Chapter 1 Chapter 1 Introduction Introduction

Chapter 1 Chapter 1 Introduction Introduction

Tinnitus

1.1 A few words on tinnitus Tinnitus is defined as a perception of sound in the absence of any external auditory stimuli (Moller, 2011). It is sometimes referred to as phantom auditory experience. The word tinnitus originates from the Latin tinnire which is translated as to ring (Saunders, 2007). This describes the sensation which most patients report very adequately: a ringing noise in their ears. Based on large epidemiological studies, tinnitus prevalence ranges from 8.2% to 30.3% (Eggermont, 2012). In Belgium, about 11.8% of subjects between 55 and 65 years old suffer from tinnitus (excluding subjects with conductive hearing loss, asymmetrical hearing loss, Menière s disease or any other otological problems) (Demeester et al., 2007). Tinnitus increases in prevalence with advancing age, increasing mainly between the fourth and sixth decade and stabilizing in the seventh decade (Eggermont, 2012). According to the origin of tinnitus, tinnitus can be divided in two groups. It can be attributable to actual somatic causes (e.g. pulsating blood vessel, spontaneous otoacoustic emissions, palatal myoclonus), then it is referred to as objective tinnitus (perhaps 5% of cases). Objective tinnitus can be heard both by the patient and the examiner and is the result of body sounds. On the other hand, subjective tinnitus, the far most common form, refers to tinnitus for which no somatic cause can be detected (Lockwood et al., 2002). It is a sound experience in the absence of any external or internal physical sound source. Even if qualified as subjective, more than half of this tinnitus population report one or more tinnitus 5

onset factors, mainly noise trauma, head and neck injuries, otologic problems or drugs (figure 1). Tinnitus is frequently associated with hearing loss. The hearing level has long been reported as one of the most important variables associated with the prevalence of tinnitus. Some studies even state that the shape itself of the audiogram has a significant, additional impact on the prevalence of tinnitus. For example, in the Belgian study, tinnitus was more common in subjects with high-frequency steeply sloping audiogram than in subjects with a flat audiogram (Demeester et al., 2007). Moreover, damage to the external, middle or internal ear (eg. Otitis media, otosclerosis, Ménière s disease ) leading to hearing loss has long been recognized as possible tinnitus trigger. When occurring together with hearing loss, neuroscientists usually agree that the initial cause of tinnitus has a peripheral origin. However, hearing loss is not a prerequisite to the development of tinnitus; people with hearing loss don t always have tinnitus and people with normal hearing can have severe tinnitus. As we previously stated, tinnitus is a frequent disorder, which might even become more frequent in the future as the population gets older and as the exposition to recreational noise and the related hearing loss increases (MP3...). This is of great important because tinnitus can severely affect quality of life. When asked, the majority of tinnitus patients reports having sleep disturbances and difficulties with daily activities because of their tinnitus (figure 2). In 2% of the tinnitus population the quality of life is considerably diminished (Goebel et al., 2006a). 6

Figure 1: Onset factors reported (data extracted from the Tinnitus Archive a freely available tinnitus database, Oregon Health and Science University) Some patients have tinnitus for decades and are not bothered by it while others are so outrageously bothered that they commit suicide. Interestingly, the same psychoacoustic properties of the tinnitus are more or less to be found in groups of individuals discommoded by their tinnitus and in those who do not complain about it(henry and Wilson, 1995). What makes the difference between those two groups is the presence of psychiatric co-morbidities (Zoger et al., 2006). In fact, severe forms of tinnitus are often accompanied by psychiatric co-morbidities like depression, anxiety, psychosis, personality disorders The relationship between tinnitus and psychiatric disorders is difficult to determine: the psychiatric disorder can evolve as a consequence of the 7

Figure2: Impact of tinnitus on daily activities (data extracted from the Tinnitus Archive a freely available tinnitus database, Oregon Health and Science University). tinnitus, evolve because of the individual s vulnerability factors or may lead to decompensation of a long standing tinnitus(landgrebe, 2011). Tinnitus and psychiatrics symptoms can also be the consequences of a third condition (Landgrebe, 2011). Nonetheless, neuroimaging studies suggest that similar brain areas are involved in tinnitus and psychiatric disorders (e.g. limbic areas) explaining why tinnitus and psychiatric symptoms can be present at the same time. 8

1.2 Tinnitus clinical evaluation As we exposed previously, tinnitus can be the symptom of a wide range of different conditions and can be accompanied by several comorbidities. Unfortunately, there is no objective parameter which makes it possible to measure the intensity of tinnitus. Therefore clinicians and scientists have no other choice than to rely on the subjective assessment of the tinnitus sufferer. We will not explain how to clinically assess and manage tinnitus but will describe some of the clinical evaluations and tests used in our studies to assess the tinnitus patients. 1.2.1 Audiometric evaluation The audiometric evaluation we performed included the following tests: Pure-tone thresholds The pure-tone audiogram was performed with the patient sitting in a soundproof booth. Pure tones of known intensity and frequency were delivered through headphones and an audiometer controlled by the examiner. We used a frequency range from 250 to 8000Hz and an intensity range of 0 to 120dB hearing level (HL). Pure tones were presented to each ear until the threshold of detection was reached. Tinnitus loudness and pitch matching Tinnitus patients were tested to identify the best match to the perceived frequency of their tinnitus. Patients identified the pure tone or white noise from the audiological examination that best matched the center frequency of their tinnitus sensation. Then the patients were asked to evaluate the 9

loudness of their tinnitus. To do that, the examiner modulated the loudness of the matched sound asking the tinnitus patient whether the sound was louder, softer or matches the tinnitus perceived loudness. 1.2.2 Questionnaires A detail history of the tinnitus patient as well as an evaluation of tinnitus severity by means of validated questionnaires were acquired. The case history questionnaire offers the advantage of using standardized questions to provide reliable and complete information on tinnitus patients. Tinnitus questionnaires on the other hand allow us to quantify tinnitus severity (Langguth, 2011). You will find the French version of all the questionnaires used in our studies in the appendix 2. Case history questionnaire (Patient tinnitus questionnaire) During a Tinnitus Research Initiative (TRI) consensus workshop in 2006, a list of questions to be included in a tinnitus case history questionnaire was compiled and a consensus document for patient assessment was published (Langguth et al., 2007). The case history questionnaire we used is the French version of this specific questionnaire. It includes questions about the history and descriptive characteristics of the tinnitus, about specific behavioral, social, interpersonal, and emotional consequences of tinnitus, about factors that may either exacerbate or reduce tinnitus severity, about relevant comorbidities and previous tinnitus treatments. Tinnitus Handicap Inventory (THI) This questionnaire is composed of 25 items with three possible responses for each of these items (Yes/4 points Sometimes/2 points No/0 10

points)(newman et al., 1996). The total score range from 0 to 100, the higher the score the greater the tinnitus severity. Handicap severity categories have been developed based on quartiles calculated for the total THI score: 0-16: no; 18-36: mild; 38-56: moderate; 58-100: severe. The THI is the most widely used tinnitus questionnaire and is one of the questionnaire recommended by the TRI for clinical studies. Tinnitus Questionnaire (TQ) This questionnaire is composed of 52 items. Tinnitus patients have to indicate their level of agreement to each statement using one of the three possible answers: true/2 points; partly true/1 point; non true/0 points(hallam et al., 1988). Seven items are scored inversely (items 1,7,32,40,44,49 and 52) and for these items scores applied are:true/0 points; partly true/1 point; non true/2 points. Twelve items (6, 23, 24, 29, 30, 32, 40, 42, 45, 46, 49 and 52) are not taken into account for the total score or for scoring into the 5 subscales. These items could not be classified unequivocally by factors analysis (Meeus et al., 2007). This questionnaire mainly assesses the emotional distress, the auditory perceptual difficulties and the sleep disturbances caused by tinnitus. Loudness scoring We asked the tinnitus patients to score the tinnitus loudness they experienced during the scanning session directly after the session on a numeric rating scale, ranging from of 0 (none) to 10 (loudest imaginable tinnitus). 11

Tinnitus Models

Despite its high prevalence, there is little consensus regarding the neuropathological origin of tinnitus. The prevailing opinion is that tinnitus is a perceptual consequence of altered patterns of intrinsic neural activity generated along the central auditory pathway following damage to peripheral auditory structures (Eggermont and Roberts, 2004). While the loss of afferent input to the central auditory system can initiate tinnitus, thereafter, central mechanisms are thought to play an important role in its maintenance (Adjamian et al., 2009). A number of animal and human neuroimaging studies have been performed contributing to our current knowledge of the mechanisms underpinning tinnitus. These studies confirmed modification of brain activity in tinnitus patients and tried to disentangle the interactions between auditory, sensory, affective, cognitive and memory areas. Based on these studies and clinical evaluations, several scientists provided tinnitus models trying to explain the underlying neural mechanisms. In the next paragraphs, we will briefly present the most preeminent of these models. 2.1 The Neurophysiological Model of Tinnitus Jastreboff s model implies that two distinct mechanisms are involved in tinnitus(jastreboff, 2011). The first one leads to the generation of the tinnitus perception and is responsible for the sound sensation. The second one deals with the tinnitus-evoked negatives reactions. This division comes from the observation that both high- and low-distress tinnitus patients share the same psychoacoustic characteristics of their 15

tinnitus and that there is a correlation between the tinnitus loudness and the perceived severity of tinnitus. These observations led to the conclusion that different mechanisms should be involved in the generation of the neural signal that causes the tinnitus perception and in the development of negative reactions to this signal (one person s reaction to the tinnitus). Jastrebroff s neurophysiological model proposed that these two systems should necessarily be included when dealing with tinnitus. According to Jastreboff s theory, the neural activity that causes tinnitus is generated within the auditory system, while non-auditory regions are involved in encoding the conscious percept as well as the emotional evaluation of it. The auditory system, while needed for perception of tinnitus, plays a secondary role for clinically relevant tinnitus. The proposed tinnitus model (figure 3), includes different brain systems in order to represent the mechanism of tinnitus. It combines two loops; the higher loop includes conscious areas of the cerebral cortex and account for the acute perception of the tinnitus, the lower loop includes subconscious centers of the brain (limbic system). Indeed, Jastreboff states that in chronic tinnitus patients, the negative effects linked to tinnitus are a consequence of a conditioned reflex mainly involving the subconscious centers. Two categories of negative effects can be observed in tinnitus patients; physiological responses and behavioral responses to tinnitus and two majors systems are involved in these negative effects; the limbic and the autonomic nervous system. 16

Figure 3: The neurophysiological model of tinnitus (reproduced from (Jastreboff, 2011)) Based on this model, Jastreboff developed a famous and widely used tinnitus therapy: the Tinnitus Retraining Therapy (TRT). This therapy combines counseling (modulation of the higher loop) to sound therapy (modulation of the lower loop). It aims at blocking the spread of the signal to extra-auditory regions of the brain, particularly to the limbic and autonomic nervous system, in order to stop the production of the negative effects (habituation of reaction). If successful, a person may still perceive their tinnitus but tinnitus will not bother her/him. 17

2.2 The Neuroscience of Tinnitus: neural synchrony and neural plasticity. Eggermont s theory of the pathophysiology of tinnitus is based on animal studies and implies that peripheral lesions in the cochlea or the auditory nerve produce dysfunctional input to central auditory structures(eggermont and Roberts, 2004; Eggermont, 2005). For example, following hearing loss, the neural transmission of certain frequencies is diminished leading to an imbalance between excitation and inhibition at multiple levels of the projection pathway (midbrain, auditory cortex, and brainstem). This results in diminished lateral inhibition at higher levels of the auditory pathway causing an overall increase in cortical excitability. These dysfunctions alter the neural activity along the auditory pathway leading to modified neural activity and plastic changes. One change that has been well documented in tinnitus animal studies is the cortical reorganization of the auditory cortex following cochlear damage induced by noise trauma. After noise trauma, tonotopic organization in the cortex is changed such that cortical neurons with characteristic frequencies in the frequency region of the hearing loss no longer respond according to their place in the tonotopic map but reflect instead the frequency tuning of their less affected neighbors, expanding the cortical representation of the edge frequencies(eggermont and Roberts, 2004). This modification of the tonotopic representation suggests an analogy with the reorganization of the somatosensory cortex in human amputees. Because hearing loss is a putative cause of tinnitus, neuroscientists believed that this expended representation of the edges 18

frequencies was the substrate of tinnitus. But this expansion, even if functional, is not itself the neural substrate of tinnitus. In fact, when asked, subject identifies their tinnitus sensation to frequencies spanning the region of their hearing loss and not the edges frequencies specifically. Cochlear lesions also lead to modified neuronal activity. Indeed, the neurons with the characteristic frequency of the hearing loss show two forms of modified neural activity: increased spontaneous firing rate (SFR) and increased synchronous firing. Increased synchrony, unlike increase in SFR, might actually be more closely linked to the percept of tinnitus. Indeed, increased synchrony is confined to the reorganized region of the auditory cortex and appears immediately after noise trauma while increased SFR is observed within and outside the reorganized region of the cortex and appears with a delays of days(eggermont and Roberts, 2004). For Eggermont, alterations of neural activity in the auditory structures (increased spontaneous ring rates, increased neuronal synchrony and plastic changes) are putative neuronal correlates of phantom sounds. Chronic tinnitus is a plasticity disorder. 19

2.3 A global Brain Model of Tinnitus Schlee and Weisz proposed a theory to explain how a conscious perception of tinnitus is formed and maintained in the brain(schlee, 2011). It is based on the Deheane global neural workspace hypothesis (Dehaene et al., 1998; Dehaene et al., 2006). According to this model, in order to form a conscious percept of a stimulus two conditions are required: first, there should be a neuronal activity of the sensory cortex of the respective modality and second, this activity should spread to the global neuronal workspace consisting of fronto-parietal-cingulate regions. Applied to tinnitus, two levels of tinnitus-related neuronal processing can be distinguished: the local (or sensory) level refers to the activity in the auditory areas. The global level refers to long-range cortical network of functionally connected brain areas. For a conscious perception of tinnitus an abnormal activity at the sensory level as well as at the global level is required (figure 4). This fits neuroimaging data on tinnitus showing altered activity in the central auditory and non-auditory regions as well as crosstalk between auditory and non-auditory areas in tinnitus(schlee et al., 2008; 2009b; Vanneste et al., 2010a; 2010b; 2011b). Dehaene also argues that neuronal activity of the sensory areas can be enhanced by global workspace top-down influence. Similarly, higher tinnitus distress could be associated with stronger top down amplification. 20

Figure 4: Abnormal activity at the sensory level (auditory) and at the global level is required for conscious perception of tinnitus. A higher top-down influence of the global level to the auditory areas is associated with higher tinnitus distress (reproduced from (Schlee, 2011)). 21

2.4 The Tinnitus Network This model has the particularity to combine all previous exposed theories and to include new concepts based on results obtained from electroencephalographic (EEG) studies. This model has been designed to explain how phantom perceptions (e.g. tinnitus) arise from activity in the brain(de Ridder et al., 2011a). Deprivation of sensory inputs triggers changes in the central nervous system. For tinnitus, peripheral lesion of the cochlea induced abnormal auditory cortex activity (modification of tonotopy, increased spontaneous firing rate, synchronization of firing of neuronal groups and oscillatory changes). But this abnormal activity alone does not lead to phantom perception (tinnitus). For tinnitus to be perceived we need the coactivation of perceptual systems, formed by interconnected circuits forming distributed systems, to be active. De Ridder s claim is that any abnormality in these networks could generate tinnitus. The different networks involved in tinnitus are (figure 5): i. Auditory network (in brown): Sensory deafferentation causes neuroplastic changes resulting in increased synchrony in the primary auditory cortex and, in turn, to topographical map reorganization (see Eggermont theory). ii. Perception network (in blue): The tinnitus percept, per se, is not encoded in the auditory cortex. Awareness of the tinnitus arises when the auditory cortex activity is connected to a larger coactivated global workspace (frontal and parietal) or a selfperception network (see Deheane-Schlee theory). 22

iii. Distress network (in red): Hearing has sensory and affective dimensions and can induce an avoidance behavior. The presence of distress in tinnitus is related to a network consisting of the anterior cingulate cortex (sgacc and dacc), anterior insula, and amygdala. iv. Salience network (in yellow): Salience to tinnitus is reflected by activation of dacc and anterior insula. Both salience (yellow) and distress (red) network brain areas overlap with brain regions involved in central control of the autonomic system (see Jastreboff theory). v. Memory areas (in green): The persistence of the phantom percept is due to memory mechanisms involving the parahippocampal area, amygdala, and hippocampus (green). Figure 5: Brain networks involved in phantom perception (e.g. tinnitus) (reproduced from (De Ridder et al., 2011a)). 23

Functional MRI

While the study of human brain function was previously based on animal studies or on histological observations, new imaging techniques have allowed the study of the brain in vivo. 3.1 Functional brain imaging techniques Functional imaging techniques are traditionally opposed to structural imaging techniques. The latter are well known to the general public and are used in daily medical practice. They include radiography, CT scanning and magnetic resonance imaging (MRI). They allow the analysis of the structural properties of different body tissues and the study of anatomy. Functional imaging, on the other hand, allows the analysis of brain function (e.g. determination of which brain region is involved when performing a specific task). Functional imaging techniques are mainly divided into two categories. The first group includes techniques based on the study of brain electrical activity such as the electroencephalography (EEG) or magnetoencephalography (MEG). These techniques provide high temporal resolution (10-100ms) but a poor spatial resolution (<1cm). The second group includes techniques allowing the detection of changes in local brain activity. It includes the Positron Emission Tomography (PET) and functional magnetic resonance imaging (fmri). 3.1.1 Magnetic resonance imaging The principle behind the use of MRI machines is that body tissue contains lots of water (and hence protons) which gets aligned in a large magnetic field. When inside the scanner, the powerful magnetic field forces the 27

protons to align with the direction of the field. The application of a brief specific radio frequency flips the spin of the aligned protons. After this additional magnetic field is turned off, the protons return to equilibrium. The measurement of the energy released when the protons get back to their initial position gives information on the chemical composition and therefore on the nature of biological tissues at each point of the scanned volume. 3.1.2 Functional magnetic resonance imaging Functional magnetic resonance imaging is a technique used for measuring hemodynamic changes after enhanced neural activity. It uses the Blood Oxygen Level Dependent (BOLD) signal which is based on the different magnetic properties of oxyhemoglobin and deoxyhemoglobin. The deoxyhb is slightly paramagnetic with respect to the brain tissue while the oxyhb is generally diamagnetic. Since the 1890s it has been shown that changes in blood flow and blood oxygenation in the brain are closely linked to neural activity(roy and Sherrington, 1890). The increase in brain activity induces a regional increase of energetic substrate consumption and blood flow. Because the increase in cerebral blood flow is always larger than the oxygen consumption rate, the net effect of neural excitation is thus a drop in the deoxyhemoglobin concentration which in turn increases the BOLD signal (BOLD hemodynamic response). fmri is widely used to detect and delineate regions of the brain that change their level of activation in response to specific stimuli and tasks. fmri studies typically use echo-planar sequences that are sensitive to 28

changes in BOLD signal which reflects, indirectly, neuronal activation. Using fmri, one can obtain activation maps that depict the average level of engagement or response to a specific task or stimuli of different regions of the brain. To do so, task-activation fmri methods are usually performed, revealing the activation patterns produce by specific conditions. These activation patterns are detected when comparisons are made between a task-state and a control-state. While being helpful for brain mapping this technique does not allow to reveal how brain regions interacts with one another. However looking at connectivity is essential to provide a complete understanding of the (ab)normal brain. Function of any brain region cannot be understood in isolation but only in conjunction with regions it interacts with (Seghier et al., 2010). Current theories consider that the human brain is an integrative network of spatially distributed, but functionally linked regions that continuously share information with each other (van den Heuvel et al., 2009). The study of this network is generally described as functional connectivity. While initially focusing on brain activity in response to specific tasks, neuroscientists now also study the brain spontaneous activity, in other words they are looking at the functional organization of the brain at rest. 3.2 Functional connectivity Currently, numerous methods are used to detect or measure functional connectivity using fmri. However, while the methods vary they all focus on correlations measures derived from BOLD data. 29

The methods used to study functional connectivity can be divided in two categories: those that use apriori knowledge or hypotheses to limit the analysis to a restricted set of regions (Model-dependent methods) and those that are data driven and attempt to map connectivity in the whole brain (Model-free methods)(rogers et al., 2007). 3.2.1 Model-dependent This first category includes, among others, seed-voxels correlation maps. Seed-voxel analysis is the most straightforward way to examine functional connection of a specific brain region with all other brain regions. It requires the extraction of the BOLD signal time series of a pre-defined region of interest (ROI), usually called seed, and then the identification of regions showing significant correlation with this time-series. This technique gives functional connectivity maps showing which regions are functionally linked to the initial ROI. Advantages of this technique are its relative simplicity and the straight forwardness of the results (Buckner and Vincent, 2007; van den Heuvel et al., 2009). Disadvantages are the need of apriori assumptions since the information gained are related to and only to the functional connections of the selected seed region. 3.2.2 Model-free To examine whole-brain connectivity patterns, model-free methods have been introduced. These methods do not require a priori assumption and allows the exploration of multiple whole-brain networks. Several modelfree methods have been applied for the studies of fmri time series, 30

including principal component analysis (PCA), independent component analysis (ICA) and hierarchical, Laplacian and normalized cut clustering. Of those models, the most commonly employed and the one showing the most consistency is ICA. ICA is a computational method used for separating a multivariate signal into additive subcomponents supposing the mutual statistical independence of the non-gaussian source signals. Applied to fmri, ICA divides the fmri BOLD signal into independent components by maximizing the statistical independence of the estimated components(mckeown et al., 1998). Each independent component is a combination of a spatial map and a corresponding time course. Each network can be interpreted as a network of similar BOLD activity. As opposed to seed-voxel approach, ICA does not need any apriori assumption. It is a hypothesis-free method that allows the exploration of multiple whole-brain networks(rosazza and Minati, 2011). It is then easily applied to the study of pathologies for which detailed apriori models of the brain activity are not available yet. Moreover, each independent component either corresponds to neural functional network/common physiological activity or to neuroimaging artifact, making it a good tool to separate neuronal from artifactual signal. However, one of the possible disadvantages of the technique is that the components are not always easy to interpret. 3.2.3 Effective connectivity The methods we have just described consider only correlation and ignore issues of causation and influences. Methods that attempt to address 31

directionality exist; this category of techniques is referred as effective connectivity. They include among others psychophysiological interaction (PPI), Granger Causal Mapping or Dynamic Causal Modeling (DCM). The most recent and promising technique is DCM, a hypothesis-led approach to understand distributed neuronal architectures underlying brain responses (Friston et al., 2010). It determines causal relationships across potentially distributed neural networks and aims to explain, quantitatively and mechanistically, how observed (fmri) responses are generated (Friston et al., 2003; Stephan et al., 2010). Generally, hypotheses about the neuronal mechanisms that underlie the experimental measurements of brain responses are created. Then, a Bayesian model selection is used to quantify the evidence for one hypothesis over another(friston et al., 2010). So, DCM is a hypothesisdriven approach that test specific hypotheses on brain architecture. 3.3 Resting-state fmri Initially functional brain imaging methods have emphasized task-induced increases in regional brain activity associated with the execution of a wide variety of tasks. More recently, however, a great interest has been developed to the study of the functional organization of the brain at rest(bressler and Menon, 2010; Fox and Raichle, 2007). Indeed, associated to cortical increases in activity depending on the task performed, researchers also encountered task-induce decreases. While these deactivations could be task-specific (decrease of the activity in sensory 32

system not involved on the processing of the stimulus) some were independent of the activity performed, varying little across various tasks (figure 6). This observation served as a start for the concept of a baseline or default-mode of the brain which can be observed when subjects are awake and resting with their eyes closed. Figure 6: Regions showing decrease of activity independently of the task 3.3.1 The brain is organized into dynamic, anticorrelated functional networks The default-mode network, encompassing precuneus, mesiofrontal and temporo-parietal junction areas(raichle et al., 2001), appears to have strong negative correlations (anti-correlations) with a network of brain regions commonly activated during the performance of goal-directed cognitive tasks termed the task-positive network(raichle et al., 2001). This anticorrelation organization seems to be an intrinsic feature of the 33

resting human brain. Fox and colleagues showed that widely distributed neuro-anatomical networks are organized through both correlated spontaneous fluctuations within a network and anticorrelations between networks. Some neuroscientists even stated that anticorrelations may reveal key aspects of functional organization of the brain (Fox et al., 2005; Kelly et al., 2008). 3.3.2 Resting-state brain networks The previously described organization into anticorrelated networks is not the only interesting features about the resting brain. The use of functional connectivity techniques has helped neuroscientists to explore the functional organization of the brain at rest. Functional connectivity studies of the resting-state examines correlation in slow (<0.1Hz) spontaneous fluctuations of the BOLD signal. Biswal and colleagues, in 1995, were the first to show that correlation of low frequency fluctuations is a manifestation of functional connectivity of the brain(biswal et al., 1995). They showed that under resting conditions the BOLD signal of the left sensorimotor cortex correlated with fluctuations of the BOLD signal in contralateral sensorimotor cortex. Indeed, these fluctuations are shown to be coherent across widely separated (although functionally related) brain regions, constituting resting state networks (Beckmann et al., 2005; De Luca et al., 2006). Not only can we find the Default network but other networks involved in visual, motor, language, and auditory processing can also be consistently found in healthy subjects(laird et al., 2011; van den Heuvel et al., 2008). The default 34

networks and other networks can be separated from each other from a single resting-fmri dataset using their distinct temporal characteristics. Multiple analysis techniques have been applied to the study of BOLD spontaneous activity. The most frequently used are the seed-voxel approach and the ICA. Although different, studies that compared the two techniques showed similar results (Rosazza and Minati, 2011). Using both methods, functional connectivity studies reported several network of functionally connected regions: resting-state networks also referred as components. And as we said before, these networks are quite consistent across population. They also have a functional relevance. In fact, resting networks mirror networks of brain regions observed during active paradigms(smith et al., 2009). In other words, resting networks represent regions that are known to share and support cognitive functions(rosazza and Minati, 2011). Regions that tend to be activated together during active tasks show correlations in their spontaneous activity. In a recent study comparing resting-state ICA map and task-based activity map (derived from the BrainMap dataset) through spatial correlation, Smith et al.(2009) found that 10 paired of map were unambiguously paired between datatsets. These ten networks could then be considered as neuronal and could easily be linked to behavioral domains. These are the ten networks we decided to look at in our third study (figure 7). We will briefly introduce them and, based on Smith at al., briefly describe to which behavioral domains they are linked (Smith et al., 2009). 35

Auditory network: Included regions: superior temporal gyrus, Heschl s gyrus and posterior insular. It corresponds to action-execution-speech, cognition-language-speech, and perception-audition paradigms. Cerebellum network: Included regions: cerebellum. This network corresponds to action-execution and perceptionsomesthesis-pain paradigms. Default mode network (DMN): Included regions: precuneus/posterior cingulate (PCC), lateral parietal cortex, mesial prefrontal cortex. It is the resting-state networks that received the most attention from the neuroscientific community. This DMN is the network that is seen as deactivated in task-based fmri experiments. It has been associated to introspective mental processes. External control network (ECN - right or left): Included regions: frontoparietal regions. These components are the only ones to be strongly lateralized. They correspond to several cognition/language paradigms. In addition, the right ECN corresponds strongly to perception-somesthesis-pain paradigms and the left ECN to cognition-language paradigms. Salience network: Included regions: several medial-frontal areas including the anterior cingulate and paracingulate. It corresponds to several cognition paradigms, as well as action-inhibition, emotion and perceptionsomesthesis-pain paradigms. 36

Sensorimotor network: Included regions: supplementary motor area, sensorimotor cortex, secondary somatosensory cortex. As we said before, this network was the first RSN to be identified by Biswal and colleages in 1995. It corresponds to action-execution and perception-somesthesis paradigms. Visual networks: Included regions: medial, occipital pole and lateral visual areas. They correspond to visual paradigms. In addition, the occipital lobe corresponds also to cognition-language-orthography and the lateral visual maps to cognition-space paradigms. Figure 7: representation of the ten resting-state network on brain slices(maudoux A., in prep.) 37

3.3.3 Origin of the BOLD fluctuations Works has been done to unravel the neurophysiological origin of the blood flow changes and how these modifications are related to the neuronal activity. Even if a lot have been learned about the neurobiology of functional brain signal we still don t know the exact significance of the change in blood flow. The notion that these changes simply reflect an adaptation to the energy demands of the brain (oxygen and glucose), as attractive as it might seem, is oversimplified because an adequate reserve of oxygen and glucose is already available without the need of increase in blood flow (unextracted oxygen in the blood and glycogen reserve in the astrocytes)(raichle and Mintun, 2006). Several researches looking at the metabolism of glucose in the brain and the role of astrocytes led to the hypothesis that astrocytes and the glycolytic machinery they house are involved in the coordination of the metabolic and circulatory requirements associated with changes in brain function(kasischke et al., 2004). Studies looking at what in the behavior of neurons accounts for blood flow changes concluded that the fmri BOLD signal correlates better with local field potentials than with the spiking activity of neurons(logothetis et al., 2001; 2003; 2004). Similarly, the understanding of the neurophysiology underlying coherent BOLD fluctuation in a resting brain is still not fully understood. What we know is that these fluctuations seem to reflect functional connectivity (Biswal et al., 1995). We also know that they cannot only be the expression of unconstrained cognition (mind-wandering, daydreaming) (Raichle and Snyder, 2007). First, because imposed tasks evoked responses contribute to a small increase in brain energy consumption and there is no reason to think that unconstrained though might be more energetic. Second, 38

because resting-state activity persists in slow wave sleep and during anesthesia (Vincent et al., 2007), two states during which cognition is absent or at least very weak. Third, because coherent spontaneous BOLD fluctuation are observed across a wide variety of brain regions and it will be difficult to imagine that as many brain regions could be modulated by the same behavior, each in its own coherent manner(raichle and Snyder, 2007). Raichle suggests that spontaneous BOLD fluctuations observed during the resting state may be partitioned into two layers. The first layer relates to the unconstrained behavior or conscious mentation that subjects perform during resting conditions. The second layer (the intrinsic activity), underlies the first layer and persists across different states or conditions. This layer of intrinsic activity is probably responsible for most of the spontaneous BOLD fluctuations observed during awake resting conditions(fox and Raichle, 2007) and may represent the fundamental and intrinsic property of functional brain organization(vincent et al., 2007). 3.3.4 Why using resting-state fmri to study tinnitus? The study of functional connectivity in resting-state tinnitus patients is important for 3 reasons (based on (Rosazza and Minati, 2011)). 1- Spontaneous activity is the most metabolic demanding component of neural activity. The brain represents about 2% of the total body weight yet it accounts for 20% of all the energy consumed. Moreover, the additional energy consumption associated with imposed task is remarkably small (Raichle and Mintun, 2006). Studying task-evoked activity can then only shed light to a small part of the functional activity performed by the brain. 39

2- Resting-state fmri do not rely on active participation by the patients. It is free from the potentially confounding effects of differences in the level of task performance. 3- This acquisition technique is easily applied in clinical setting. Indeed, fmri is a noninvasive, increasingly available technique with relatively high spatiotemporal resolution. Moreover, maps of spontaneous network correlations have been proposed to provide tools for the understanding of clinical conditions. fmri resting-state paradigms have, for example, been applied to the study of hypnosis [16], anesthesia [17] and various neurological disorders including dementia [18,19], depression [20] disorder of consciousness [21,22] and auditory hallucinations [23]. This techniques has be used to detect differences between patients and controls and, more importantly, to correlate the resting-state differences to clinical variables or behavior (Dosenbach et al., 2007; Fox and Raichle, 2007). 40