Summary and general discussion

Working in the field of neuroimaging is just like looking for a flower in a meadow.

The objectives of this work were to investigate the frontal-striatal and limbic circuits during emotional and cognitive processes in patients with obsessive-compulsive and related disorders and to determine the specificity of possible abnormalities for these disorders. To address these issues, task related changes in regional cerebral blood blow (rCBF) and blood oxygenation level dependent (BOLD) signal were measured, using both positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). The neuroimaging studies described in this thesis started in 2000. As can be seen from a review of this field (part I), almost half of the relevant studies have been published after 2000. This illustrates the tremendous developments within the field and stimulates to further innovation of imaging techniques and neuropsychological paradigms (part II). In this last chapter of the thesis, significance and limitations of the results of this thesis (part III) will be discussed. Subsequently, a view is presented on where we are now and where we are going.

Part I

Other people have traveled the meadow of neuroimaging before. They made maps of their routes and they described the flowers they picked on their way. Some of their paths appear to have a dead-end; others seem to promise infinite views. Before starting your walk, it is important to study the maps of your predecessors. Which paths look most promising? And are there any hidden paths that have never been walked before?

Part I of this thesis critically reviewed the results of neuroimaging studies in obsessive-compulsive disorder (OCD) and panic disorder (PD), and illustrated some of the strengths and limitations of the various imaging techniques available.
Chapter two reviewed two decades of neuroimaging research in OCD, ranging from morphological measurements to functional experiments at baseline, after symptom provocation, and during neuropsychological performance. The overview clearly shows a shift in thinking about OCD over time. Theoretical models were first based on the idea of an exclusive, or at least prominent, role of the striatum. Subsequent theories emphasized the functional significance of the frontal-striatal circuits1, and the differentiation (mainly dorsal versus ventral) within these circuits. Recently, limbic structures, mainly the amygdala, have been implicated as well2.
To date, imaging findings in OCD may be summarized as follows: 1) resting state in OCD patients is associated with increased activity of the ventral parts of frontal-striatal circuits - probably reflecting ongoing emotional and cognitive processing related to tonic symptomatology - and decreased activity in the dorsal parts of the prefrontal cortex, 2) during symptom provocation, activation of limbic structures (mainly the amygdala) and additional recruitment of ventral frontal-striatal regions may reflect the processing of salient information and emotional responses, and 3) exaggerated responses (anxiety and/or distress) may, at least in part, be the result of insufficient suppression or top-down control by the dorsal frontal-striatal circuit. The results of the present thesis provide a major contribution to this conclusion. Nevertheless, it should be realized that this description is a simplification of the complex mechanisms underlying the disorder and that it does not fully answer the question about the disorder-specificity of the assumed pathophysiological processes.

In order to investigate the issue of disorder-specificity, comparisons between OCD and related disorders are warranted. The comparison with PD enables the differentiation between aspecific distress or anxiety related characteristics and OCD specific deficits. Chapter three illustrated the impact of neuroimaging experiments on the understanding of PD. Neuroimaging studies in PD, compared with those in OCD, originate from a different tradition. So far, the work in PD largely consists of pharmacological challenge studies in order to induce panic attacks, and receptor ligand studies, most of which have addressed the functioning of the GABA-benzodiazepine complex. Theoretical models of PD are based on the assumption that the main components of the illness - anticipatory anxiety, panic attacks and phobic avoidance - are linked to distinct neuroanatomical systems in the brain: the temporal lobe structures (mainly the amygdala, hippocampus and parahippocampal gyrus), the brainstem, and the prefrontal cortex, respectively3;4. Results of imaging studies in PD only partly support this hypothesis. As in most imaging studies a clear description of the emotional state at the time of data acquisition is lacking, proper neuroanatomical differentiation between panic attacks, anticipatory anxiety and other related symptoms is not yet possible.

Part II

Working in the field of neuroimaging is like looking for a rare flower in a meadow. The flower of interest is often overgrown by weeds. And the colorful spot seen in the distance might just as well be a plastic bag.

Noise and artifacts are the weeds and plastic bags of neuroimaging. It requires a critical attitude towards the experimental data acquired in order to improve signal-to-noise ratios and to differentiate between true and artificial findings. Only this attitude will lead to further innovation of imaging techniques. In Part II, two methodological aspects were investigated.
In chapter four an important methodological aspect of PET imaging has been investigated in detail, namely the effect of subject motion on the quality of the attenuation correction. Correction for tissue attenuation is a vital step in obtaining quantitative PET data. The most commonly used method to correct for tissue attenuation is by direct measurement using a separately acquired transmission scan. However, subject motion between transmission and emission scans may result in misalignment and, therefore, in erroneous attenuation correction. When standard attenuation correction is used, this mismatch can result in either underestimation or overestimation of regional activity concentrations. In case of random motion, transmission-emission mismatch will result in diminished signal-to-noise ratios and false-negative findings (type-II errors). On the other hand, as illustrated by a clinical case report, task related motion will result in systematic reconstruction artifacts and consequently in false-positive results (type-I errors). During symptom provocation experiments in psychiatric patients task related motion may easily occur when the presented emotional stimuli elicit muscle contractions accompanying emotions of fear, disgust or discomfort. For that reason, head positions may differ between the provoked and neutral state.
It was investigated whether the implementation of an image registration (IR) method, which allows for motion-corrected attenuation correction, improved the accuracy of H215O PET analyses. The IR method, first described by Andersson et al.5, is based on 3 assumptions: 1) the transmission scan and the first emission scan are well-aligned, 2) transformation matrices can be derived accurately from non-attenuation corrected emission scans, and 3) forward projected transmission scans result in the same attenuation correction factors as those derived directly from transmission scans when patient motion is absent. The implementation of the IR method consisted of the following five steps: 1) reconstruction of emission scans without attenuation correction, 2) calculation of transformation matrices between the first and the subsequent emission scans, 3) application of inverse transformation matrices to the reconstructed transmission scan, 4) forward projection of the transformed transmission scans to yield motion-corrected attenuation correction factors, and 5) application of these factors in the reconstruction of the emission scans. To evaluate the accuracy of this method, phantom studies (using a solid homogeneous 20-cm-diameter cylindrical phantom and a 3D brain Hoffman phantom) as well as studies in human subjects were performed. The results were compared with three alternative methods: 1) standard measured attenuation correction without motion correction, 2) calculated contour-based attenuation correction, and 3) no attenuation correction.
In case of subtraction rather than quantitative analyses, attenuation correction might not be necessary, thereby bypassing the problem of subject motion. Indeed, results of the clinical case report and human validation study confirmed that motion induced false-positive results, as obtained with standard attenuation correction, greatly reduced in case of subtraction analyses without attenuation correction. This method, however, is likely to generate extracranial (or border) artifacts, and is not suitable in the case of quantitative analysis (e.g. such as required in ligand studies). In addition, the calculated contour based method proved to be suboptimal, probably due to difficulties in accurately defining contours.
The elaborate evaluation of the IR method showed that this method reduces noise for the group subtraction analysis and removes type-I errors in case of task related motion, while improving the signal from expected activated areas. The phantom data showed excellent accuracy of the algorithms for image registration, reconstruction, and forward projection of the transmission scan data. Therefore, the IR method should be considered as the first choice for attenuation correction in PET activation studies. Although promising, the application of this method for dynamic ligand studies still needs validation.

Another domain of methodological consideration in neuroimaging research is the experimental paradigm used to visualize a specific cognitive and/or emotional state. An almost infinite battery of neuropsychological tasks can be used to investigate various neuropsychiatric disorders. To prevent contamination with aspecific findings, it is important to invest in the design of hypothesis driven paradigms. Comparison of task related neuronal correlates between different groups requires a sensitive paradigm that specifically addresses the function of interest and incorporates a correction for potential differences in task performance.
In chapter five the design of a parametric self-paced pseudo-randomized event-related fMRI-version of the Tower of London task has been described. This planning task, originally developed by Shallice6, is supposed to address the flexibility of the frontal-striatal and visuo-spatial circuits. A self-paced, parametric design allows for flexibility in response as well as for comparisons between subjects and/or groups at each task level, providing the opportunity to investigate groups with varying levels of performance. As performance is likely to deteriorate at higher complexity, control for performance differences is especially relevant for parametric designs. In addition to variation in reaction times, performance may also vary in hit frequency. Event-related analysis enables a selective analysis of correct responses, by separate modelling of the false events. Moreover, adequate randomization of trials is possible using an event-related design.
To evaluate this version of the Tower of London task, 22 healthy control subjects were investigated. Compared with baseline, planning activity was correlated with increased blood oxygenation level dependent (BOLD) signal in the dorsolateral prefrontal cortex (DLPFC), striatum, pallidum, (pre)motor cortex, supplementary motor area (SMA), and visuo-spatial system (precuneus and inferior parietal and parietal-occipital cortex). Task load was associated with increased activity in the same regions, and additional recruitment of the left anterior prefrontal cortex, a region supposed to be specifically involved in 3rd order executive functioning.
Two different aspects need to be considered when interpreting these results. First, compared with the main effects (i.e. planning versus baseline contrast), task load effects are more specific, although less sensitive, in their measurement of planning related changes in BOLD response. Second, while increasing task load, other processes, not specific for planning, might also increase (e.g. working memory). Therefore, the investigation of the neuronal correlates of planning might be best visualized by the combination of main and task load effects.

Another methodological issue in neuroimaging of psychiatric conditions is the experimentally induced state of interest. Although not addressed in part II, probably the most important questions remain: in which state is the patient at the moment of data acquisition and to what extent do researchers succeed in inducing and maintaining a desired emotional and/or cognitive state in subjects within an experimental setting? It is questionable whether the usual subtraction designs really succeed in isolating only the neuropsychological function of interest, even with a carefully matched baseline condition (e.g. for stimulus complexity and motor demands). Moreover, aspecific factors, such as arousal or distress, may disturb the visualization of the task related neuronal correlates.
Being aware of the difficulty of capturing the mental state of interest, an attempt was made to develop proper paradigms in order to investigate three different mental states: 1) emotional perception during provoked contamination fear, 2) higher-order cognitive function during executive performance, and 3) the interaction between emotional and cognitive processes during a paradigm addressing attentional bias.

Part III

We started our walk. And how lucky we were! We found beautiful flowers on our way.

Chapter six described the neurophysiological correlates of experimentally induced symptoms of contamination fear in OCD patients. This was not the first study to investigate the symptomatic state in OCD patients. Inconsistencies in earlier experimental results due to various methodological concerns (lack of control groups, medicated subjects, idiosyncratic tactile stimuli, off-on designs, limited sample size), however, asked for replication of the findings in 1) a homogeneous group, 2) of medication-free OCD patients, 3) in comparison with healthy control subjects, 4) during a randomized design, 5) with standardized visual stimuli. To this end, a symptom provocation experiment was performed in 11 medication-free OCD patients with contamination fear and 10 healthy control subjects. Using oxygen-15 water (H215O) PET, task related changes in rCBF were measured during a randomized block design containing visual presentations of 'dirty' and 'clean' surroundings. To prevent type-I errors due to task related, subject motion induced transmission-emission mismatches, the IR method (as described in chapter 4) was used to reconstruct images.
Behavioral data showed that the provocation design was successful in inducing both subjective distress and obsessionality. Whereas obsessionality scores increased in OCD patients as well as in healthy control subjects, subjective distress scores only increased in the patient group. In other words, subjects in both groups experienced 'ritualism', but only OCD patients also became anxious. Moreover, OCD patients became sensitised rather than habituated as a result of the contamination related stimuli. Imaging findings corresponded with these behavioral results. Healthy control subjects showed provocation induced increased rCBF in the left DLPFC and right caudate nucleus. The recruitment of frontal-striatal regions in controls in response to emotional stimuli seems to reflect 'normal ritualism' or top-down control of the emotional response. In contrast, in OCD patients the provoked symptomatic state was correlated with increased rCBF in the left amygdala. Moreover, in this group a time by condition interaction effect was found in the right amygdala, reflecting sensitization. The involvement of the amygdala in OCD during the symptomatic state is in agreement with the literature on the central role of the amygdala in evaluating the emotional significance of external stimuli and fear responses7;8. In OCD research, however, the role of the limbic system, mainly the amygdala, in the pathophysiology of the symptomatic state has been underestimated so far. Based on these results, it is suggested that the observed differences between OCD patients and controls reflect a failure of the frontal-striatal circuitry in OCD patients to control the processing of negative disease-relevant stimuli, resulting in an inadequate fear response, involving both amygdalae.

The flexibility of the frontal-striatal system in OCD during an emotionally neutral state was investigated using an fMRI version of the Tower of London task, as described in chapter seven. To control for differences in performance between OCD patients and healthy control subjects, and to increase the specificity of the experimental effect of interest, a parametric self-paced pseudo-randomized event-related design (as described in chapter 5) was used.
Twenty-two medication-free OCD patients and 22 healthy control subjects were included. Behavioral data showed decreased performance scores across all levels in OCD patients compared with control subjects, whereas reaction times (RTs) were significantly longer in OCD patients only during the two easiest task levels. Within groups, performance was not significantly correlated with symptom severity and subjective distress. Compared with controls, imaging results showed decreased task-associated activation in OCD patients in several regions previously found to be involved in planning, in particular DLPFC, basal ganglia and parietal cortex. Task load correlated with increased activity in the left DLPFC in control subjects compared with OCD patients. In contrast, the OCD group showed increased - presumably compensatory and/or stress-related - involvement of bilateral cingulate, ventrolateral prefrontal and parahippocampal cortices, left anterior temporal cortex and dorsal brain stem. The decreased responsiveness of the frontal-striatal system, described in OCD patients, was not correlated with ratings for symptom severity and subjective distress.
The results further support the involvement of a frontal-striatal dysfunction in the pathophysiology of OCD. Some alternative explanations might, however, be postulated in relation to the present differences between OCD patients and healthy control subjects, and some important concerns remain. First, it might be argued that only the parametric task load contrast represents a valid and specific comparison between different groups, since the main effects for task might have been confounded by differences in baseline activity. Baseline differences may result from both resting state differences between patients and controls, and dissimilar cognitive or emotional processes during the baseline task in patients compared with controls. Second, the finding of decreased frontal-striatal responsiveness during planning performance might be at least partly explained by poor matching for intelligence. Although only correct responses were selected to control for performance differences, and in spite of the fact that post-hoc analyses of covariance with regard to educational level were performed (showing that the crucial fMRI task by group interaction effects in the striatum and DLPFC persisted after correcting for differences in education), the present findings need replication in an appropriately matched sample. Third, the question remains whether the frontal-striatal dysfunction during planning is specific for OCD, or whether it reflects an aspecific characteristic of anxiety or even neuropsychiatric disorders in general. This issue can be addressed by comparing the task related activation patterns across different patient groups. The present data set enables a comparison between OCD, PD and hypochondriasis, and this analysis will be performed in the near future. If patients with PD and hypochondriasis show normal frontal-striatal recruitment during planning, similar to controls, it might be concluded that the observed frontal-striatal dysfunction in OCD is specific for the pathophysiology of OCD. In contrast, if patients with PD and/or hypochondriasis also have decreased frontal-striatal responsiveness during planning performance, in common with OCD patients, a closer look at the shared features will contribute to a better understanding of common characteristics across disorders and /or aspecific factors confounding the task related activation patterns.

In chapter eight, addressing attentional bias, emotional and cognitive processes are strongly interwoven. Difficulty in inhibiting irrelevant information is a key feature of OCD. Because most of their attentional resources are allocated to threat cues related to their concerns, OCD patients are limited in their ability to selectively attend to relevant stimuli, whilst simultaneously ignoring irrelevant competing stimuli. As the critical process of gating, i.e. inhibiting irrelevant information, has been linked to frontal-striatal function and the evaluation of emotional stimuli to limbic function, the investigation of attentional bias to disease-relevant emotional cues might contribute to the understanding of the altered function of both frontal-striatal and limbic circuits in OCD. Moreover, comparisons across related disorders are needed to address the specificity of the dysfunction.
To investigate the neuronal correlates of attentional bias across related anxiety disorders, cognitive and emotional Stroop task related BOLD responses were measured in medication-free patients with OCD (N=16), PD (N=15), and hypochondriasis (N=13), and the behavioral and imaging results of these patient groups were compared with those of 19 healthy control subjects. Contrasts of interest were the cognitive interference effect (incongruent versus congruent color words), the OCD related emotional interference effect (OCD related negative words versus neutral words) and the panic related emotional interference effect (panic related negative words versus neutral words).
Cognitive interference in all patient groups relative to controls was correlated with recruitment of additional posterior brain regions, but performance was impaired only in OCD patients. In OCD, color naming of disorder specific (OCD related) words only was associated with increased activation of ventral frontal-striatal and limbic regions, including bilateral amygdala, although performance was not abnormal. In contrast, patients with PD and hypochondriasis showed increased activation of ventral and widespread dorsal frontal-striatal regions during both OCD and panic related words. In addition, in PD patients, the speed of color naming panic related words was significantly reduced, which was associated with increased activation of the right amygdala and hippocampus.
These results imply clear differences between OCD patients on the one hand and PD and hypochondriasis patients on the other. The disorder specific neuronal response in OCD mainly involves ventral brain regions, which are assumed to be implicated in emotional appraisal of emotional cues and unconscious fear responses. Attentional bias was found to be more generalized in patients with PD and hypochondriasis, involving both ventral and dorsal brain regions, which reflects not only unconscious emotional stimulus processing, but also increased cognitive elaboration of the initial emotional response.

Where we are now?

Now that we've returned from our journey the time has come to ask ourselves what we actually found. Did we really pick the specific beautiful flowers we were looking for, or have we been seduced by the painted flowers of van Gogh?

Taking together the results described in part III, it can be hypothesized that altered dorsal frontal-striatal function in OCD patients is responsible for 1) decreased inhibition of ventral frontal-striatal and limbic recruitment in response to disease-relevant emotional cues and 2) decreased executive performance. This hypothesis, if confirmed, leaves several questions to be answered.
First, the term frontal-striatal appears to be too broad to capture the subtle alterations of brain functioning in OCD during the various emotional and/or cognitive conditions of interest. A distinction between dorsal and ventral frontal-striatal circuits is generally accepted in both the neuroanatomical and neuropsychiatric literature. However, the results of the present study do not allow us to unequivocally associate any of the studied OCD emotional and cognitive deficits to a specific striatal region or a particular prefrontal cortical circuit. A key issue seems to be how cognitive and emotional functions interact or, in other words, dorsal and ventral frontal-striatal circuits might influence each other. Moreover, there appear to be differences in interpretation as to connectional characterization of the dorsal and ventral striatum in the neuroanatomical and the neuropsychiatric literature. Whereas in neuroanatomical models the limbic (or ventral) striatum is defined as the area that receives hippocampal and amygdaloid input and is associated with emotional and motivational functions9, in psychiatric models the hippocampus belongs to the dorsal circuit8. Close collaboration between psychiatrists and anatomists is required in developing future theoretical models of psychiatric conditions.
Second, it is questioned whether anxiety is the key feature of OCD, leading to the repetitive behaviors, or whether OCD is a primary 'cognitive' disorder and anxiety just the spin-off. In other words, which level of emotional processing is most abnormal: the appraisal of the emotional significance of a stimulus, the subsequent fast and unconscious behavioral response, or the higher-order evaluation and modification of this initial response? This issue is hard to investigate and longitudinal designs are needed to differentiate between cause and effect. The present working hypothesis implies a primary failure (hypofunction) in the dorsal frontal-striatal circuit rather than a primary deficit (hypersensitivity) of the limbic or ventral frontal-striatal circuits. Support for a primary deficit in the DLPFC stems from imaging studies in children, showing altered maturation of the DLPFC in pediatric OCD10. Although replication is needed, this suggests a primary 'cognitive deficit'. To date, however, it has not been possible to prove this theory and the opposite hypothesis still cannot be rejected. As stated by Damasio11 and LeDoux7, trying to disentangle the issue might lead to a Cartesian error. Separation of emotion and cognition implicitly involves an artificial divorce of two closely interacting partners. Interactions between the different functional neuronal circuits subserving emotional and cognitive processes and the effects of the various neurotransmitters, from early development to the mature adult brain, are poorly understood and need further elaboration.
Third, the hypothesis suggests a direct top-down control mechanism from the DLPFC to the limbic circuit (mainly the amygdala). Although this sounds attracting, neuroanatomical evidence for projections descending directly from the DLPFC to the amygdala is rather weak. Whereas connections of the amygdala with orbitofrontal and medial prefrontal areas are robust and bi-directional, connections with lateral prefrontal areas are sparse, uni-directional and primarily ascending12;13. One possible way in which diverse streams of information could guide behavior would be through the rich interconnections between prefrontal areas9, involving cortico-cortical and cortico-thalamo-cortical connections. Whereas the latter have long been thought to be organized in a strict reciprocal manner, recent evidence indicates that different cortical areas might be interconnected via the thalamus14. However, little is known about the complex direct and indirect connections subserving the communication between emotion and cognition. An important problem with the interpretation of neuroanatomical tracing studies is that these are at best carried out in non-human primates necessitating the extrapolation of the results to the more complex human brain. It is assumed that humans differ from animals in the way cognition is used to modify instinct-driven behavior. Possibly, humans differ from non-human primates in the complexity of the neuroanatomical connections between dorsolateral prefrontal areas and limbic structures (e.g. the amygdala). The same reasoning applies to the opposite direction of the reciprocal interactions between emotion and cognition. Although psychiatrists and psychologists are easily inclined to assign a role for anxiety in executive impairment in patients, it is unclear in which way the amygdala might influence dorsolateral prefrontal functioning.
Fourth, OCD is not the first or only psychiatric disorder in which a frontal-striatal dysfunction has been implicated. Not only anxiety disorders, but also psychotic disorders (e.g. schizophrenia), movement disorders (e.g. Tourette's Syndrome) and mood disorders (e.g. major depressive disorder) have been attributed to an imbalance of the frontal-striatal circuits. This phenomenon might be explained in two different ways. On the one hand, different subregions of the frontal-striatal circuits, subserving different emotional and/or cognitive processes, might be involved in different neuropsychiatric disorders. Even in the case of a disorder specific dysfunction, the question remains whether the neurophysiological alterations reflect either state or trait characteristics of the disorder. Longitudinal follow-up measurements, for instance before and after treatment, may contribute to a better understanding of this issue. On the other hand, various neuropsychiatric disorders might share a common feature. These features might be related to the illness (e.g. feelings of stress and hopelessness) or to aspecific demographic and behavioral characteristics (e.g. educational level, unemployment, and smoking). Another common feature might be temperament, i.e. the inborn psychological profile. Schwartz et al.15;16 showed that adults who, in the second year of their life, had been categorized as inhibited, display amygdala hyperresponsiveness to novel versus familiar faces in comparison with those previously categorized as uninhibited.

Where are we going?

To date, almost all neuroimaging results in OCD have been based on comparisons between OCD patients and healthy control subjects. There are two questions with respect to the issue of disorder-specificity: 1) is a neurophysiological differentiation possible between OCD subcategories, and 2) are the frontal-striatal and limbic activation patterns found in OCD specific for this disorder or is there overlap with other anxiety, mood and even psychotic disorders? To answer these questions, two opposite approaches are needed: 'splitting', i.e. contrasting subcategories within OCD, and 'lumping', or grouping together of OCD patients with patients suffering from other neuropsychiatric disorders, thereby focusing on functional dimensions rather than on DSM-categories. The latter approach appears to be the most promising one. Similarities between anxiety disorders seem to be more pronounced than differences. In addition, reported dysfunctional neuroanatomical circuits are implicated in general aspects of human, and even non-human, behavior. First, it is necessary to understand the main processes underlying normal and pathological emotional perception, and fear responses and inhibition. Only then, will it be possible to interpret, sometimes subtle, neurophysiological differences between related disorders.
Using the strategy of 'lumping', an important methodological question remains to be addressed: how to select patients and control subjects independent of the DSM-based criteria? The answer to this question depends on the aspect of the disorder to be investigated. A first step in the dimensional approach is to combine different related disorders as compared with healthy control subjects. Another possibility for subject inclusion, both in patients and in controls, is the use of a cut-off score on a scale developed to measure a specific dimension, for instance state anxiety, compulsiveness, inhibition, uncertainty, disgust, etc. The between-subject variance in the dimensional scores might be used to perform analyses of covariance with the BOLD response of interest. It will be interesting to use this approach to reanalyze the data presented in chapters seven and eight. Instead of grouping patients by diagnosis (OCD, PD and hypochondriasis), the score on the Padua Inventory Revised17;18 or a visual anxiety scale may be used to investigate the role of a possible common feature on the task related BOLD response. Moreover, the score on the Whitely Index19, a measure for obsessionality to diseases, might be used to further investigate similarities and differences in neuronal response to OCD and panic related negative words across the different patient groups.
As proposed by Phillips et al.8, in a model on emotional perception, altered behavioral responses to emotional information might be investigated at roughly three different levels. The first level concerns the identification of the emotional significance of a stimulus. It might be possible that perceptual processing of emotional stimuli is impaired in anxiety disorders due to altered amygdala function. Vuilleumier et al.20 have shown that emotion can directly affect sensory processing at an early stage of perception. If so, anxious people may not only have a different interpretation of what they see, but may also literally perceive these stimuli as abnormal themselves21. The second and third levels of emotional perception relate to the production and regulation of an affective response to this emotional information. When focusing on anxiety disorders, these processes might best be investigated by studies on fear responses, fear extension learning and higher-order modulation of behavior. Whereas almost nothing is known about the role of the DLPFC in the modulation of fear responses, studies in extinction learning, both in animals and in humans, have contributed to the understanding of the role of the ventromedial PFC in the inhibition of learned fear. Direct interconnections between the ventromedial PFC (or subgenual anterior cingulate cortex) and the amygdala have been implicated in fear extinction, which is not simply a process of 'unlearning', but rather one of 'new learning'22-24. Understanding how learned fears are diminished and how extinction learning is changed in patients with anxiety disorders might be an important step in translating neurobiological research to diagnosis and treatment of these patients.
As mentioned above, there is a lack of understanding of the modulation of emotional responses by higher-order cognitive processes. Therefore, probably one of the key issues for the coming years is to integrate research on emotion and cognition. With regard to the complex interactions between emotion and cognition, there appear to be two important questions: 1) which direct and indirect descending connections from the DLPFC to the amygdala play a role in the presumed top-down control of the amygdala response, and 2) in which way can activation of the amygdala directly or indirectly influence executive functioning? More insight in the functional neuroanatomical circuits connecting emotion and cognition, underlying normal and pathological behavior, necessitates a multidisciplinary approach, combining lesion studies, structural measurements, and functional imaging paradigms in human and non-human primates. Useful comparisons between human and non-human experiments require neuropsychological paradigms that are applicable across various species.
Finally, future research in this field may benefit most from a longitudinal and multimodal approach. Probably the most interesting cohort consists of young children, with and without obsessive-compulsive and related disorders. Longitudinal follow-up of these children into adolescence and adulthood enables visualization of natural history as well as the evaluation of long-term effects of environmental influences and treatment strategies. Multimodal designs can also contribute to a better differentiation between cause and effect. For instance, altered morphology may be related to an imbalance of interacting neurotransmitter systems, and differences in task related BOLD responses may be confounded by early maturation deficits. If we succeed in mapping neurophysiological profiles relevant for the symptomatology of patients, these profiles might be used as endophenotypes for subsequent genetic investigations.

Concluding remarks

OCD is an interesting neuropsychiatric disorder, particularly because of the simultaneous presence of cognitive and emotional processes underlying the characteristic, pathological behavior. The specificity of research findings and the sensitivity of experimental paradigms, however, seem to be limited. This may be explained by the congeniality with other neuropsychiatric disorders, and the lack of a strict dividing line between normal and pathological obsessive-compulsive behavior (or between adequate and inadequate fear responses). At presence it is not clear, which research approach will be superior in its attempt to understand the specific underlying mechanisms of the symptomatology. In view of the heterogeneity of the clinical phenomena in OCD, the main challenge for neuroimaging researchers is to define and map the clinically most relevant and specific psychopathological features of the disorder. The findings described in this thesis contribute to the knowledge about the underlying mechanisms of some of the psychopathological processes that are basic to neuropsychiatric diseases.

References
Reference List

1. Saxena S, Brody AL, Schwartz JM, Baxter LR. Neuroimaging and frontal-striatal circuitry in obsessive-compulsive disorder. British Journal of Psychiatry. 1998;173:26-37.
2. Aouizerate B, Guehl D, Cuny E, Rougier A, Bioulac B, Tignol J, Burbaud P. Pathophysiology of obsessive-compulsive disorder. A necessary link between phenomenology. neuropsychology, imagery amd physiology. Progress in Neurobiology. 2004;72:195-221.
3. Gorman JM, Liebowitz MR, Fyer AJ, Stein J. A neuroanatomical hypothesis for panic disorder. American Journal of Psychiatry. 1989;146:148-161.
4. Gorman JM, Kent JM, Sullivan GM, Coplan JD. Neuroanatomical hypothesis of panic disorder, revised. American Journal of Psychiatry. 2000;157:493-505.
5. Andersson JL, Vagnhammar BE, Schneider H. Accurate attenuation correction despite movement during PET imaging. J Nucl Med. 1995;36:670-678.
6. Shallice, T. Specific impaitments of planning. Philos Trans R Soc Lon B Biol Sci 298, 199-209. 1982.

7. Ledoux J. The emotional brain, the mysterious underpinnings of emotional life. first ed. New York: Touchstone; 1996:225-266.
8. Phillips ML, Drevets WC, Rauch SL, Lane R. Neurobiology of emotion perception I: the neural basis of normal emotion perception. Biological Psychiatry. 2003;54:504-514.
9. Barbas H. Connections underlying the synthesis of cognition, memory, and emotion in primate prefrontal cortices. Brain Research Bulletin. 2000;52:319-330.
10. Gilbert AR, Keshavan MS, Birmaher B, Nutche B, Rosenberg DR. Abnormal brain maturational trajectory in pediatric obsessive-compulsive disorder (OCD): a pilot voxel-based morphometry (VBM) study. Clinical EEG and Neuroscience. 2004;35:223.
11. Damasio A. Descarte's error: emotion, reason, an the human brain. New York: Grosset/Putman; 1994.
12. Amaral DG, Price JL. Amygdalo-cortical projections in the monkey (Macaca fascicularis). The Journal of Comparative Neurology. 1984;230:465-496.
13. Ghashghaei HT, Barbas H. Pathways for emotion: interactions of prefrontal and anterior temporal pathways in the amygdala of the rhesus monkey. Neuroscience. 2002;115:1261-1279.
14. Haber SN. The primate basal ganglia: parallel and integrative networks. Journal of Chemical Neuroanatomy. 2003;26:317-330.
15. Schwartz CE, Wright CI, Shin LM, Kagan J, Rauch SL. Inhibited and uninhibited infants "grown up": adult amygdalar response to novelty. Science. 2003;300:1952-1953.
16. Schwartz CE, Rauch SL. Temperament and its implications for neuroimaging of anxiety disorders. CNS Spectrums. 2004;9:284-291.
17. Sanavio E. Obsessions and compulsions: the Padua Inventory. Behav Res Ther. 1988;26:169-177.
18. van Oppen P, Hoekstra RJ, Emmelkamp PMG. The structure of obsessive-compulsive symptoms. Behav Res Ther. 1995;33:15-23.
19. Pilowsky I. Dimensions of hypochondriasis. British Journal of Psychiatry. 1967;113:89-93.
20. Vuilleumier P, Richardson MP, Armony JL, Driver J, Dolan RJ. Distinct influences of amygdala lesion on visual cortical activation during emotional face processing. Nature Neuroscience. 2004;7:1271-1277.
21. Adolphs R. Emotional vision. Nature Neuroscience. 2004;7:1167-1168.
22. Quirk GJ, Russo GK, Barron JL, Lebron K. The role of ventromedial prefrontal cortex in the recovery of extinguished fear. The Journal of Neuroscience. 2000;20:6225-6231.
23. Quirk GJ, Likhtik E, Pelletier JG, Paré D. Stimulation of medial prefrontal cortex decreases the responsiveness of central amygdala output neurons. The Journal of Neuroscience. 2003;23:8800-8807.
24. Phelps EA, Delgado MR, Nearing KI, LeDoux JE. Extinction learning in humans: role of the amygdala and vmPFC. Neuron. 2004;43:897-905.