Anatomic and Functional Brain Imaging in Alzheimer's Disease

Stanley I. Rapoport

Currently, there is no certain biological marker for Alzheimer's disease (AD), and the pathophysiological mechanisms of neurodegeneration in AD are not understood. Furthermore, diagnosis is not absolutely certain and depends in large part on excluding other causes of dementia. In life, therefore, AD is diagnosed as "possible" [progressive decline in single cognitive sphere (usually memory), or atypical course of presentation of dementia, or other illness sufficient to cause dementia but not considered the cause of dementia] or "probable" (progressive memory deficit plus additional cognitive defect), according to NINCDS-ADRDA criteria (54). "Definite" AD is diagnosed with a history of dementia and postmortem evidence of critical densities of senile (neuritic) plaques and of neurofibrillary tangles with paired helical filaments within the brain. Possible or probable AD are considered "dementia of the Alzheimer type" (16).

Before the introduction of in vivo brain imaging, information was limited about the brain metabolic and anatomic abnormalities that underlay the signs and symptoms of AD in a given individual. Furthermore, until brain imaging was employed, AD could not be readily diagnosed in life nor readily distinguished from other causes of dementia, which include vascular dementia, normal pressure hydrocephalus, stroke, hemorrhage, neoplasm, frontal lobe atrophy of Pick's disease, caudate nucleus atrophy of Huntington's disease, Wernicke–Korsakoff disease, and some cases Parkinson's disease (4, 17).

Structural imaging in the form of computer-assisted x-ray tomography (CT) was introduced just 20 years ago (39), after which diagnostic accuracy for AD rose from 43% to greater than 70% (4). CT replaced invasive methods of pneumoencephalography and cerebral angiography which had been used sparingly in the elderly because of associated morbidity. More recently, magnetic resonance imaging (MRI) has been introduced to examine brain structure. In the future, magnetic resonance spectroscopy (MRS) looks promising for measuring brain metabolites and metabolite fluxes through brain compartments (72).

With regard to functional imaging, the low-resolution 133Xe clearance technique provided initial evidence for a relation between reduced regional cerebral blood flow (rCBF) in the cerebral cortex and dementia severity in AD (30). This technique was replaced in the early 1980s by the more quantitative positron emission tomography (PET), which provided numerical values for (a) regional cerebral metabolic rates for glucose (rCMRglc) and for O2 (rCMRO2) and (b) the regional oxygen extraction fraction (rOEF) (19, 22; also see Methodical Issues in the Neuropathogy of Mental Illness andPositron and Single Photon Emission Tomography: Principals and Applications in Psychpharmacology ), in deep brain structures as well as in the neocortex. PET also has elucidated blood–brain barrier integrity, dopamine metabolism, and specific receptors in brains of AD patients (56, 59, 71, 74). More recently, single photon emission computed tomography (SPECT) has been introduced as a cheaper and more available imaging tool than PET, for clinical measurements of rCBF in AD, and to distinguish AD from vascular dementia (37; see also Tardive Dyskinesia: Epidemiological and Clical Presentation). Dynamic MRI also promises to allow measurements of rCBF in AD patients, with temporal resolutions in seconds and anatomic resolutions in millimeters, free of radiation exposure (47). In this chapter, we consider each of these in vivo imaging techniques with regard to the nature and diagnosis of AD, and we show how the cognitive deficits of AD in individual patients are related to the local functional and structural changes in the brains of these patients.

Because AD is a heterogeneous disease with regard to severity of dementia and cognitive profile (33, 53), our consideration of brain imaging in AD will be most informative if the results are correlated explicitly with cognitive status. Dementia severity can be defined by scores on the Mini Mental State Examination (MMSE): mild, 22–30; moderate, 11–21; and severe, 0–10 (20). Cognitive profiles are summarized with regard to dementia severity in Table 1, for a group of AD patients as compared with healthy controls, as mean scores on tests of different spheres of cognitive function: memory; attention, planning, and abstract reasoning; language; and visuospatial function. The details of these tests and why they were chosen are described elsewhere (31). Moderately demented patients had significantly reduced means scores on all the cognitive tests that were administered. In contrast, mildly demented patients had significantly reduced mean scores only on measures of verbal and visual recent memory (Wechsler Memory Scale), attention to a complex set (Trailmaking B and Stroop Color Interference Task), and planning (Porteus Mazes), but performance on most tests of "focal" neuropsychological functions was not significantly impaired. These findings confirm that a memory deficit usually is the first prominent complaint in AD and show that attention, planning and abstract reasoning are also affected. Indeed, with an appropriate history and neurologic examination and laboratory testing, a memory deficit is sufficient to make a diagnosis of "possible" AD (16, 54).

Cognitive profiles even for patients of equivalent dementia severity are heterogeneous, and their rates of change, which likely are biphasic, can differ markedly among individual patients (33, 53). When the Wechsler Adult Intelligence Scale (WAIS) and the Dementia Rating Scale were repeated over periods of 2.7–6.8 years in 16 AD patients, an initial plateau phase, during which language and cognitive functions did not change for 9–35 months, was observed in five patients who presented with an isolated memory impairment (33). Once nonmemory functions began to decline in the second phase in these patients, the rate of decline was remarkably steady in each patient but varied two- to threefold among patients.

Computer-assisted x-ray tomography (CT), whether analyzed qualitatively or quantitatively, cannot easily distinguish mildly or moderately demented AD patients from healthy age-matched controls, when causes of dementia other than AD are excluded (12, 50). This lack of sensitivity is due largely to a marked overlap in brain atrophy between AD patients and healthy elderly (9, 17, 42).

Measures of "cortical atrophy" which have been used to try to discriminate individual AD patients from normals include the width of the largest sulci and the outlined perimeter of the brain, standardized to cranial size (12, 50). In an attempt to overcome limitations of irregular ventricular shape, net ventricular volumes have been calculated by summing ventricular area on serial CT slices, and then multiplying the sum by interslice distance. However, volumetric measurements also have proven of limited use (12, 24).

CT-derived estimates of volume demonstrated statistically significant reductions in mean gray matter volume and in the mean gray/white matter volume ratio, and significant increases in mean cerebrospinal fluid (CSF) and lateral ventricular volumes, in relation to dementia severity in a cross-sectional study (9). When brain atrophy then was examined in a longitudinal design, that the mean rate of enlargement of linear indices of ventricular size over 1 year was significantly larger in mildly demented AD patients than in controls (24). In another longitudinal study, rates of change of lateral ventricle volume on CT (cm3/year) completely separated a group of 12 male AD patients (including 8 with mild dementia), examined during a mean interval of 1.4 years, from controls who were examined during a mean interval of 3.3 years (51). Furthermore, rates of ventricular enlargement in individual AD patients were correlated significantly with rates of decline on a composite neuropsychological test battery.

A more extensive longitudinal CT study was conducted on 11 men and 9 women with presumed AD (6 of whom had an isolated memory impairment), who were followed for 9 months to 7 years, and on 9 male and 8 female controls (11). The rate of total lateral ventricle enlargement (cm3/year) differed significantly between patients and controls, with a 94% specificity (the ability to make a correct positive diagnosis) and 90% sensitivity (the ability to correctly exclude AD) (one control and two patients were misdiagnosed). For six initially "possible" AD patients, the rate of ventricular enlargement during the period of isolated memory impairment was significantly less than that after the appearance of focal neuropsychological deficits, suggesting a biphasic atrophic process. The diagnostic power of the volumetric measurements from two CT scans taken 1 year apart was only 0.33 in the mildly demented patients (11).

In summary, single CT scans, whether analyzed qualitatively or quantitatively, cannot distinguish AD patients from normals because of overlap of brain atrophy in the two groups. Serial quantitative CT of lateral ventricular volume has demonstrated a distinct difference in rate of dilatation between AD and normal subjects, and it deserves to be evaluated further for identifying AD patients.

Limitations of CT include its low spatial resolution and low tissue contrast differences, as well as an artifactual elevation of brain CT density adjacent to the skull ("bone hardening artifact") which renders CT unreliable for measuring subarachnoid CSF and cortical atrophy. MRI can overcome some of these limitations, because MRI requires no ionizing radiation, repeated measures are without known risk, and images are free of a bone hardening artifact (although MRI scans do contain spectral inhomogeneities) (17, 58). A thorough description of MRI methodology, and of the limitations and advantages of MRI, is presented elsewhere in this volume (Methodical Issues in the Neuropathogy of Mental Illness). Thus MRI, but not CT, can be used to quantitate lobar and cortical atrophy as well as volumes of subarachnoid CSF. MRI is more costly than CT ($1200–$1800 as compared to $450 for contrast studies of the head), and it should be used clinically when CT is considered inadequate for diagnosis.

Whereas a cross-sectional volumetric CT study of the brain demonstrated no mean difference between mildly demented AD patients and controls (see above) (9), a comparable MRI study did show statistically significant group differences. Mildly demented AD patients had significantly smaller mean cerebral brain matter and temporal lobe volumes and significantly larger volumes of the lateral ventricles and of temporal lobe CSF than did controls (58). Severity of dementia correlated significantly with reduced brain matter volume and increased lateral ventricular volume, confirming progression of brain atrophy. Sparing of caudate, lenticular, and thalamic nuclear groups was evident, consistent with their lesser neuropathology post mortem (64) and their lesser PET-derived metabolic reductions in life (see below).

Discriminant analysis is a statistical procedure that constructs a linear combination of observed variables that best describes group differences, and it can classify group membership of any individual. When applied to MRI volumetric data, a discriminant analysis distinguished each of 31 mildly demented AD patients from sex- and age-matched controls (13). Age and brain volume were the most significant discriminators for men, whereas temporal lobe and CSF volumes were best for women. Because 10 of the patients had a diagnosis of "possible" AD, with impaired memory as the only apparent cognitive deficit (54), it appears that discriminant analysis using volumetric MRI variables can add diagnostic certainty in mildly demented, "possible" AD patients and should be explored further in this regard.

Of 2143 demented patients reported as of 1988 with a diagnosis of AD and/or vascular dementia, having 15% pathological confirmation, 51% had a diagnosis of AD alone, 23% had a diagnosis of vascular dementia alone, and 15% were considered to have AD plus cerebrovascular disease (4). Thus, vascular disease contributes to 38% of reported dementias, and its distinction from AD is a major clinical problem. This is particularly important because many of the cardiovascular risk factors for vascular dementia can be prevented or treated (29). There are a number of different causes of vascular dementia, including arteriosclerotic encephalopathy (lacunar state, multiple small infarcts, large cerebral infarcts), hypertensive arteriosclerosis (including mixed cortical and subcortical leukoencephalopathy of Binswanger), and congophilic angiopathy (7, 40). In subcortical disease (leukoencephopathy), CT and T2-weighted MRI images demonstrate periventricular and deep white matter changes, referred to as leukoaraiosis.

To date, diagnostic criteria to distinguish vascular dementia from AD have not formally incorporated neuroimaging, which published data nevertheless demonstrate can be very helpful (7, 54, 69). DSM-III-R identifies the following diagnostic criteria for AD: dementia, insidious onset with a generally progressive deteriorating course, and exclusion of other specific causes. Diagnostic criteria for multi-infarct (vascular dementia), as listed by DSM-III-R, are as follows: dementia; stepwise deteriorating course with "patchy" distribution of deficits, focal neurological signs and symptoms, and evidence of significant cerebrovascular disease judged to be etiologically related to the disturbance (16). A Hachinski Ischemia Scale Score exceeding 6, with emphasis on focal neurological signs, points to vascular dementia (7, 69).

The relation between white matter hyperintensities on CT or MRI and cognitive and metabolic function is not entirely evident. Studies indicate that 30–80% of elderly individuals without neurologic signs have focal density abnormalities in cerebral white matter, and that the frequency of these abnormalities increases with hypertension (7). Abnormalities can be small focal or confluent areas of increased signal intensity on T2-weighted MRI, which demonstrates them more frequently than does CT (17). Abnormalities often are scattered throughout deep white matter and basal ganglia, and they cap lateral ventricular margins and are thought to indicate increased tissue water due to vascular disease (17). However, cognitive testing has failed to demonstrate differences between healthy aged adults with and without white-matter hyperintensities (1).

In the absence of hypertension, AD patients do not demonstrate a higher frequency of grade 2–3 white-matter abnormalities (17%) than do elderly nonhypertensive healthy controls (27%) (44). Nevertheless, in demented AD and vascular disease patients, severity of cognitive impairment, particularly in tests of subcortical function, has been related to the grade of white-matter hyperintensities by MRI (1). This suggests that in the absence of hypertension, white-matter hyperintensities can represent brain changes that are clinically subthreshold, but that the subcortical pathology they represent in the presence of hypertension is more severe and can add to cognitive impairment in vascular dementia as well as AD patients.

In this regard, 14 AD patients with severe leukoencephalopathy on MRI, 9 of whom were hypertensive, were compared by cognitive testing and PET (see below) with 13 age- and severity-matched nonhypertensive AD patients free of leukoencephalopathy (10). No significant difference was found on any psychological test score or grade of leukoencephalopathy between the groups, but a ratio analysis of PET data indicated lesser rCMRglc in the caudate nucleus and thalamus of the patients with white-matter changes than of the patients without these changes, consistent with subcortical dysfunction.

In the above study, three of the leukoencephalopathic patients who were not hypertensive have come to autopsy (10). Each demonstrated the senile plaques and tangles of AD, as well as extensive myelin pallor in white matter in the area of distribution of the white matter hyperintensities during life. Furthermore, in each, Congo-red staining revealed striking amyloid deposition (amyloid angiopathy) in meningeal and cerebral perforating arteries, but normal-appearing arteries in white matter and in the lenticulo-striate vasculature. There was no evidence of hypertensive lipohyalinosis or atherosclerosis in these vessels. Thus, white-matter changes in nonhypertensive AD patients may reflect the amyloid angiographyphy that is found frequently in the AD brain (severe in 30% of cases).

Grading of white-matter lesions on MRI to distinguish AD from vascular dementia has been of limited success. In one study, scores of periventricular lesions could not distinguish the two groups; scores of subcortical lesions, although somewhat better, showed too much overlap to be useful (5). Forty percent of the AD patients did not have subcortical white-matter changes, whereas such changes were present in all the vascular dementia patients. In another study, vascular dementia patients had higher mean frequencies of infarcts and lacunae (p < 0.001) and of focal signal hyperintensities (p < 0.05) in the basal ganglia and thalamus than did AD patients (48).

In summary, MRI or CT evidence of infarcts or lacunae, particularly in the basal ganglia and thalamus, points to a clinical diagnosis of cerebrovascular disease plus AD, or of vascular dementia alone. Whereas significant leukoencephalopathy on CT likely reflects cerebrovascular disease, on the more sensitive MRI the pathological significance is less certain. In the absence of hypertension and a high Hachinski Ischemic Score to point to vascular dementia rather than AD, white-matter lesions on MRI alone are of questionable clinical import. In nonhypertensive AD patients, they may represent amyloid angiopathy.

In vivo magnetic resonance spectroscopy (MRS) produces spectra of relatively weak magnetic signals from nuclei of phosphorus, carbon, or non-water hydrogen (the signals are weak due to the small concentrations of these nuclei). These spectra provide information about chemical compounds and the energy state within the brain, and they can be localized to specific brain regions (72).

Phosphomonoesters (e.g., phosphoethanolamine and phosphocholine) are considered anabolic precursors of membrane phospholipids, whereas phosphodiesters (e.g., glycerol-3-phosphoethanolamine and glycerol-3-phosphocholine) are thought to represent catabolic products from the breakdown of phospholipids (61). Using in vivo 31P MRS, AD patients were reported to have a larger-than-normal concentration of phosphomonoesters in the temporoparietal cortex (6). This observation, as well as evidence that phosphomonoesters and phosphodiesters are abnormal in the postmortem AD brain (61), suggested that regenerative processes involving phospholipids occur early in AD, whereas degenerative process occur later on. The phosphocreatinine-to-inorganic phosphorus ratio with 31P MRS was reported to distinguish AD from vascular dementia patients (6).

Disturbed phospholipid metabolism in AD also is suggested by 1H MRS evidence of an increased ratio (by 11–22%) of myoinositol to creatine in the parietal and occipital cortices of AD patients compared to controls, when the ratio of the neuronal marker N-acetylaspartate to creatine was reduced by 5–11% (55). Because myoinositol is part of the phosphatidylinositol molecule, its increased brain concentration in AD could reflect accelerated breakdown of phosphatidylinositol and other phospholipids.

Despite these reports, a careful study which used 1H MRS to localize and calculate dimensions of a brain volume of interest, and 31P MRS to quantitate phosphorus metabolites in this volume of interest when employing internal standards (57), found no significant difference between AD patients and controls in absolute concentrations of adenosine triphosphate, phosphocreatine, inorganic phosphate, phosphomonoesters, or phosphodiesters. Nor was any absolute concentration or concentration ratio in AD related to dementia severity, or to rCMRglc as measured with PET, in the volume of interest. It was concluded that reduced rCMRglc in AD (see below) is unrelated to rate-limited delivery of glucose or oxygen to brain, and that normal levels of high-energy phosphorus metabolites are maintained even with severe dementia. However, because the volume of interest studied with MRS was large and included primary as well as association brain regions, localized changes in association areas may have been obscured. A quantitative MRS study of association cortex in severely demented patients is needed to confirm the conclusions (57).

In the future, in vivo MRS combined with anatomic localization should help to clarify whether there is an energy deficit in AD, and whether abnormalities of phospholipid turnover or of phospholipids in second messenger systems are part of the AD process. In vivo MRS also could be used to determine the actual flux of 13C-glucose moieties into brain glutamate, glutamine and aspartate compartments, in relation to rCMRglc (72).

Measurements of rCMRglc, rCMRO2, rOER, and rCBF with PET have contributed to our understanding of brain functional activity and oxidative metabolism in AD. As described in detail elsewhere in this volume (Methodical Issues in the Neuropathogy of Mental Illness and Positron and Single Photon Emission Tomography: Principals and Applications in Psychpharmacology), with PET, a positron-emitting compound that is administered systemically is taken up by brain, where it releases positrons (positively charged electrons). These collide with electrons and are annihilated, to release two gamma rays at 180° to each other. A ring or sphere of radiation detectors surrounding the head identifies, by coincidence counting with appropriate reconstruction and attenuation algorithms, the quantities and locations of radioactivity within the brain. 18F-2-fluoro-2-deoxy-D-glucose (18F-FDG; radioactive half-life of 18F is 110 min) or 11C-DG (radioactive half-life of 11C is 20 min) has been used to measure rCMRglc with PET, whereas H215O and 15O2 or 15O-CO2 (radioactive half-life of 15O is 2.03 min) have been used to generate rCMRO2 and rCBF images, respectively.

PET can be performed during specific cognitive or pharmacologic activation, or in the absence of direct activation when the subject is in a "resting state" (eyes covered, ears plugged to reduce sensory input) or otherwise at rest with visual and auditory inputs uncontrolled. In the absence of activation, rCBF is proportional (coupled) to both rCMRglc and rCMRO2 in normal subjects and in patients with chronic stable brain disease, including AD (63). During focal stimulation, however, coupling is maintained between rCBF and rCMRglc, but can be disrupted between each of these measures and rCMRO2 (21), implying glycolytic at the expense of oxidative metabolism. Thus, rCBF or rCMRglc, but not rCMRO2, can be employed to quantify focal activation.

More than 20 cross-sectional PET studies of brain functional activity have been reported for AD patients who were diagnosed by DSM-III and/or NINCDS-ADRDA criteria (16, 54). Individual publications and supporting references are summarized elsewhere (66). These data were obtained with scanners differing with regard to anatomic resolution [full width at half-maximum (FWHM) of line spread function], sensitivity, and attenuation correction. Patients studied differed with regard to dementia severity and scanning conditions, whereas controls differed with regard to quality of medical screening. Despite these differences, the overall picture is remarkably consistent. Metabolic reductions were reported throughout the neocortex, more so in association than in primary areas, in AD patients of equivalent dementia severity. Reductions were more severe in relation to dementia severity, ranging from -17% in the prefrontal association cortex of mildly demented patients to -54% in the parietal association cortex of severely demented patients.

The most extensive study with a high-resolution multislice PET scanner [Scanditronix PC 1024-7B, 6-mm in-plane and 10-mm axial resolution] was performed by Kumar et al. (45) on 47 carefully screened AD patients of differing dementia severity, as well as on 30 controls, in the "resting state." Table 2 illustrates rCMRglc values from this study for representative right-sided regions. With the exception of the caudate nucleus, mean metabolic rates even in mildly demented AD patients were significantly less than control means. At each dementia severity in AD, rCMRglc generally was lower in association than in primary neocortices or subcortical nuclei. For example, whereas percent (of control) association cortex metabolism in mildly demented subjects ranged from 74% to 86%, the percent equaled 87% for the primary cortices and 90–93% for the subcortical nuclei. Fig. 1 illustrates such selectivity in PET scans from a severely demented patient with a disease of 8 years' duration.

When high-resolution Scanditronix data were subjected to discriminant analysis (see above), a discriminant function of rCMRglc values in frontal and parietal association areas was derived which correctly identified 87% of mildly to moderately demented AD patients and controls (2). This function thus could be used to convert a "possible" to a "probable" AD diagnosis in mildly demented patients. In this regard, this discriminant function later identified as "AD" an individual with an isolated memory impairment and a family history for autosomal dominant AD, whose PET scan had normal absolute and ratio values of rCMRglc (62). The patient subsequently developed severe dementia and a reduced parietal rCMRglc.

Independently of clinical history and evaluation and of CT or MRI, PET (like SPECT, see below) cannot easily distinguish AD from vascular dementia. Both syndromes demonstrate global reductions in brain metabolism and flow in relation to dementia severity (22), as well as heterogeneity of cognitive and local metabolic deficits. Large asymmetric metabolic or flow reductions that correspond to CT or MRI abnormalities, or reductions in the basal ganglia or thalamus, suggest vascular dementia, whereas sparing of primary as compared with association cortical areas suggests AD (3). Neither AD nor vascular dementia is accompanied by an elevated rOER (22). However, prior to the appearance of dementia in some subjects with leukoaraiosis and hypertension (a major risk factor for vascular dementia), rOER can be elevated (77). Recently, PET demonstrated that the rCBF response to hypercapnia is normal in AD but defective in vascular dementia of the Binswanger type (46).

Mean PET measures of brain metabolism and blood flow (Table 2) mask distinct but heterogeneous metabolic patterns that are found in different AD patients. Four

statistically significant patterns have been identified by a principal components analysis of high-resolution rCMRglc data, normalized via Z scores, from 16 regions of 36 mildly to severely demented AD patients (28). The group patterns are illustrated in Fig. 2. The most common pattern (Group 1, 17 patients) had rCMRglc reduced in superior and inferior parietal lobules and posterior medial temporal lobe. Group 2 (paralimbic) patients (8 of 36) had reduced metabolism in orbitofrontal and anterior cingulate gyri, whereas parietal regions were relatively spared. Group 3 (5 of 36) patients showed reduced left hemisphere metabolism, whereas Group 4 (6 of 36) patients had reductions in frontal, parietal, and temporal cortices.

Each patient group had a characteristic neuropsychological–behavioral profile. Group 2 patients had poorer verbal performance and fluency than did Group 1 (parietal/temporal) patients, but better visuospatial performance and spatial memory. Group 3 patients (left hemisphere) had worse verbal memory, verbal fluency, and calculating ability than did Group 1 patients, but better visuoperceptual performance and drawing. Group 1 patients were likely to be depressed, whereas Group 4 patients tended to show inappropriate behavior and psychotic symptoms. Group 2 patients demonstrated agitation, inappropriate behavior, and personality change, whereas those in Group 3 frequently had depressive symptoms (28).

Different PET metabolic patterns do not appear to be related to etiology in AD. No relation exists between metabolic asymmetry and early- or late-onset AD (27), nor can PET distinguish familial from sporadic AD (35). Another etiologic subgroup is Down's syndrome, in which AD neuropathology and dementia become evident after 35 years of age (70). In nondemented Down's syndrome adults, patterns and absolute values of rCMRglc are the same as in age-matched controls, whereas in demented Down's syndrome adults, PET abnormalities appear which cannot be distinguished from abnormalities in AD.

We have seen that reductions in mean PET measures of regional brain metabolism are more severe in more demented AD patients, that they are more profound in association than primary neocortical areas at each level of dementia severity, and that they fall into at least four statistically independent patterns. PET also has been used to show that cognitive deficits in individual AD patients correspond to specific metabolic deficits. Thus, PET has extended our understanding of brain networks that subserve specific cognitive processes, conceived of from lesion and stimulation studies in humans and higher primates (49). In this section, we discuss right/left asymmetries and parietal/frontal gradients of metabolism in individual AD patients, in relation to test scores of cognitive functions thought to be mediated by the right and left hemispheres or by the parietal and frontal lobes, respectively.

Based on evidence that Extended Range Drawing and Visual Recall tests reflect right neocortical function, and that Syntax Comprehension and Verbal Recall Tests reflect left neocortical function, AD patients and controls were ranked separately on these test scores and differences between ranks were quantified as a "drawing/comprehension discrepancy" or a "visual recall/verbal recall discrepancy" to reflect hemispheric functional asymmetry (31). These discrepancies then were correlated with a metabolic asymmetry index derived with high-resolution PET data, where rCMRglc,right is metabolic rate in a right hemisphere region and where rCMRglc,left is the rate in the homologous left hemisphere region:

Metabolic asymmetry index (%)

Asymmetry indices for four association and two primary cortical areas are illustrated in Fig. 3 for mildly to severely demented AD patients and controls. As shown by asterisks, patients had significantly greater variances (SD2) than did controls in association but not in primary cortices. Distinct metabolic asymmetries were visually evident in PET scans of individual patients (Fig. 2, Group 3). Abnormal variances of metabolic asymmetry were found at each dementia severity (Table 3). They were correlated significantly and in expected directions with cognitive discrepancies in moderately but not in mildly demented patients (Table 4). In the moderately demented patients, relatively lower right-sided metabolism corresponded to worse drawing and visual recall test scores, compared with syntax comprehension and verbal recall test scores, respectively. The opposite was true for relative left-sided hypometabolism (31).

PET scans of AD patients also can display gradients in metabolism between the parietal and frontal lobes (Fig. 2, Groups 1 and 3). Accordingly, cognitive tests of parietal lobe function—Arithmetic Subtest of the WAIS, Syntax Comprehension Test, Extended Range Drawing Test, Block Tapping Span test—were compared in terms of rank ordering and cognitive discrepancies (see above) with tests of prefrontal integrity—Controlled Word Association (FAS) and Trailmaking (Trail A) tests. The cognitive discrepancies then were correlated with parietal/frontal metabolic ratios in AD patients of differing dementia severity (32). As noted in Table 5, statistically significant correlations in the expected directions were evident between cognitive discrepancies and the metabolic ratios in moderately but not in mildly demented patients.

Longitudinal studies of individual AD patients with PET demonstrated frequent retention of initial directions of right/left metabolic asymmetry and parietal/frontal metabolic ratios over many years. The initial direction of metabolic asymmetry in each of 11 mildly demented AD patients was shown not to change for up to 4 years (Fig. 4) (27). Likewise, Spearman correlations between initial and follow-up metabolic ratios at least 1.5 years later

ranged from 0.67 to 0.86 (p < 0.01) (32). These results indicate that the AD process leading to regional hypometabolism may start in one hemisphere before the other [there is no predilection for either hemisphere because mean asymmetries do not differ from unity (Table 3)], or in the parietal before the frontal lobe, and that subsequently rates of decline in both lobes or hemispheres are sufficiently consistent in a given patient to prevent crossover in direction.

This consistency likely explains why metabolic asymmetries in initially mildly demented AD patients were significant predictors of expected cognitive discrepancies that appeared 1–3 years later (language worse with initial left-sided hypometabolism, visuospatial function worse with initial right-sided hypometabolism), and why parietal as compared with frontal hypometabolism in mildly demented patients accurately predicted worst scores on cognitive tests of parietal than of frontal integrity (27, 31). Thus, in initially mildly demented AD patients in whom the direction of metabolic asymmetry was maintained (Fig. 4), Spearman rank-sum correlations between asymmetries and appropriate neuropsychological discrepancies were significant at the last but not at the first evaluation (Table 6) (31).

In summary, PET is a more sensitive marker of early AD than are neuropsychological measures of cognitive functions mediated by the neocortex. Metabolic reductions, asymmetries, and parietal/frontal gradients can be found in mildly demented patients, where they indicate very early dysfunction within neocortical association areas. The directions of asymmetries and gradients are retained over many years in individual patients, in whom they predict either (a) deficits in language as compared with visuospatial function that later appear or (b) discrepancies of parietal compared to prefrontal cognitive test scores, which appear in the later stages of disease. Thus, metabolic asymmetry or an abnormal prefrontal/parietal metabolic ratio in a patient with a diagnosis of "possible" AD patient and only a memory deficit implies a diagnosis of "probable" AD.

If rCMRglc largely represents functional activity of terminal synapses and their postsynaptic dendritic connections (see Proto-Oncogens: Beyond Second Messengers, Methodical Issues in the Neuropathogy of Mental Illness, and Positron and Single Photon Emission Tomography: Principals and Applications in Psychpharmacology) a reduced rCMRglc in early AD could be due to synaptic pathology and reduced synaptic efficacy (67). Indeed, synaptic markers are down in the AD brain, and presynaptic elements have been shown by biopsy to drop out very early in disease (15). Reduced synaptic functioning is evidenced in life by fewer significant positive correlations between pairs of PET-derived rCMRglc values in AD as compared with control subjects (38). Fewer correlations imply, furthermore, that rCMRglc in regions without marked pathology (e.g., basal ganglia or thalamus; see Table 2) may decline because of disrupted connections with pathological regions elsewhere in the brain. The fact that uptake of 18F-fluoro-DOPA into the basal ganglia was not altered in AD patients even with extrapyramidal signs, whereas uptake into the putamen was reduced in Parkinson's disease, further indicates that basal-ganglia–substantia-nigra circuitry is not directly affected in AD (74).

In demented patients proven on autopsy to have AD, neurofibrillary tangle densities were common in cortical association areas which had the most reductions in rCMRglc prior to death, but were much less common in primary cortical areas (14). This correlation is consistent with other reports on gradients of tangle distribution and atrophy in the neocortex in AD (64, 66). Because focal atrophy in AD may artificially reduce PET metabolic values due to partial voluming, anatomic registration will be needed in future PET studies to obtain true estimates of metabolism or of flow per gram tissue (see Methodical Issues in the Neuropathogy of Mental Illness and Positron and Single Photon Emission Tomography: Principals and Applications in Psychpharmacology).

Metabolic deficits and neuropathology in association as compared with primary neocortex represent a more general AD neurodegeneration in regions closely connected with the association neocortex—the hippocampal formation, amygdaloid complex, entorhinal cortex, nucleus basalis of Meynert, and catecholaminergic projecting neurons from the brainstem and pons. These regions constitute a telencephalic "association system" that evolved coherently and preferentially in higher primates, particularly in hominids, leading to the hypothesis that AD is a human phylogenic, or evolution-related, neurodegenerative disease (64, 65).

Reduced brain metabolism in AD is unlikely to result from blood–brain barrier damage or ischemia. PET with 68Ga-EDTA demonstrates normal blood–brain barrier permeability in AD (71). Furthermore, a 31P MRS study indicates normal brain concentrations of phosphocreatine, adenosine triphosphate, and inorganic phosphate in AD (58), and PET shows that the rOER is not reduced (22). PET also indicates that transfer coefficients for glucose between blood and brain are not abnormal in AD (23), although capillary density of the glucose carrier is reduced in affected brain regions (41).

By localizing and quantifying focal activation, PET promises to provide a new order of information about neural networks that are defective in AD, about the basis and possible reversibility of these defects, and about how these defects might be altered by appropriate drugs. Thus, Grady et al. (26) used H215O and a high-resolution Scanditronix multislice scanner to examine rCBF in occipitotemporal visual association regions while AD patients or normals performed a control or face-matching task. rCBF during the control task (a button was pressed alternately with right and left thumbs in response to a neutral visual stimulus) was subtracted from rCBF during the face-matching task (the button was pressed with the appropriate thumb after deciding whether the right or left face was to be matched) to produce a "difference" image. Mildly to moderately demented AD patients were chosen who were capable of performing the face-matching task as accurately [85 ± 8 (SD)% correct choices] as the normals (92 ± 5%, respectively).

As illustrated in Table 7, a reduced rCBF ratio during the control task, between association occipitotemporal (Brodmann areas 19 and 37) and primary occipital (Brodmann areas 17) regions, indicated impaired functional activity in a visual association area of the AD patients. However, during face matching, the mean increment in absolute rCBF (ml/100 g/min) did not differ between patients and controls (26, 67). Thus, reduced unstimulated functional activity in visual association areas in some mildly demented AD patients is not accompanied by a reduced capacity of these areas to respond fully during face recognition, suggesting a degree of reversibility of functional failure, possibly through recovery of synaptic efficacy (67). It also was noted that frontal lobe regions were activated in the AD patients but not in the normals during face matching (26). This was ascribed to greater effort by the patients, whose reaction times during the task were more variable than in normals; 3.34 ± 1.46 (SD) sec as compared with 2.07 ± 0.54 sec.

Another study measured rCMRglc twice during the same PET session, in AD patients and normal subjects who first performed a control and then a reading memory task (18). Patients and normals had equivalent global activation [11% ± 13 (SD)% compared to 15 ± 15% (as percent control task)] and equivalent regional activations.

Limited useful data are available with regard to PET and drug effects. In one study, AD patients who were scanned before and during physostigmine infusion demonstrated increments as well as decrements of normalized rCMRglc (73). The one patient who improved clinically had the maximum elevation of rCMRglc. With regard to efficacy of cholinergic therapy, three mildly to moderately demented AD patients, treated orally with the cholinesterase inhibitor tetrahydroaminoacridine for several months, were shown by PET to have increased uptake of (S)(-) 11C-nicotine in the frontal and temporal cortices, suggesting restoration of nicotinic cholinergic receptors and improved rCMRglc as well as neuropsychological performance (59). In another study, when phosphatidylserine was administered to AD patients over a 6-month period, both resting and visual stimulation rCMRglc values were higher after treatment than before, whereas decreases were observed after 6 months in AD patients not given phosphatidylserine (34). Cognitive performance was not markedly changed in drug and nondrug groups.

A large number of positron-emitting ligands for receptors have been developed for use with PET, but few have as yet been applied to AD. These include 11C-nicotine (see above). 11C-Carfentanil has been used to demonstrate early loss of mu opiate receptors in the amygdaloid complex of AD patients, even after correction for atrophy (56).

Its high cost (approximately $4000) and its requirements for extensive cyclotron and technical support limit PET as a routine clinical tool for the diagnosis and characterization of AD. In response, SPECT has been introduced into clinical settings to estimate rCBF and receptor densities. SPECT is cheaper than PET ($500–$700) and requires less personnel and technical support. As described elsewhere in this volume ( Positron and Single Photon Emission Tomography: Principals and Applications in Psychpharmacology), high-sensitivity special-purpose gamma cameras can be used with 133Xe gas to image single slices during 4–7.5 min with a resolution (FWHM) approaching 8 mm. Alternatively, low-sensitivity rotating gamma cameras can provide volumetric information, using commercially available long-lived 123I- or 99mTc-labeled radiopharmaceuticals (36, 37). The use of long-lived isotopes with a need for long signal acquisition times extends scanning time to 30–60 min and allows only 1 scan per day. SPECT is not truly quantitative because photon scatter and attenuation compensation have not been adequately addressed in its reconstruction paradigms. Ratios of radioactivity between one brain region and another usually are determined, and arterial input functions for absolute flow or density calculations usually are not obtained.

Estimates of specificity in diagnosing AD, based on studies using 123I-N-isopropyl-p-iodoamphetamine SPECT, range from 50% to 100%, whereas estimates of sensitivity range from 97% to 100% (43). Because most of these reports lacked pathological confirmation and used moderately to severely demented patients, their relevance to the early diagnosis of AD is limited.

Whereas four statistically distinct PET patterns of rCMRglc have been reported in AD (see above) (28), seven qualitative rCBF patterns have been identified using 99mTc-HMPAO SPECT from a prospective study of 132 consecutive demented patients of varying dementia severity (36). Probability of AD was 82% for bilateral temporoparietal defects, 72% for bilateral temporoparietal plus additional defects, 57% for unilateral temporoparietal defects, 43% for frontal defects, 18% for large focal defects, and 0% for multiple small cortical defects. In another study of mildly to moderately demented AD patients, 99mTc-HMPAO SPECT demonstrated depressed rCBF ratios of frontal, anterior, posterior temporoparietal, and occipital regions to cerebellum (60). However, a normal SPECT pattern has been reported in 45% of mildly demented patients (68). Another study concluded that 99mTc-HMPAO SPECT has minimal diagnostic utility in mildly demented AD patients unless considerable doubt about the diagnosis exists on clinical ground (8).

A rule of thumb derived from SPECT studies of moderately to severely demented patients is that a temporoparietal reduction in rCBF is more likely with AD, whereas a patchy whole brain reduction is more likely with vascular dementia (48). But this rule is not always followed, because normal SPECT images occur in mildly demented AD and vascular dementia patients, and an asymmetric presentation in AD may be indistinguishable from pattern of cerebral infarction in another patient (37). A parietal deficit may favor the diagnosis of AD, whereas involvement of the motor cortex may favor vascular dementia, but some patients may have a mixture of both diseases. In a study of AD and vascular dementia patients matched for dementia severity (rarely done), with controls, 99mTc-HMPAO ratios of temporo-parietal cortex, parietal cortex, or frontal cortex to cerebellar radioactivity did not distinguish between the two dementia groups (76).

With regard to neuropharmacology, 0.5 mg intravenous physostigmine was shown by SPECT to increase rCBF in the left posterior parietotemporal region of AD patients whose baseline rCBF was reduced, but not of normals, suggesting up-regulation of postsynaptic cholinergic sensitivity (25). The muscarinic M2-receptor subtype is selectively lost in AD (52), but currently no M2-selective ligand which can penetrate the blood–brain barrier is available for use with SPECT. However, the radioligand 123I-QNB has different pharmacokinetic properties for the M1 as compared with the M2 receptors. These properties may allow SPECT imaging with 123I-QNB to quantitate loss of the M2 subtype in the AD brain, if the regional M2/M1 receptor ratio normally is 1/1 or higher (78). This is unlikely in the parietal cortex, however, where M1 = 88 nM and M2 = 12 nM in normal brain (78). Focal reductions of 123I-QNB binding by SPECT, in excess of reductions in rCMRglc by PET, have been reported in the thalamus and frontal cortex of AD patients, and they suggest selective loss of M2 receptors in these regions (75).

In summary, SPECT measures of rCBF are of limited application for identifying mildly demented AD patients, except those with an atypical disease presentation, and of limited value for distinguishing moderately to severely demented patients with AD or vascular dementia. If a diagnosis of probable AD has been made by clinical history and examination, by evaluation of the Hachinski Ischemic Scale, and by anatomical imaging, the addition of SPECT usually is not necessary. Further use of SPECT for exploring receptor changes in AD would be of interest.

Prior to the introduction of functional brain imaging, AD was considered a progressive global dementia whose neuropsychological defects were an expression of overall damage and not dependent on the localization of more severe tissue changes. It was argued that there are homogeneous stages of deterioration, involving simultaneous worsening of aphasic, apraxic, and agnosic symptoms (66). Consistent with staging is evidence that mean metabolic reductions within the neocortex are proportional on average to dementia severity in AD patients (Table 2), and that memory and attention test scores are defective in mildly demented AD patients, whereas moderately to severely demented changes have more global defects (Table 1). Furthermore, volumetric CT demonstrates progressive dilatation of the lateral ventricles in relation to overall dementia severity (9).

Nevertheless, quantitative MRI now indicates more atrophy in the neocortex than in the basal ganglia and thalamus (58), in agreement with in vivo PET measurements and postmortem observations that AD is a disease of a phylogenically distinct, telencephalic association system which does not include these subcortical regions (64). With regard to the neocortex, PET-derived right–left metabolic asymmetries, posterior/anterior metabolic ratios, and metabolic group patterns defined by a principal-component analysis (28) have demonstrated marked metabolic heterogeneity in AD, and that staging of metabolic reductions as homogeneous is an oversimplification. Metabolic heterogeneities are not related to disease etiology, but they correlate with appropriate patterns of cognitive dysfunction and clinical profiles in individual patients and probably have underlying heterogeneous neuropathological profiles as their basis (14).

The consensus from cross-sectional and longitudinal studies of AD patients, using PET and cognitive testing, is that the disease has an initial and maintained predilection for the frontal, parietal, temporal, and occipital association neocortices compared with primary sensory and motor regions. Right–left metabolic asymmetries and abnormal posterior/anterior metabolic ratios appear in mildly demented patients, are maintained over years, and predict and correspond to neuropsychological discrepancies between language and visuospatial test scores, and between parietal and frontal lobe test scores, that appear with progressing dementia. Vulnerability of the association cortex early and throughout the disease course reflects an overall vulnerability, confirmed by neuropathology, for a telencephalic association system that evolved disproportionately during higher primate evolution (64, 65).

Initial PET studies suggest that some brain regions with low resting rCBF can be fully activated during cognitive stimulation (18, 26). Such stimulation studies deserve to be extended, because they suggest that early functional failure in AD might be reversed by administering drugs which increase synaptic efficacy (67). Dynamic MRI should be of help in this regard, because it can provide signals proportional to rCBF with a spatial resolution of a few millimeters and a temporal resolution of a few seconds, in the absence of ionizing radiation (47).

Future advances in anatomic and functional imaging technology and in tracer development promise to provide further insights into AD, and additional diagnostic applications. High-resolution, high-sensitivity PET scanners, combined with better localization and atrophy correction through MRI–PET superimposition (see Methodical Issues in the Neuropathogy of Mental Illness and Positron and Single Photon Emission Tomography: Principals and Applications in Psychpharmacology), should afford critical information about functional activation of the amygdaloid formation and hippocampal complex, regions known to be pathological in AD, and perhaps help to identify mildly demented patients with only a memory deficit, or even subjects at risk for AD.

published 2000