probably steered by a small number of basic principles: (i) relatively random outgrowth of neuronal connections and establishment of new contacts, whereby the distance between neuronal elements plays a major role; (ii) selective reinforcement and weakening of established connectins under the influence of activity, in particular levels of synchronizatin between neurons and brain areas. The bidirectional interaction between these two processes seem to determine to a large extent the complex structure of the adult brain, optimized for efficient information processing. Moreover, the balance between outgrowth or loss of connections is regulated by a homeostatic principle, whereby the mean level of activity (firing rates) is kept within narrow limits. From a dynamical point of view, our brain is a “critical” system, balancing at the border between stability and instability; by now it has become clear that precisely critical systems are optimal for flexible and optimal information processing.

Neurological disorders are processes that interfere in different manners with this structural and functional self-organization and plasticity of the brain. Remarkably, a number of clearly recognizable patterns of dysfunction can be recognized. The first lesson of network research is that local lesions can have widespread effects, that are determined by the location of the damaged area within the network. With network analysis this general concept can be quantified exactly using centrality measures. Another pattern is the shift from the normal small-world structure to a less organized random topology. This pattern can be discerned in various different disorders such as Alzheimer ‘s disease, schizophrenia, brain tumors and brain trauma. In Alzheimer’s disease, the randomization of network structure seems to be due to a selective disruption of so-called network hubs. How this comes about, and whether this scenario is specific for Alzheimer’s disease, is a topic of current research. A second pattern consists of pathologically increased connectivity between brain areas; this has been seen during epileptic seizures and in the interictal state, but also in obesitas and to a certain extent in Parkinson’s disease. Pathological hypersynchronization in the theta band could well become important in the diagnosis of epilepsy in daily practice. However, the underlying mechanism is not clear yet. Possible explanations are the disinhibition of the cortex due to failure of the basal ganglia, or pathological plasticity (with excessive outgrowth of new connections) following leasions (traumatic, brain tumor). These hypotheses are currently investigated in an animal model and the macroscopic brain model.

PhD theses 2005-2014:

References:

For comments and suggestions: send an e-mail

copyright C.J.Stam
Contact information: Department of Clinical Neurophysiology VU University Medical Center
Postal address: De Boelelaan 1118 Postal code: 1081 HV Amsterdam The Netherlands
P.O Box: 7057 Postal code: 1007 MB Amsterdam The Netherlands
Phone: 020 4440727 Fax: 020 4444816