Concept of functional imaging of memory decline in Alzheimer’s disease

Available online at www.sciencedirect.com

Methods 44 (2008) 304–314 www.elsevier.com/locate/ymeth

Concept of functional imaging of memory decline in Alzheimer’s disease q Alexander Drzezga

*

Department of Nuclear Medicine, Klinikum rechts der Isar, Technische Universita¨t Mu¨nchen, Ismaninger Str. 22, D-81675 Mu¨nchen/Munich, Germany Accepted 13 February 2007

Abstract Functional imaging methods such as Positron Emission Tomography (PET) and functional Magnetic Resonance Imaging (fMRI) have contributed inestimably to the understanding of physiological cognitive processes in the brain in the recent decades. These techniques for the first time allowed the in vivo assessment of different features of brain function in the living human subject. It was therefore obvious to apply these methods to evaluate pathomechanisms of cognitive dysfunction in disorders such as Alzheimer’s disease (AD) as well. One of the most dominant symptoms of AD is the impairment of memory. In this context, the term ‘‘memory’’ represents a simplification and summarizes a set of complex cognitive functions associated with encoding and retrieval of different types of information. A number of imaging studies assessed the functional changes of neuronal activity in the brain at rest and also during performance of cognitive work, with regard to specific characteristics of memory decline in AD. In the current article, basic principles of common functional imaging procedures will be explained and it will be discussed how they can be reasonably applied for the assessment of memory decline in AD. Furthermore, it will be illustrated how these imaging procedures have been employed to improve early and specific diagnosis of the disease, to understand specific pathomechanisms of memory dysfunction and associated compensatory mechanisms, and to draw reverse conclusions on physiological function of memory. Ó 2007 Elsevier Inc. All rights reserved. Keywords: Alzheimer’s disease; MCI; Dementia; PET; fMRI; Activation studies; Memory decline; Functional imaging

1. Introduction The neurodegenerative disorder Alzheimer’s disease (AD) represents the most frequent cause for the development of dementia, which is characterized by an insidious onset and a progressive impairment of cognitive functions, finally leading to an impairment in the capabilities of daily living [1,2]. In the public perception, AD is strongly associated with memory impairment. Although the equation of AD with memory impairment is simplifying the complexity

q Function (in philosophy): ‘‘A normative relation of objects or events to their use or consequences’’ (Function. 2007, January 17. In Wikipedia, The Free Encyclopedia. Retrieved 18:32, January 29, 2007, from http:// en.wikipedia.org/w/index.php?title=Function&oldid=101298822). * Fax: +49 89 4140 4841. E-mail address: [email protected]

1046-2023/$ - see front matter Ó 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ymeth.2007.02.002

and diversity of cognitive problems conditioned by the disorder, memory problems do actually represent one of the earliest symptoms of AD and one of the most serious and subjectively affecting cognitive deficits for the individual. The term ‘‘memory’’ itself represents a simplified summarization of an entire set of different associated functions, such as short-term, long-term, procedural, declarative, semantic or episodic memory. Furthermore, memory can be subdivided in functions occupied with encoding and with retrieval of information. The hallmark symptom of AD is an impairment of the declarative memory. The term ‘‘declarative memory’’ describes the aspect of human memory that stores facts and experiences, which can be explicitly discussed, or declared by the individual. The declarative memory is subdivided in the semantic memory (non-context specific fact, word and object memory) and the episodic memory (memory of events, including time,

A. Drzezga / Methods 44 (2008) 304–314

place and associated emotions). In AD, the episodic memory is among the earliest affected functions [3]. Typically, AD results in a deficit in the establishment of new episodic memories, whereas events dating from more remote periods in the past are better preserved. In later stages of dementia of the Alzheimer type, most other memory domains are hampered as well. For this reason, memory impairment is also a fundamental criterion of the diagnosis of AD, which today is mostly based on neuropsychological evaluation [2]. A definite diagnosis of Alzheimer’s disease, however, is currently only possible by post mortem histopathological analysis of the brain. In contrast to other disease entities, the in vivo extraction of tissue for histopathology (e.g. by biopsy) can hardly be reasonably justified for diagnosis of dementia. Besides this limitation regarding the definite diagnosis of AD by neuropsychology, it is well accepted that the pathology of AD starts years to decades before onset of cognitive symptoms, which limits the value of clinical examination also for early detection of the disease [4]. Following a period free of symptoms, it is assumed that patients go through a stage of mild cognitive impairment (MCI) for 5–10 years (often particularly in the memory domain), before clinically manifest dementia can be diagnosed [5–7]. Thus, the MCI group is regarded a risk population for AD, important for scientific exploration of the disease in early stages. However, the diagnosis ‘‘MCI’’ can also be based on other factors (vascular pathology, depression, etc.) and not all MCI-patients will develop Alzheimer dementia. Clinical/neuropsychological examination does not permit a reliable identification of the patients with early Alzheimer’s disease in this group [5,8]. In summary, the early and definite diagnosis of Alzheimer’s disease is hampered by the low sensitivity and specificity of clinical/neuropsychological evaluation and the limited accessibility of brain tissue for histopathological analysis. Furthermore, the specific brain pathology underlying selected cognitive deficits can not be uncovered in vivo without the help of suitable diagnostic tools. Consequently, the search for specific surrogate markers of the disease and for tools for the non-invasive assessment of regional brain pathology in vivo has been intense. Regarding the fact that imaging procedures have been successfully applied for the scientific evaluation of cognitive functions such as memory in vivo, it suggests itself to direct efforts towards the application of different neuroimaging modalities for the improvement of early and specific diagnosis of Alzheimer’s disease in the stage of MCI and for a better understanding of ongoing pathomechanisms of the disease. The general purpose of the assessment of cognitive deficits in AD with imaging tools has been manifold: (1) to improve early diagnosis of the disease in the stage of MCI or even before measurable cognitive symptoms arise, (2) to allow a specific differential diagnosis of the type of disorder, underlying cognitive symptoms, (3) to assess the association of particular symptoms such as memory impairment with type and localization of underlying

305

pathology and in such, to improve understanding of specific pathomechanisms, (4) to learn about the physiologic role particular brain regions play in healthy cognitive function, using AD as a ‘‘lesion model’’. It has been shown that modern imaging technologies, such as Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET), allow the detection of structural, functional and molecular pathological changes associated with the disease progression and are superior to neuropsychological testing regarding early diagnosis of Alzheimer’s disease [9,10]. In addition to the improvement of diagnosis, functional imaging procedures have enabled a better correlation of cognitive functional deficits with particular affected brain regions. Due to the sheer abundance of literature on this matter, the current article will exclusively focus on methods of imaging that allow the visualization of functional changes of neural activity associated with memory decline in AD. Other imaging procedures (MRI volumetry, voxel-based morphometry, etc.) directed to the detection of structural abnormalities (such as hippocampal atrophy), imaging studies of neuronal transmitter and receptor status and modern molecular imaging procedures, which allow detection of specific pathology (amyloid plaque imaging) appearing in the course of AD cannot be discussed here. 2. Resting brain metabolism and memory impairment With the radiolabeled glucose analog [18F]Fluorodeoxyglucose (FDG), a PET-tracer for in vivo assessment of regional cerebral glucose metabolism is at hand. The uptake of [18F]FDG parallels the transport of glucose into cells, subsequently the tracer is phosphorylized and trapped in the cell and thus allows regional assessment of regional cerebral glucose metabolism (rCGMglc) [11]. It is well known that glucose constitutes the relevant source of energy for the brain and that glucose metabolism is tightly coupled to regional neuronal function. Recently, it has been shown that this coupling may be mediated by glial cells. Apparently, energy-demanding synaptic activity of a neuron leads to increased glucose metabolism in surrounding glial cells, which subsequently transfer lactate as an energy carrier to the neuron [12]. If the [18F]FDG PETexamination is performed in resting conditions, the tracer uptake is mainly driven by basal neuronal activity, thus representing a measure for general neuronal integrity [13]. Inversely, impairment of neuronal function leads to a decrease of regional glucose turnover. In patients with AD, characteristic deficits of regional glucose metabolism in affected brain regions have often been demonstrated. Typically, temporoparietal cortex, posterior cingulate cortex and frontal cortex show hypometabolic changes, whereas regions such as the sensorimotor cortex and the primary visual cortex are spared (Fig. 1) [14–16]. These metabolic deficits have been assigned to neurodegenerative processes such as loss of synapses, neuronal death, but also to an impaired metabolic efficiency [17]. The resulting pat-

306

A. Drzezga / Methods 44 (2008) 304–314

Fig. 1. Typical [18F]FDG PET-findings in healthy subjects, patients with mild cognitive impairment (MCI) and patients with Alzheimer’s disease. Surface projections of cerebral glucose metabolism in both hemispheres (Red/yellow colour: normally high cerebral glucose metabolism, green colour: abnormally low metabolism). (For interpretation of the references in colour in this figure legend, the reader is referred to the web version of this article.)

tern of metabolic changes has a high specificity for AD and, thus, [18F]FDG PET represents a valuable tool for differential diagnosis [15,16,18–20]. Furthermore, a high sensitivity of [18F]FDG PET for the diagnosis of dementia in very early stages such as in MCI has often been demonstrated [21–23]. Patients with MCI who show metabolic abnormalities in specific cortical region such as posterior cingulate cortex accompanied by hypometabolism in parietotemporal cortex have a high risk for the development of AD. Finally, [18F]FDG PET can be used for objective follow-up of disease progression and, thus, as a potential marker for therapy control [24]. Deficits in a cognitive domain are linked to metabolic abnormalities in the corresponding brain region, and the pattern of metabolic abnormalities in AD tightly reflects the typical functional loss in this disease. E.g. AD-typical cognitive deficits in memory or spatial thinking are mirrored in the typical pattern of hypometabolism in brain regions usually linked with these functions such as the temporal cortex and the parietal cortex, respectively. In contrast, functions which are preserved in AD such as movement and sensory abilities or visual perception are mirrored in the usually relatively less impaired glucose metabolism in sensorimotor and primary visual cortical regions. However, this association of cognitive deficits with regional brain pathology is limited for several reasons: (1) Cognitive functions are usually not sustained by a single cortical region but by a complex network. Different regional brain defects may result in similar cognitive deficits, and the association of regional brain pathology with a specific cognitive problem may not allow the reverse conclusion that this brain region is specifically responsible for the particular function. (2) Ongoing neurodegeneration in AD is not restricted to a certain cerebral region but

involves extended portions of the brain. Likewise, ADpatients usually show more than one cognitive problem. Thus, it may be difficult to identify the particular regional defect underlying a single cognitive deficit. For these reasons, studies correlating metabolic deficits of the brain with particular neuropsychological test scores may sometimes be more useful to identify the validity of the test than to identify the true connection of a functional deficit and regional brain pathology. Consequently, it may be necessary to compare the results from functional imaging studies in patients with cognitive impairment with data obtained in healthy subjects to draw valid conclusions on pathophysiology and physiological function. In this context, many studies regarding functional neuroanatomy of memory functions have been performed previously in healthy subjects, and different theoretical models of the memory exist. These models and the results of the associated studies can not be discussed here in detail; generally it has been shown to be very complex to assign particular brain regions to a function as complex and diverse as ‘‘memory’’. There is great consent, however, that many memory functions appear to be tightly linked to temporal regions of the brain. Particularly, a system in the medial temporal lobes consisting of anatomically related regions (the hippocampal region and the adjacent perirhinal, entorhinal and parahippocampal cortices) has been shown to be involved in declarative memory functions [25,26]. Some evidence has been collected in animal and functional imaging studies that episodic memory, which is particularly affected in AD, strongly relies on the hippocampus itself. Studies were able to show that the hippocampus is crucially involved in encoding and retrieving of episodic memory [27,28] and of autobiographical information in general [29]. In addition to the hippocampus, other structures of the brain have been identified to contribute to episodic/ autobiographic memory, i.e. the prefrontal cortical regions, the temporoparietal junction, the posterior cingulate cortex and also the cerebellum [28,30]. Generally, regions on the left hemisphere appear to be stronger involved in encoding (particularly the hippocampus), whereas the right hemisphere seems to be more strongly involved in retrieval (particularly the prefrontal cortex) of episodic memory according to the so-called HERA (hemispheric encoding and retrieval asymmetry) concept [31]. This neuroanatomical concept is well supported by structural imaging studies in patients with early AD, showing dominant cortical atrophy in the hippocampal cortex [32–34], corresponding to their episodic memory deficits, followed by extension of atrophy to neocortical regions with progression of the disease. The results of several functional imaging studies using [18F]FDG PET in AD also fit very well into the described neuroanatomical concept of episodic memory. Eustache et al. examined the correlation of autobiographical memory from differently remote periods in AD-patients with cerebral metabolism [35]. They found a significant correlation of the integrity of autobiographical memory of the last

A. Drzezga / Methods 44 (2008) 304–314

5 years with glucose consumption in right hippocampal, middle and inferior frontal and middle temporal gyrus. Intactness of memory of middle age was associated with glucose metabolic rate in bilateral prefrontal cortex and memory of the childhood with left middle frontal glucose metabolism. In a recent study, the performance of ADpatients in a standard neuropsychological test battery (CERAD: Consortium to Establish a Registry for Alzheimer’s Disease [36]) on episodic memory, verbal fluency and naming was correlated with [18F]FDG PET [37]. In this study, a strong correlation of test performance with left-hemispheric temporal and prefrontal metabolic rate was demonstrated in AD-patients [37]. These results imply that brain regions associated with the tested cognitive functions, indeed show a selective metabolic impairment in association with the observed cognitive deficits. Furthermore, these data demonstrate a susceptibility of the CERAD test battery to left-hemispheric brain lesions. Desgranges et al. showed in study a correlation of verbal episodic memory function in another AD with rCGMglc in predominantly left-hemispheric limbic structures, such as mesial temporal cortex, thalamus and cingulate cortex [27]. In addition to these findings, which agree well with results on physiological functional neuroanatomy of episodic memory, the authors found a correlation with right parietotemporal and frontal association cortical glucose metabolism. As these cortical regions would not be expected to participate in the examined cognitive functions in healthy subjects, Desgranges et al. interpreted these right-hemispheric correlations in the sense of potential compensatory effects. In another, more recent study, Desgranges et al. evaluated the neural substrate of episodic memory impairment in AD in relation to the degree of deterioration. Whereas the impairment of episodic memory correlated with the hypometabolism in hippocampal and retrosplenial cortex in the less severe group (as expected), in the more severe subgroup correlations with mainly the left temporal neocortex were observed. Again they interpreted the involvement of these neocortical areas, which are usually involved in semantic rather than episodic memory, as a form of inadequate compensatory mechanism [38]. Semantic memory (non-context specific fact-, word- and object-memory), the second subcategory of the declarative memory functions, is also impaired in later stages of AD and in Semantic Dementia (SD), a different type of disease, belonging to the frontotemporal lobar degenerative disorders. The agreement between data on functional neuroanatomy of semantic memory and results of imaging studies on affected brain structures in patients with semantic memory deficits appears less concordant than for episodic memory. Functional activation studies on semantic memory identified an extended associated network, including prefrontal and caudo-lateral temporal areas [39]. Studies on brain pathology in SD, however, predominantly demonstrate a very concise affection of the rostral temporal lobes. For example, several studies on structural

307

imaging (MRI) mainly demonstrated rostral temporal lobe atrophy in SD [40–43]. Correspondingly, studies of cerebral glucose metabolism in SD-patients report rostral temporal hypometabolism [44–47]. A recent comparative study by Nestor et al. of [18F]FDG PET-findings in AD- and SDpatients showed predominant rostral hypometabolism exclusively in SD-patients, thus demonstrating that a functional impairment of this part of the temporal cortex may result in semantic memory deficits [46]. Interestingly, in this study, comparable hypometabolism was detected in medial temporal lobe (MTL) regions in AD- and SD-patients, although semantic memory impairment was only observed in SD and episodic memory impairment only in AD. Thus, the authors concluded that the preserved episodic memory function in SD compared to AD cannot be explained by less affection of the MTL in SD, as previously suggested. Reversely, the authors argued that the exclusive episodic memory deficits in AD in their study cannot be explained by MTL abnormalities alone, but rather appear to be induced by dysfunction of the MTL accompanied by dysfunction of additional components of a diencephalic network (mamillary bodies, dorsomesial thalamus and posterior cingulate). The resolution of [18F]FDG PETimaging may however not be sufficient to reliably differentiate between anatomical MTL-substructures (entorhinal, perirhinal, parahippocampal, hippocampal cortex) and, thus, a potentially different affection of these substructures in the examined AD- and SD-groups can not be excluded. These results further underline that the association of a cerebral lesion with a cognitive deficit does not necessarily allow the reverse conclusion that the affected brain region is exclusively responsible for the impaired cognitive function under physiological conditions. Particularly, it should be kept in mind that different brain pathologies may result in similar cognitive deficits. Several studies have also been performed to identify specific correlations of cerebral metabolic deficits with semantic memory dysfunction in ADpatients. Some of these studies demonstrated correlations with regional abnormalities similar to Semantic Dementia. E.g. Hirono et al. were able to show a correlation of cerebral glucose metabolic rate in left inferior temporal gyrus with performance in three verbal semantic memory tests [48]. Other studies showed correlations of semantic memory problems in AD with more extended metabolic abnormalities. E.g. Zahn et al., for example, showed a correlation of left-hemispheric hypometabolism (anterior temporal, posterior inferior temporal, inferior parietal and medial occipital) with semantic impairments in AD-patients. Desgranges et al. found a correlation of semantic memory scores with glucose metabolism in a set of cortical regions similar to the physiological cerebral network associated with semantic memory function (temporoparietal and frontal association cortices of the left hemisphere). Correlation analyses of brain metabolism and cognitive function have not only been performed in advanced stages of neurodegenerative disorders. First studies were able to correlate cognitive deficits with regional cerebral hypome-

308

A. Drzezga / Methods 44 (2008) 304–314

tabolism in MCI-patients as well. Perneczky et al. showed significant correlation of posterior cingulate hypometabolism with the degree of cognitive impairment in the Clinical Dementia Rating scale (CDR), a neuropsychological test including recent and long-term memory assessment [49]. This agrees well with the known fact that posterior cingulate hypometabolism is one of the earliest clues, predicting the development of AD-type dementia in MCI. Explicit correlation of hippocampal glucose metabolism with neuropsychological performance is aggravated by the limited spatial resolution of PET regarding the medial temporal cortex, thus data thereto are limited. However, using specific image processing technology, a high predictive value has been attributed to hippocampal hypometabolism in MCI as well [50]. Chetelat et al. performed high-resolution T1-weighted volume MRI and resting-state [18F]FDG PET-scans in a group of MCI-patients with isolated memory impairment [51]. They assessed episodic memory encoding and retrieval performance in these patients and calculated positive correlations between memory scores and grey matter density and partial-volume effect corrected brain glucose utilization (rCMRglc). Both encoding deficits and retrieval deficits were correlated with declines in hippocampal region grey matter density. However, the hippocampal rCMRglc only correlated with the encoding subtest, whereas the retrieval subtest correlated with the posterior cingulate cortical rCMRglc, exclusively. These results nicely underline the crucial role of the hippocampus in memory encoding. Furthermore, they illustrate a divergence of structural and functional changes associated with memory impairment. The selective correlation of memory retrieval with hippocampal atrophy and posterior cingulate rCMRglc decrease (without reference to local atrophy in this region) may point to a remote effect of hippocampal atrophy, i.e. through functional disruption, on posterior cingulate synaptic function. Some studies were able to establish an association between changes in memory function and cerebral metabolism even in healthy subjects at risk for AD. Ercoli et al. performed a Memory Functioning Questionnaire and [18F]FDG PET studies in healthy subjects with subjective memory complaints [52]. They subdivided the group in carriers and non-carriers of the Apolipoprotein (APOE) e4-allele, a genetic risk factor for AD. Interestingly, selfreported frequency of applied mnemonic strategies significantly correlated with metabolic decline in the temporal regions in APOEe4 carriers but not in non-carriers. The authors concluded that this use of compensatory strategies, as indicated by more frequent mnemonics use in APOEe4-carriers, may reflect underlying metabolic changes in the brain regions associated with prodromal Alzheimer disease. Another [18F]FDG PET study by De Leon et al. was able to demonstrate that baseline metabolic reductions in the entorhinal cortex of healthy subjects accurately predicted the conversion from normal to MCI within a 3-year follow-up period. In these subjects who declined, the baseline entorhinal metabolism pre-

dicted longitudinal memory and temporal neocortical metabolic reductions [53]. SPECT-tracers for imaging of cerebral perfusion such as Tc99m-ECD or Tc99m-HMPAO generally show very similar characteristics as [18F]FDG PET regarding imaging in AD-patients. This is due to the fact that patterns of regional metabolism are predominantly highly comparable to patterns of regional brain perfusion at rest. Consequently, a decrease of cerebral perfusion can be demonstrated in corresponding regions of the brain in AD as shown for hypometabolism with [18F]FDG PET. However, spatial resolution of SPECT in general and therefore sensitivity for early detection of the disease is regarded inferior to [18F]FDG PET [54]. Nevertheless, in some studies similar associations of the degree of memory impairment with cerebral perfusion defects as measured with SPECT have been shown in analogy to the PET-findings mentioned above. For example, Nobilli et al. were able to demonstrate a positive correlation of performance in a memory test (SRT: selective reminding test) with cerebral perfusion in left-hemispheric post-central, parietal/precuneus, inferior parietal and middle temporal cortex on the left hemisphere as well as with right-hemispheric middle temporal cortex [55]. 3. The concept of cognitive reserve It has been shown that the same degree of measurable AD-pathology does not necessarily result in an inter-individually comparable degree of cognitive impairment. In up to 25% of individuals who fulfill neuropathological criteria for Alzheimer’s disease (AD) at autopsy, no cognitive impairment has been obvious during their lives [56]. Also, it has been shown that education modifies the relation of AD-pathology to the level of cognitive function and that individuals with higher education show more advanced AD-pathology compared to less educated subjects at comparable clinical levels of impairment [57,58]. To explain this phenomenon, the hypothesis of a ‘‘cognitive reserve’’ has been proposed [57,59,60]. This hypothesis suggests that certain factors must exist, leading to a delay of the onset of AD, respectively to a reduced clinical manifestation of the pathological processes of AD in some individuals. Higher education and a more cognitively stimulating lifestyle in general have been shown to contribute to this effect [61]. However, until today, the neurobiological substrate of this cognitive reserve is still unknown. Possibly, a higher synaptic density, a greater number of neurons or a greater brain size might provide a structural basis of this phenomenon. To examine the functional correlates of cognitive reserve, a small number of imaging studies have been performed in patients with AD. A recent study by Perneczky et al. in AD-patients was able to demonstrate stronger metabolic deficits in the posterior parietal cortex in patients with higher level of education as compared to less educated patients with a comparable degree of cognitive impairment (Fig. 2) [62]. This study confirms the findings from neuro-

A. Drzezga / Methods 44 (2008) 304–314

Fig. 2. Negative linear regression between cerebral metabolism and level of education in AD-patients with equal degrees of cognitive impairment (SPM2-analysis, p < 0.001 uncorrected, red colour: lower cerebral metabolism in patients with higher education). See also: Perneczky et al. [62]. (For interpretation of the references in colour in this figure legend, the reader is referred to the web version of this article.)

pathology and provides in vivo evidence for an effect of cognitive reserve in the patients with higher educational level. Meanwhile, similar effects have been shown for other types of dementia as well, and, thus, should be taken into account whenever a correlation of metabolic brain deficits with the level of cognitive impairment is performed [56]. 4. Activation studies in AD and MCI In addition to the imaging techniques mentioned above, which allow the assessment of neuronal baseline function at rest, different methods of functional imaging may be used to identify brain structures that are involved in the performance of specific cognitive tasks. These ‘‘activation studies’’ require the design of a suitable control condition and of particular paradigms which lead to a precise activation of the involved brain regions. The most established methods to assess cerebral activation include [15O]water PET and functional MRI (fMRI). The principle of these methods is based on the fact that an increase of synaptic activity leads to a corresponding increase of perfusion in the brain region involved. The mentioned imaging procedures allow the measurement of this change in cerebral perfusion, and thus enable conclusions on regional brain activation. With [15O]water PET regional blood-flow can be assessed by measurement of radiolabeled tracer-delivery to the concerned brain regions, whereas with fMRI bloodflow changes can be registered by assessment of the level of oxygenated blood, which rises in better perfused brain regions. Using these procedures, various alterations of cerebral activation patterns were already demonstrated during different types of cognitive work in neurodegenerative disorders like AD, as compared to healthy controls. In general, the classification of this type of changes is not trivial. In part, they have been interpreted in the sense of compensatory reorganization processes (functional reserve capacity) and in part in the sense of unspecific decompensation of functional systems [63]. Due to these assumptions, it may be complex and possibly misleading to directly link changes in the cerebral activation pattern to specific cognitive deficits such as memory impairment. Pruvlovic et al. summarized the basic hypotheses about the

309

mechanisms underlying changed activation patterns in neurodegenerative disorders in a comprehensive model [63] (Fig. 3). In short, this model implies that, in a slightly damaged system, stronger cerebral activation and possibly activation of auxiliary units (outside of the original network) is required even at lower task-difficulty levels, to yield the same performance as a healthy system. In contrast, in a system with stronger damage no comparable activation levels can be achieved, which results in a measurable decrease of task performance. In context with this hypothesis, a lower task performance in AD appears to be accompanied by significant activation deficits in related brain structures in many studies [63–72]. Several studies were able to demonstrate a reduced activation in AD-patients particularly during performance of memory tasks. Studies on retrieval of episodic memory showed less activation in hippocampal and parietal cortex [73–75] in AD-patients. Regarding encoding of memory, many studies were able to demonstrate significantly reduced medial temporal activation in AD-patients as compared to healthy subjects [66,74,76]. Furthermore, in many of these studies, reduced activation in temporal neocortex and frontal and cortical regions has been found during encoding of visual and verbal information in AD [66,71,76]. Complementary to these examples of reduced activation in severely damaged brain regions in AD, the model by Prvulovic predicts stronger activation of task-relevant brain structures, in a stage of limited damage, representing a compensatory effort. Indeed, in the group of MCI, which contains AD-patients in less progressive stages, stronger activation of memory-associated brain regions has been demonstrated during performance of memory tasks. For example, Dickerson et al. showed a hyperactivation of the medial temporal cortex during an episodic memory encoding task [77]. Strikingly, in this study the extent of temporal cortical activation correlated positively with the memory performance, but on the other hand predicted further cognitive decline. Also Yetkin et al. demonstrated increased extent of activation in MCI during a working memory task [78]. Other studies in MCI-subjects, however, demonstrated reduced cerebral activation during memory tasks. Machulda et al. showed decreased medial temporal activation during encoding in MCI similar as in AD, Johnson et al. observed less activity in posterior cingulate cortex during recognition of previously learned items and less right hippocampal activation during encoding of novel items [66,79]. Small et al. found a pattern of dysfunction similar to that found in elders with AD, involving all hippocampal regions, in a subgroup of elderly with isolated memory decline. They suggested that this may represent ongoing AD-pathology in these subjects [76]. Although these findings are considerably controversial, they may still fit into the basic activation model. Dickerson et al. demonstrated that greater hippocampal activation

310

A. Drzezga / Methods 44 (2008) 304–314

Fig. 3. Integrative activation model (adapted from Prvulovic et al. [63]. PU, processing unit; MOD, modulation unit; AUX, auxiliary unit.

was present during face-name associative memory encoding in MCI, whereas hippocampal hypoactivation was observed during performance of the same task in manifest AD [80]. They discussed that a phase of increased medial temporal activation in an early prodromal stage may be followed by a decrease of activation in later stages of AD. Similar effects may not only emerge between MCI and AD but also within the MCI category and even in healthy subjects at risk. This hypothesis is nicely confirmed by a study of Celone et al. who showed paradoxical hyperactivation of the hippocampus in less impaired MCI, whereas significant hypoactivation was present in more impaired MCI-subjects during performance of an associative memory paradigm [81]. In general, dysfunctional hypoactivation and compensatory hyperactivation may be present at the same time in a brain affected by neurodegeneration, depending on the stage of disease. Recently a number of similar studies have been performed in healthy subjects at risk for AD such as carriers of the APOEe4-genotype or subjects with a familiar risk for the disorder. Again, several studies showed reduced activation in memory-associated brain systems during encoding and recall of information in subjects at risk of AD and interpreted these findings as potential early neurodegeneration [82–85]. Lind et al. demonstrated reduced brain activity in APOEe4-carriers in regions typically

affected by AD during semantic categorization tasks [84]. Trivedi et al. found reduced activation in medial temporal lobe during encoding episodic memory in a risk population [85]. Elgh et al. showed lower prefrontal activity in highrisk subject during episodic encoding and retrieval [83]. In contrast to these studies, a large number of studies performed in healthy subjects at risk demonstrated greater magnitude and extent of activation during memory-associated cognitive work. Bookheimer et al. found that the magnitude and the extent of brain activation during a memory-activation task (learning and recalling unrelated word pairs) were greater among the carriers of the APOE epsilon4 in regions affected by Alzheimer’s disease, including the left hippocampal, parietal, and prefrontal regions [86]. This may point to already ongoing neurodegeneration and resulting compensatory increased cognitive effort in these subjects. Indeed the detected brain activation differences in this study correlated with the degree of memory decline, retested two years later. Many other studies were able to show similar effects in subjects at risk of AD [3,82,87–90]. Bondi et al. showed stronger brain activation during learning of new pictures in APOEe4-carriers, Bassett et al. demonstrated more intense and extensive frontal and temporal/hippocampal activations during a paired associates memory task in subjects with familiar risk for AD [82,87].

A. Drzezga / Methods 44 (2008) 304–314

Neurodegenerative processes do not evenly affect the entire brain in AD. For this reason, even in patients with clinically manifest Alzheimer’s disease, it can be expected that some less affected cerebral regions will maintain compensatory functions. Furthermore, it has been discussed that patients with neuronal damage, affecting the activation of a regional brain structure, may recruit additional ‘‘auxiliary’’ brain structures to solve a task, representing a potential heterotopic compensation effect [91]. Indeed, studies were able to demonstrate stronger activation and additional activation foci in AD in brain regions which are not activated in healthy subjects during task performance. Yetkin et al. showed increased extent of activation and recruitment of additional brain regions in AD during a working memory task [78]. Several studies found increased activation of left prefrontal cortex in presence of reduced hippocampal activation during a memory encoding and retrieval task [73,74]. During retrieval of short-term verbal memory, Becker et al. demonstrated in AD-patients a higher activation of regions of cerebral cortex normally involved in auditory-verbal memory, as well as activation of cortical areas not activated by normal elderly subjects [92]. Woodard et al. were able to demonstrate that this type of over-activation correlated with preserved task performance in AD-patients [93]. Grady et al. demonstrated increased activity in the right prefrontal, anterior cingulate and left amygdale during a visual working memory task in AD-patients [94]. A correlation of activation increase in the amygdala with better performance was observed in AD, whereas, in healthy subjects, performance showed a correlation with prefrontal cortical activation, indicating a shift in the relevance of brain structures for task performance. Also in this study, a connectivity analysis revealed a functional disconnection between the prefrontal cortex and the hippocampus in Alzheimer’s disease. In another study, Grady et al. observed activation of a unique network of bilateral dorsolateral prefrontal and posterior cortices in AD-patients during semantic and episodic memory tasks. They concluded that AD-patients can use additional neuronal resources for compensation [91]. Regarding this type of studies, it becomes clear that no linear relation between the cognitive deficits and the changes of cerebral activation can be established, regarding that compensatory over-activation in some brain regions and breakdown of systems in other regions may be present simultaneously during the course of ongoing neurodegeneration. Furthermore, it has been demonstrated that activation patterns during cognitive work may be influenced by cognitive reserve effects, similar as shown for functional imaging studies at rest [95]. Another limitation of activation studies in general and of memory studies in particular may be the fact that the extent of brain activation is dependant on the individually different task-difficulty level [96]. It may be particularly hard to objectively quantify this difficulty level in AD-patients, regarding that a higher difficulty may result in a stronger cognitive effort, but not necessarily in an impaired performance.

311

Thus, the interpretation of activation patterns in ADpatients has to be carried out utterly careful. Nevertheless, the detection of such specific functional reorganisation effects in persons at risk could lead to the diagnosis of ongoing neurodegeneration in an early stage (even in presence of still uncompromised cognitive performance) [86,97]. Furthermore, it has been shown that effects of therapeutic interventions on cognitive performance are mirrored in changes of the activation patterns. Thus, activation studies may help to understand and monitor treatment effects [98]. 5. Functional deactivation and the principle of the defaultmode network Besides the necessary activation of task-related brain regions, the ability to specifically inhibit or even deactivate brain regions unrequired for the task has been established in healthy subjects. E.g. it has been shown that the visual cortex is deactivated during tasks that put high demand on the auditory system. This cross-modal inhibition probably facilitates the focussing on the required cerebral functions [99,100]. In patients with MCI and AD not only altered activation patterns were demonstrated but also impaired deactivation of non-required brain regions during cognitive work [64,101]. This leads to the hypothesis that, in neurodegeneration, necessary inhibitory processes may be impaired. Such inhibitory deficits may account for the decreased ability of AD-patients to focus their attention on relevant modalities and for their general vulnerability to distraction. Closely related to this aspect, measurement of the cerebral ‘‘default-mode’’ activity or ‘‘resting-state’’ activity has moved into the centre of interest recently [102]. Modern statistical analytic procedures such as independent component analysis (ICA) allow the detection of the functional connectivity of certain brain regions by identification of synchronous low-frequency fluctuations in the resting fMRI-signal. The default-mode network comprises these brain regions which are coherently activated in awake resting state, without performance of a dedicated cognitive task. Different functions of perception of the environment and the own person, planning of future actions and particularly memory processes have been assigned to the default-mode network [103,104]. As soon as brain activity is directed towards a specific cognitive task, the default-mode network in total or at least its unrequired parts are being deactivated. This effect has been interpreted as a focussing of the available capacities on the task-relevant brain regions. In contrast to [18F]FDG PET, the default-mode network does not represent a measure for the general basal neuronal activity but for synchronously interacting neurons in the resting state. Anatomical regions which have been identified to be participating in the default-mode network are the posterior cingulate cortex and parietal and medial prefrontal cortical areas. Furthermore, prominent coactivation of the

312

A. Drzezga / Methods 44 (2008) 304–314

hippocampus has been detected, suggesting that the default-mode network is closely involved with episodic memory processing [105]. Recently, the striking topographical similarity of the default-activity pattern in healthy subjects with the defects in glucose metabolism and the patterns of amyloid plaque deposition and atrophy in AD has been described. Furthermore, a strong concordance of the default-activity pattern with the cerebral activation pattern associated with successful memory retrieval in healthy subjects has been noticed [104], underlining the close relation of default-activity with memory functions. It has been speculated that lifetime regional cerebral default activity may predispose cortical regions to AD-related changes, including amyloid deposition, metabolic disruption and atrophy [101,104]. In return, the induced pathology in the brain regions associated with the default-mode would then lead to a functional impairment, particularly concerning memory function. First studies on default activity in AD-patients were already performed. Greicius et al. were able to demonstrate decreased resting-state activity in memory-associated brain regions, namely the posterior cingulate and hippocampus. They suggested that disrupted connectivity between these two regions may partially account for the posterior cingulate hypometabolism commonly detected in Positron Emission Tomography studies of early AD [105]. Furthermore, an impaired deactivation of the default-mode network in MCI and AD during task-directed cognitive work has been shown [103,106]. 6. Summary A set of different imaging methods is at hand today to study functional processes in the brain at rest and during cognitive work. In patients with Alzheimer’s disease, these techniques proved to be highly valuable to assess diverse mechanisms of cerebral functional pathology accompanying the decline of cognitive functions, such as memory impairment, in the course of this neurodegenerative disorder. Regarding the improvement of early and specific diagnosis of the disease, some of these imaging methods have already reached relevance in clinical routine. For example, the assessment of the characteristic regional deficits of cerebral glucose metabolism with [18F]FDG PET as a measure for reduced neuronal baseline function has frequently been demonstrated to be highly valuable for early and reliable diagnosis of AD, even in the stage of MCI. Other detected abnormalities, such as changes in the pattern of cerebral activation during active cognitive work, impaired deactivation of task-irrelevant modalities or affection of the cerebral default-activity network have not yet reached clinical application, but offer valuable insights in the complexity of functions involved in cognitive processing and the multiple deleterious effects of the disease. Many of these imaging studies were able to confirm current hypotheses on functional anatomy of memory functions using AD as a ‘‘lesion model’’. However,

it has been shown to be considerably complex to establish a specific causal association of cognitive deficits such as the AD-related memory decline with underlying regional brain pathology. A linear correlation between cerebral pathology and resulting cognitive impairment can hardly be established for several reasons. First, in the process of neurodegeneration, regional cerebral pathology may induce local and remote compensatory effects before reaching a stage of decompensated dysfunction. These effects of compensation and decompensation may even be present at the same time in different regions of the brain and, thus, cannot always be easily differentiated. This may be the reason for the confusing fact that imaging studies report more extended activation patterns as well as reduced activation during performance of cognitive tasks in AD-patients and risk populations, compared to healthy controls. Second, it has been shown that the degree of measurable AD-pathology does not necessarily result in an inter-individually comparable degree of cognitive impairment. A number of confounding factors, such as e.g. the level of education, are apparently modulating this causal interrelation and have been summarized under the concept of ‘‘cognitive reserve’’. Although these limitations hamper the direct correlation of detected brain abnormalities with cognitive decline, the methods of functional imaging may actually open the unique opportunity to study exactly these mechanisms, which influence the expression of cognitive symptoms in AD. This may be utterly important to understand defense strategies of the brain against the ongoing neurodegeneration in AD. In summary, functional imaging may offer the chance to objectively assess the extent of pathology in the brain in AD, independently of the degree of measurable cognitive impairment. It may help to understand not only the mechanisms of pathology but also effects of compensation and cognitive reserve. This may be crucial for the development and evaluation of treatment strategies for this debilitating disorder. References [1] H. Forstl, A. Kurz, Eur. Arch. Psychiatry Clin. Neurosci. 249 (1999) 288–290. [2] G. McKhann, D. Drachman, M. Folstein, R. Katzman, D. Price, E.M. Stadlan, Neurology 34 (1984) 939–944. [3] B.J. Small, J.L. Mobly, E.J. Laukka, S. Jones, L. Backman, Acta Neurol. Scand. Suppl. 179 (2003) 29–33. [4] E. Braak, K. Griffing, K. Arai, J. Bohl, H. Bratzke, H. Braak, Eur. Arch. Psychiatry Clin. Neurosci. 249 (Suppl. 3) (1999) 14–22. [5] R.C. Petersen, G.E. Smith, S.C. Waring, R.J. Ivnik, E.G. Tangalos, E. Kokmen, Arch. Neurol. 56 (1999) 303–308. [6] R.C. Petersen, J. Intern. Med. 256 (2004) 183–194. [7] R.C. Petersen, R. Doody, A. Kurz, et al., Arch. Neurol. 58 (2001) 1985–1992. [8] E. Arnaiz, O. Almkvist, R.J. Ivnik, et al., J. Neurol. Neurosurg. Psychiatry 75 (2004) 1275–1280. [9] E. Zamrini, S. De Santi, M. Tolar, Neurobiol. Aging 25 (2004) 685–691. [10] J.R. Petrella, R.E. Coleman, P.M. Doraiswamy, Radiology 226 (2003) 315–336.

A. Drzezga / Methods 44 (2008) 304–314 [11] M.E. Phelps, H.R. Schelbert, J.C. Mazziotta, Ann. Intern. Med. 98 (1983) 339–359. [12] P.J. Magistretti, L. Pellerin, Philos. Trans. R. Soc. Lond B Biol. Sci. 354 (1999) 1155–1163. [13] A.B. Rocher, F. Chapon, X. Blaizot, J.C. Baron, C. Chavoix, Neuroimage 20 (2003) 1894–1898. [14] K. Herholz, Alzheimer Dis. Assoc. Disord. 9 (1995) 6–16. [15] S. Minoshima, Neuroimaging Clin. N. Am. 13 (2003) 769–780. [16] D.H. Silverman, G.W. Small, C.Y. Chang, et al., Jama 286 (2001) 2120–2127. [17] J. Valla, J.D. Berndt, F. Gonzalez-Lima, J. Neurosci. 21 (2001) 4923–4930. [18] J.C. Mazziotta, R.S. Frackowiak, M.E. Phelps, Semin. Nucl. Med. 22 (1992) 233–246. [19] E. Salmon, B. Sadzot, P. Maquet, et al., J. Nucl. Med. 35 (1994) 391–398. [20] G.W. Small, West. J. Med. 171 (1999) 293–294. [21] A. Drzezga, T. Grimmer, M. Riemenschneider, et al., J. Nucl. Med. 46 (2005) 1625–1632. [22] S. Minoshima, B. Giordani, S. Berent, K.A. Frey, N.L. Foster, D.E. Kuhl, Ann. Neurol. 42 (1997) 85–94. [23] L. Mosconi, Eur. J. Nucl. Med. Mol. Imaging 32 (2005) 486–510. [24] A. Drzezga, N. Lautenschlager, H. Siebner, et al., Eur. J. Nucl. Med. Mol. Imaging 30 (2003) 1104–1113. [25] L.R. Squire, Neurobiol. Learn. Mem. 82 (2004) 171–177. [26] L.R. Squire, C.E. Stark, R.E. Clark, Annu. Rev. Neurosci. 27 (2004) 279–306. [27] B. Desgranges, J.C. Baron, V. de la Sayette, et al., Brain 121 (Pt. 4) (1998) 611–631. [28] B. Desgranges, J.C. Baron, F. Eustache, Neuroimage 8 (1998) 198–213. [29] M. Moscovitch, R.S. Rosenbaum, A. Gilboa, et al., J. Anat. 207 (2005) 35–66. [30] E. Svoboda, M.C. McKinnon, B. Levine, Neuropsychologia 44 (2006) 2189–2208. [31] E. Tulving, S. Kapur, F.I. Craik, M. Moscovitch, S. Houle, Proc. Natl. Acad. Sci. USA 91 (1994) 2016–2020. [32] M.J. de Leon, A. Convit, S. DeSanti, et al., Neuroimaging Clin. N. Am. 5 (1995) 1–17. [33] M.J. de Leon, S. DeSanti, R. Zinkowski, et al., J. Intern. Med. 256 (2004) 205–223. [34] N.C. Fox, M.N. Rossor, Rev. Neurol. (Paris) 155 (Suppl. 4) (1999) S33–S37. [35] F. Eustache, P. Piolino, B. Giffard, et al., Brain 127 (2004) 1549–1560. [36] J.C. Morris, A. Heyman, R.C. Mohs, et al., Neurology 39 (1989) 1159–1165. [37] S.J. Teipel, F. Willoch, K. Ishii, et al., Neurobiol. Aging 27 (2006) 681–690. [38] B. Desgranges, J.C. Baron, C. Lalevee, et al., Brain 125 (2002) 1116–1124. [39] R. Cabeza, L. Nyberg, Curr. Opin. Neurol. 13 (2000) 415–421. [40] D. Chan, N. Fox, M. Rossor, Neurology 58 (2002) 838. [41] D. Chan, N.C. Fox, R.I. Scahill, et al., Am. Neurol. 49 (2001) 433–442. [42] R.R. Davies, J.R. Hodges, J.J. Kril, K. Patterson, G.M. Halliday, J.H. Xuereb, Brain 128 (2005) 1984–1995. [43] C.J. Galton, K. Patterson, K. Graham, et al., Neurology 57 (2001) 216–225. [44] J. Diehl, T. Grimmer, A. Drzezga, M. Riemenschneider, H. Forstl, A. Kurz, Neurobiol. Aging 25 (2004) 1051–1056. [45] J. Diehl-Schmid, T. Grimmer, A. Drzezga, et al., Dement. Geriatr. Cogn. Disord. 22 (2006) 346–351. [46] P.J. Nestor, T.D. Fryer, J.R. Hodges, Neuroimage 30 (2006) 1010–1020. [47] B. Desgranges, V. Matuszewski, P. Piolino, et al., Neurobiol. Aging (2006).

313

[48] N. Hirono, E. Mori, K. Ishii, et al., Dement. Geriatr. Cogn. Disord. 12 (2001) 15–21. [49] R. Perneczky, J. Hartmann, T. Grimmer, A. Drzezga, A. Kurz. Cerebral metabolic correlates of the clinical dementia rating scale in mild cognitive impairment. J. Geriatr. Psychiatry and Neurol. (2007), in press. [50] L. Mosconi, W.H. Tsui, S. De Santi, et al., Neurology 64 (2005) 1860–1867. [51] G. Chetelat, B. Desgranges, V. de la Sayette, et al., Brain 126 (2003) 1955–1967. [52] L. Ercoli, P. Siddarth, S.C. Huang, et al., Arch. Gen. Psychiatry 63 (2006) 442–448. [53] M.J. de Leon, A. Convit, O.T. Wolf, et al., Proc. Natl. Acad. Sci. USA 98 (2001) 10966–10971. [54] K. Herholz, H. Schopphoff, M. Schmidt, et al., J. Nucl. Med. 43 (2002) 21–26. [55] F. Nobili, A. Brugnolo, P. Calvini, et al., Clin. Neurophysiol. 116 (2005) 364–375. [56] R. Perneczky, A. Drzezga, J. Diehl-Schmid, et al., Brain reserve capacity in frontotemporal dementia: a voxel-based 18F-FDG PET study, Eur. J. Nucl. Med. Mol. Imaging (2007), Available online 12 January 2007. [57] N. Scarmeas, Y. Stern, Curr. Neurol. Neurosci. Rep. 4 (2004) 374–380. [58] D.A. Bennett, R.S. Wilson, J.A. Schneider, et al., Neurology 60 (2003) 1909–1915. [59] Y. Stern, J. Clin. Exp. Neuropsychol. 25 (2003) 589–593. [60] Y. Stern, J. Int. Neuropsychol. Soc. 8 (2002) 448–460. [61] Y. Stern, S. Albert, M.X. Tang, W.Y. Tsai, Neurology 53 (1999) 1942–1947. [62] R. Perneczky, A. Drzezga, J. Diehl-Schmid, et al., J. Neurol. Neurosurg. Psychiatry 77 (2006) 1060–1063. [63] D. Prvulovic, V. Van de Ven, A.T. Sack, K. Maurer, D.E. Linden, Psychiatry Res. 140 (2005) 97–113. [64] A. Drzezga, T. Grimmer, M. Peller, et al., PLoS Med. 2 (2005) e288. [65] C.L. Grady, J.V. Haxby, B. Horwitz, et al., Neurobiol. Aging 14 (1993) 35–44. [66] M.M. Machulda, H.A. Ward, B. Borowski, et al., Neurology 61 (2003) 500–506. [67] S.I. Rapoport, Ann. N. Y. Acad. Sci. 893 (1999) 138–153. [68] N. Scarmeas, K.E. Anderson, J. Hilton, et al., Neurology 63 (2004) 913–915. [69] N. Scarmeas, C. Habeck, K.E. Anderson, et al., Am. J. Geriatr. Psychiatry 12 (2004) 596–605. [70] N. Scarmeas, C.G. Habeck, J. Hilton, et al., J. Neurol. Neurosurg. Psychiatry 76 (2005) 1440–1444. [71] T. Kato, D. Knopman, H. Liu, Neurology 57 (2001) 812–816. [72] S.A. Rombouts, F. Barkhof, D.J. Veltman, et al., AJNR Am. J. Neuroradiol. 21 (2000) 1869–1875. [73] L. Backman, J.L. Andersson, L. Nyberg, B. Winblad, A. Nordberg, O. Almkvist, Neurology 52 (1999) 1861–1870. [74] F. Remy, F. Mirrashed, B. Campbell, W. Richter, Neuroimage 25 (2005) 253–266. [75] G. Gron, D. Bittner, B. Schmitz, A.P. Wunderlich, M.W. Riepe, Ann. Neurol. 51 (2002) 491–498. [76] S.A. Small, G.M. Perera, R. DeLaPaz, R. Mayeux, Y. Stern, Ann. Neurol. 45 (1999) 466–472. [77] B.C. Dickerson, D.H. Salat, J.F. Bates, et al., Ann. Neurol. 56 (2004) 27–35. [78] F.Z. Yetkin, R.N. Rosenberg, M.F. Weiner, P.D. Purdy, C.M. Cullum, Eur. Radiol. 16 (2006) 193–206. [79] S.C. Johnson, T.W. Schmitz, C.H. Moritz, et al., Neurobiol. Aging 27 (2006) 1604–1612. [80] B.C. Dickerson, D.H. Salat, D.N. Greve, et al., Neurology 65 (2005) 404–411. [81] K.A. Celone, V.D. Calhoun, B.C. Dickerson, et al., J. Neurosci. 26 (2006) 10222–10231.

314

A. Drzezga / Methods 44 (2008) 304–314

[82] S.S. Bassett, D.M. Yousem, C. Cristinzio, et al., Brain 129 (2006) 1229–1239. [83] E. Elgh, A. Larsson, S. Eriksson, L. Nyberg, Int. Psychogeriatr. 15 (2003) 121–133. [84] J. Lind, J. Persson, M. Ingvar, et al., Brain 129 (2006) 1240–1248. [85] M.A. Trivedi, T.W. Schmitz, M.L. Ries, et al., BMC Med. 4 (2006) 1. [86] S.Y. Bookheimer, M.H. Strojwas, M.S. Cohen, et al., N. Engl. J. Med. 343 (2000) 450–456. [87] M.W. Bondi, W.S. Houston, L.T. Eyler, G.G. Brown, Neurology 64 (2005) 501–508. [88] A.S. Fleisher, W.S. Houston, L.T. Eyler, et al., Arch. Neurol. 62 (2005) 1881–1888. [89] H.A. Wishart, A.J. Saykin, L.A. Rabin, et al., Am. J. Psychiatry 163 (2006) 1603–1610. [90] C.D. Smith, A.H. Andersen, R.J. Kryscio, et al., Neurology 58 (2002) 1197–1202. [91] C.L. Grady, A.R. McIntosh, S. Beig, M.L. Keightley, H. Burian, S.E. Black, J. Neurosci. 23 (2003) 986–993. [92] J.T. Becker, M.A. Mintun, K. Aleva, M.B. Wiseman, T. Nichols, S.T. DeKosky, Neurology 46 (1996) 692–700. [93] J.L. Woodard, S.T. Grafton, J.R. Votaw, R.C. Green, M.E. Dobraski, J.M. Hoffman, Neuropsychology 12 (1998) 491–504. [94] C.L. Grady, M.L. Furey, P. Pietrini, B. Horwitz, S.I. Rapoport, Brain 124 (2001) 739–756.

[95] N. Scarmeas, E. Zarahn, K.E. Anderson, et al., Arch. Neurol. 61 (2004) 73–78. [96] P.M. Grasby, C.D. Frith, K.J. Friston, et al., Brain 117 (Pt. 6) (1994) 1271–1282. [97] S.I. Rapoport, C.L. Grady, Int. J. Neurosci. 70 (1993) 39–56. [98] M.L. Furey, P. Pietrini, J.V. Haxby, et al., Proc. Natl. Acad. Sci. USA 94 (1997) 6512–6516. [99] P.J. Laurienti, J.H. Burdette, M.T. Wallace, Y.F. Yen, A.S. Field, B.E. Stein, J. Cogn. Neurosci. 14 (2002) 420–429. [100] R. Kawashima, B.T. O’Sullivan, P.E. Roland, Proc. Natl. Acad. Sci. USA 92 (1995) 5969–5972. [101] C.L. Grady, M.V. Springer, D. Hongwanishkul, A.R. McIntosh, G. Winocur, J. Cogn. Neurosci. 18 (2006) 227–241. [102] M.E. Raichle, A.M. MacLeod, A.Z. Snyder, W.J. Powers, D.A. Gusnard, G.L. Shulman, Proc. Natl. Acad. Sci. USA 98 (2001) 676–682. [103] C. Lustig, A.Z. Snyder, M. Bhakta, et al., Proc. Natl. Acad. Sci. USA 100 (2003) 14504–14509. [104] R.L. Buckner, A.Z. Snyder, B.J. Shannon, et al., J. Neurosci. 25 (2005) 7709–7717. [105] M.D. Greicius, G. Srivastava, A.L. Reiss, V. Menon, Proc. Natl. Acad. Sci. USA 101 (2004) 4637–4642. [106] S.A. Rombouts, F. Barkhof, R. Goekoop, C.J. Stam, P. Scheltens, Hum. Brain Mapp. 26 (2005) 231–239.