Series of Reports
Department of Neurology
No. 37
|
University of Kuopio Series of Reports Department of Neurology No. 37
|
MRI of Hippocampus In Incipient Alzheimer's DiseaseBy KUOPION YLIOPISTO
MIKKO LAAKSO Academic Dissertation
To be presented with the assent of the Faculty of Medicine of the University of Kuopio for public examination in Auditorium L1 of the Canthia Building of the University of Kuopio, on Saturday May 4, 1996, at noon. Departments of Neurology and Radiology Kuopio 1996
Docent Kaarina Partanen M.D., Ph.D. Reviewers: Docent Juha Rinne Opponent:
ABSTRACTAlzheimer's disease (AD) is the most common cause of dementia, yet impossible to diagnose precisely without invasive techniques, particularly at the onset of the disease. Therefore, a reliable diagnostic method is needed. The hippocampus is a part of the mesial temporal lobe memory system, and known to be affected early in the course of AD. Recent development of imaging techniques, particularly magnetic resonance imaging (MRI), has made the evaluation of diminutive brain structures, such as the hippocampus, conceivable. The purpose of this study was to focus on the sensitivity and specificity of different approaches of hippocampal imaging by MRI, and their applications for the diagnosis of incipient AD. Hippocampal pathology was evaluated by means of linear (interuncal distance, IUD), planimetric (hippocampal area) and volumetric measurements, complete with T2 relaxometry using an 1.5 T imager. The accuracy of hippocampal measurements was compared to that of the amygdala and the frontal lobes. Various procedures for normalization of the data to the head and brain size were compared. A total of 193 subjects were examined: 59 patients with probable AD according to NINCDS-ADRDA criteria; 43 patients with age-associated memory impairment (AAMI) according to National Institute of Mental Health criteria; nine patients with vascular dementia (VaD) according to DSM-III-R criteria; 20 patients with Parkinson's disease, eight of whom were demented, and 62 cognitively normal control subjects of whom 42 were older and 20 younger than 50 years of age. Bilateral volumetric hippocampal atrophy was a highly sensitive indicator of early AD. The best discriminant function analysis resulted in correct classification of 95 % of AD patients versus non-demented age-matched controls. The volume of the hippocampus also correlated with AD severity as assessed by Mini-Mental Status Examination and with tests assessing delayed recall. In contrast, the volume of the hippocampus was not significantly affected either by aging or AAMI. The specificity of hippocampal atrophy in comparison to other dementias, however, may be limited, since the hippocampus seem to display various patterns of atrophy in VaD and Parkinson's disease with and without dementia as well. The AD group also invariably showed smaller volumes of the amygdala and frontal lobes, smaller hippocampal areas, longer IUDs and prolonged T2. Yet, evaluation of these measurements did not produce as good an accuracy in correct grouping as did hippocampal volumetry, but was compromised by age-dependence of the variables resulting in substantial overlap between the study groups. In conclusion: volumetric hippocampal atrophy is a highly sensitive indicator in early AD. On the other hand, the specificity compared to other dementias with temporal lobe pathology may be limited. Volume of the hippocampus is not significantly affected by age or AAMI, which makes its assessment useful in detecting, or rather excluding AD and differentiating it from benign memory impairment. National Library of Medicine Classification: WM 220, WT 155 Medical Subject Headings: aging; memory disorders; Alzheimer's disease/diagnosis, Alzheimer's disease/radiography; amygdala; dementia, vascular; frontal lobe; hippocampus; Parkinson's disease |
Although history has long forgotten them, Lambini & Sons are generally credited with the Sistine Chapel floor.
-Gary Larson, Unnatural Selections
ACKNOWLEDGEMENTSThis study was carried out at the departments of Neurology and the MRI unit of the Dept. Clinical Radiology of the University and University Hospital of Kuopio, Finland in 1993-1995. The study is a part of a larger research project aimed at investigating characteristics and treatment of memory impairment and dementia. As for me, I like to tell people that I got into the project by an accident. I thought that I was supposed to beep X-rated films day in, day out, instead of X-ray films. The (boring) truth, however, is that I drifted into the project by a mere change. I was first asked to do "a small job" by Kaarina. Eventually, more and more job got done, and the recruitment was confirmed by Hilkka. I got carried away by a subject that turned out to be most interesting. The making of this thesis has really been most rewarding (educationally, that is). And now, here we are. Interest towards the subject made the process smooth, but still there are people who made the process even smoother. I thank my true guardian angels, Hilkka Soininen and Kaarina Partanen, who somehow managed to get the best out of me. No bosses better are likely to come up. Thank You Big Time. This book was officially reviewed by docents Jaakko Kinnunen and Juha Rinne, whom I also thank. Further, I'm grateful to my coworkers Merja Hallikainen, Päivi Hartikainen, Eeva-Liisa Helkala, Tuomo Hänninen, Mervi Könönen, Maarit Lehtovirta, Paavo Riekkinen Jr., and Pauli Vainio for a job well done. Besides me, the blame for poor language is on Bill Gates III (Thesaurus and proof-reader of Microsoft Word) and Jan Six. So it is pretty much unpossible that any linguistic mistakes would eventually be included. Furthermore I would like to thank professors Paavo Riekkinen Sr. and Seppo Soimakallio for providing the facilities and the opportunity. Finally, I thank the people volunteering for the study, particularly the patients, without whom none of this would ever have been possible. This study was supported by the Medical Research Council of the Academy of Finland, the Finnish Medicine Foundation, the North Savo Fund of the Finnish Cultural Foundation, the Finnish Neurology Foundation, and the Instrumentarium Science Foundation. The original articles are reprinted with a permission from the publishers. This book is dedicated to my mother. Paris de Savonia, after hours, April 1996. (For those of you who didn't get this, Kuopio has been sometimes referred to as PdS. But not very often). MPL N.B. - Almost a decade has passed since the publication of this thesis. Well, to make the story short, I still think that the contents of this thesis are valid and not markedly consumed by time. I still sign this work. Kuopio, March 6. 2005
ABBREVIATIONSAAMI Age-associated memory impairment
LIST OF ORIGINAL PUBLICATIONSThis thesis is based on following publications that in the text are
referred by their Roman numerals I- VI I Laakso MP, Soininen H, Partanen K, Helkala E-L, Hartikainen P, Vainio P, Hallikainen M, - Hänninen T, Riekkinen PJ Sr. Volumes of hippocampus, amygdala and frontal lobes in the MRI-based diagnosis of early Alzheimer's disease: correlation with memory functions. Journal of Neural Transmission [Parkinson's Disease-Dementia Section] 1995;9:73-86 II Laakso MP, Soininen H, Partanen K, Hallikainen M, Lehtovirta M, Hänninen T, Vainio P, Riekkinen PJ Sr. The interuncal distance in Alzheimer's disease and age-associated memory impairment. American Journal of Neuroradiology 1995;16:727-734 III Laakso MP, Partanen K, Lehtovirta M, Hallikainen M, Hänninen T, Helkala E-L, Vainio P, Riekkinen PJ Sr, Soininen H. MR T2 relaxometry in Alzheimer's disease and age-associated memory impairment. Neurobiology of Aging; in press IV Laakso MP, Soininen H, Partanen K, Soininen H, Lehtovirta M, Hallikainen M, Hänninen T, Helkala E-L, Vainio P, Riekkinen PJ Sr. MRI of the hippocampus in Alzheimer's disease: sensitivity, specificity and analysis of the incorrectly classified. Submitted. V Laakso MP, Riekkinen P Jr, Partanen K, Lehtovirta M, Hallikainen M, Hänninen T, Helkala E-L, Vainio P, Soininen H. Hippocampal volumes in Alzheimer's disease, Parkinson's disease with and without dementia, and in vascular dementia: a MRI study. Neurology 1996;46:678-681 VI Laakso MP, Partanen K, Lehtovirta M, Hallikainen M, Hänninen T, Vainio P, Riekkinen PJ Sr, Soininen H. MRI of amygdala fails to diagnose Alzheimer's disease. NeuroReport 1995;6:2414-2418
CONTENTS:ABSTRACT ABBREVIATIONS LIST OF ORIGINAL PUBLICATIONS 1. INTRODUCTION 2. REVIEW OF THE LITERATURE 2.1 DEFINITIONS OF THE STUDY CONCEPTS 2.2. ALZHEIMER'S DISEASE 2.3. AGE-ASSOCIATED MEMORY IMPAIRMENT 2.4. NORMAL AGING 2.5. VASCULAR DEMENTIAS 2.6. PARKINSON'S DISEASE AND DEMENTIA 3. AIMS OF THE STUDY 4. MATERIALS AND METHODS 4.1. THE SUBJECTS 4.2. EXAMINATION OF THE STUDY SUBJECTS 4.3. MAGNETIC RESONANCE IMAGING 4.4. VALIDATION STUDIES 4.5. STATISTICAL METHODOLOGY 4.6. ANALYSIS OF THE INCORRECTLY CLASSIFIED 4.7. DETERMINATION OF APOLIPOPROTEIN E GENOTYPE 5. RESULTS 5.1. DEMOGRAPHIC AND COGNITIVE DATA 5.2. VOLUMETRIC STUDIES 5.3. HIPPOCAMPAL AREA 5.4. INTERUNCAL DISTANCE 5.5. RELAXOMETRY 5.6. ANALYSIS OF THE INCORRECTLY CLASSIFIED 6. DISCUSSION 6.1. SENSITIVITY AND SPECIFICITY OF HIPPOCAMPAL MEASUREMENTS IN ALZHEIMER'S DISEASE vs. NONDEMENTED SUBJECTS 6.2. ANALYSIS OF THE INCORRECTLY CLASSIFIED 6.3. THE AMYGDALA 6.4. THE FRONTAL LOBES 6.5. T2 RELAXOMETRY 6.6. HIPPOCAMPAL ATROPHY AMONG NON-ALZHEIMER CONDITIONS AND DEMENTIAS
CONLUSIONS 7. REFERENCES
|
____________
|
________________
|
1. To evaluate hippocampal involvement in incipient AD using volumetry,
measurement of hippocampal area, interuncal distance and T2 relaxometry. |
2. To compare the accuracy of these hippocampal measurements to
those of the amygdala and the frontal lobes. |
3. To compare findings in the AD group to those in nondemented young
and elderly controls, thus evaluating the effect of aging on the normal
hippocampus. |
4. To compare these findings to those in subjects suffering from
AAMI, and to evaluate the state of AAMI with respect to normal aging versus
AD. |
5. In order to assess the specificity of the findings to Alzheimer's
dementia, to compare them to those found in common dementias and degenerative
disorders other than AD, that is, VaD and PD with and without dementia.
|
In summary, to find a method, that would improve the accuracy of
diagnosis of incipient AD. |
The final study population consists of a total of 193 subjects: 59 patients fulfilling the NINCDS-ADRDA criteria for probable AD (McKhann et al. 1984), 44 subjects fulfilling the National Institute of Mental Health criteria for AAMI (Crook et al. 1986), 9 patients with vascular dementia (VaD) according to DSM-III-R criteria (American Psychiatric Association 1987), 20 patients with idiopathic Parkinson's disease (PD), out of whom 8 demented (PDD), 42 cognitively normal age-matched elderly controls (old controls; OC), and 20 cognitively normal controls younger than 50 years of age (YC). Due to different sample sizes at different times in the study segments, and some missing data, this number may vary a little, but on those occasions this is stated separately. Brief clinical characteristics of the study groups are presented in table 3.
Table 3. Basic clinical characteristics of the study groups
| AD | VaD | PDD | PD | AAMI | OC | YC | |
|---|---|---|---|---|---|---|---|
| Nr | 59 | 9 | 8 | 12 | 44 | 42 | 20 |
| M/F | 30/29 | 3/6 | 5/3 | 6/7 | 11/32 | 19/23 | 10/10 |
| Age, Years | 70 +/- 8 | 76 +/- 4 | 71 +/- 2 | 68 +/- 5 | 70 +/- 5 | 72 +/- 4 | 28 +/- 7 |
| Education, Years | 7 +/- 3 | 3 +/- 1 | 6 +/- 3 | 5 +/-3 | 8 +/- 3 | 10 +/- 3 | 16 +/- 3 |
| Duration, Years | 3.0 +/- 1.6 | 3.5 +/- 1.4 | 4.2 +/- 1.5 | ||||
| Duration of PD, yrs | 7.2 +/- 3.3 | 5.4 +/- 2.6 | |||||
| MMSE | 22 +/- 4 | 16 +/- 4 | 19 +/- 5 | 27 +/- 2 | 28 +/- 2 | 28 +/- 1 |
AD, Alzheimer's disease; VaD, vascular dementia; PDD,
Parkinson's disease with dementia; PD, Parkinson's disease; AAMI, age-associated
memory impairment; OC, old controls; YC, young controls; MMSE, Mini-Mental
Status Examination. Possible significant differences may be found in the
results chapter.
The ethics committee of Kuopio University and University Hospital approved
the study. All subjects provided their informed consent for participation
in the study following an explanation of the study protocol.
4.2.1. THE ALZHEIMER PATIENTS
The AD patients were recruited at diagnostic examinations or were recently diagnosed. They underwent a structured interview, a general physical and clinical neurological examination; assessment of clinical severity by Mini Mental Status Examination (MMSE) (Folstein et al. 1975), Clinical Dementia Rating scale (CDR) (Hughes et al. 1982) and Brief Cognitive Rating Scale (Reisberg et al. 1983); assessment of extrapyramidal signs using the Webster Parkinson's Disease scale (Webster 1968); assessment of depressive signs by the Hamilton scale (Hamilton 1960); an extensive battery of laboratory tests to exclude secondary causes of dementia; comprehensive neuropsychological testing listed below; EEG and event related potentials; SPECT scan; and MRI of the brain. All patients scored less than four in the modified ischemic scale (Rosen et al. 1980). Seventeen AD patients had a history of coronary heart disease and 10 had well controlled hypertension. According to the CDR scale, 5 AD patients had questionable dementia (0.5), 34 had mild (1) and 15 had moderate (2) dementia.
4.2.2. THE AGE-ASSOCIATED MEMORY IMPAIRMENT SUBJECTS AND THE ELDERLY CONTROLS
The AAMI and the OC groups were examined alike. They were derived from a randomly selected population of 1049 individuals who participated in an epidemiological study on the prevalence of dementia and memory disorders in the Kuopio area (Koivisto et al. 1995). The AAMI subjects and the age-matched controls were randomly drawn from this population. Finally, subjects were included in the study if they were willing to complete the whole study protocol.
The criteria of Crook et al. (1986) were used to identify AAMI. The AAMI subjects were 50 years of age or older, had subjective memory problems disturbing their everyday life, showed an objective impairment in memory tasks, and got a score of seven and less on the Benton Visual Retention Test (Benton 1967) or 13 or less on the Paired Association Subtest of the Wechsler Memory Scale (WMS) (Wechsler 1945). Otherwise, the cognitive capacity of AAMI subjects was within the normal range: they scored 32 or more on the Wechsler Adult Intelligence Scale Vocabulary Subtest (Wechsler 1955) and 24 or more on the MMSE. They had no other disease or medication accounting for memory disorders. Four AAMI subjects and one elderly control had a coronary heart disease, and four AAMI subjects and four elderly controls had hypertension. The screening also included an interview concerning medical history, registration of subjective memory complaints with a Memory Complaint Questionnaire (Crook et al. 1992), memory tests mentioned in the AAMI criteria, and short neuropsychological tests to identify potential dementia cases (Koivisto et al. 1992). The subjects also filled in the Geriatric Depression Rating Scale (Yesavage et al. 1983) to assess occurrence of depressive symptoms. Further investigation included clinical neurological examination, comprehensive neuropsychological testing, EEG, event related evoked potentials, and MRI.
4.2.3. THE VASCULAR DEMENTIA AND THE PARKINSON GROUPS
All the VaD patients suffered from multi-infarct dementia. All the PD patients had a typical idiopathic PD with the three classical symptoms of PD: tremor, hypokinesia and rigidity. The test battery for these groups basically included the same clinical and laboratory tests and comprehensive neuropsychological testing as the AD, AAMI and OC groups. Special attention was paid to excluding secondary causes of parkinsonism.
4.2.4. THE YOUNG CONTROLS
The young controls were students or staff members volunteering for the study. They were healthy and had no history of CNS or systemic diseases or medication.
4.2.5. NEUROPSYCHOLOGICAL TESTS
A comprehensive battery of neuropsychological tests was used to assess the cognitive performance of the study subjects. Verbal memory was examined with the list learning test using shopping items (Helkala et al. 1988). A "yes" or "no" recognition of the words in the list was asked after a 30 minute delay filled with other psychometric tests. The story recall test with the Boston approach was used as well (Millber et al. 1986). Recall of the story was tested immediately and after a 30 minute delay. Visual memory was examined with the Heaton Visual Reproduction Test (Russell 1975). Recall of the figures was tested both immediately and after a 30 minute delay.
Nelson's version of the Wisconsin Card Sorting Test (Nelson 1976), Trail Making test A and B (Reitan 1958), and Verbal Fluency test (Borkowski et al. 1967) were used to assess executive functions. The maximum time of 150 seconds for Trail Making A and 300 seconds for Trail Making B was allowed. If the test was not completed in the time allowed, the missing letters or numbers were scored as omissions. In the Verbal Fluency Test, the subject was asked to produce as many words as they could beginning with letters P, A and S in one minute for each letter. The score was the number of words correctly named.
Furthermore, verbal functions were assessed with the Boston Naming Test (Kaplan et al. 1983), visuospatial functions with cube copying test, clock setting test and the Block Design subtest of Wechsler Adult Intelligence Scale (Wechsler 1981; Goodglass and Kaplan 1972), and praxic functions of the hand using Luria's method (Helkala et al. 1988) (data not shown).
4.3.1. MRI TECHNIQUE FOR VOLUMETRIC STUDIES
The subjects were scanned with a 1.5 T Magnetom (Siemens, Erlangen) using a standard head coil and a tilted coronal 3D gradient echo sequence (magnetization prepared rapid gradient echo: time of repetition 10 ms, time of echo 4 ms, time of inversion 250 ms, flip angle 12°, field of view 250 mm, matrix 256x192, 1 acquisition) resulting in contiguous T1 weighted partitions with a slice thickness of 1.5 2.0 mm oriented perpendicular to the long axis of the hippocampus (figure III.1.A).
The boundaries of the ROI were outlined by a trackball driven cursor and the number of voxels within the region was calculated by using an in house developed program for standard work console. The outlining of the boundaries always proceeded from anterior to posterior in a sequential fashion. The measurements were performed by a single rater blinded to the clinical data or diagnostic category of the study subjects (I, IV-VI).
4.3.2. DETERMINATION OF VOLUMES
Data from standard anatomical atlases of the human brain (Duvernoy 1988; DeArmond et al. 1989) as well as from several previous articles (Naidich et al. 1987; Bronen and Cheung 1991 a; b; c; Amaral et al. 1992; Jack et al. 1992 a.; Tien et al. 1992; Watson et al. 1992; Bartzokis et al. 1993) were used as guidelines to determine the boundaries of the amygdala and the hippocampus in oblique coronal MRI sections.
The hippocampus
The hippocampus included the dentate gyrus, the hippocampus proper and the subicular complex. The rostral end of the hippocampus when it first appears below the amygdala, was the anatomical starting point. The caudal end of the hippocampus was taken as the section in which the crura of the fornices depart the hippocampal tail (I, IV-VI). Examples of hippocampal delineation are presented in figure 1.
Figure 1. Examples of delineation of the
hippocampus of an Alzheimer patient,
most posterior slice drawn presented at bottom right.
The amygdala
The amygdala is a heterogeneous structure composed of several nuclei in cortical and subcortical areas. The whole amygdala as it appears on the screen was contiguously outlined (I, VI). In detail, the amygdala is considered to include the deep nuclei (which include lateral, basal, accessory basal, and paralaminar nuclei), the superficial nuclei of the amygdala (the anterior cortical nucleus, the medial nucleus, the nucleus of the lateral olfactory tract, the periamygdaloid cortex, and the posterior cortical nucleus), as well as the remaining nuclei (the anterior amygdaloid area, the central nucleus, the amygdalohippocampal area, and the intercalated nuclei) (Amaral et al. 1992). Examples of amygdaloid delineation are presented in figure 2.
Figure 2. Examples of delineation of the amygdala. Alzheimer patient.
The frontal lobes
The frontal lobes were selected as a ROI in order to ascertain whether the atrophy rate of temporal lobe structures exceeds that of the frontal lobes. The frontal lobes were outlined on every third slice. Therefore, due to thin slices, the measurement was done at intervals of 5 mm. The number of slices measured ranged between 14 and 18. The most anterior slice was the one with clearly visible gyri. The most caudal slice included in the measurement was the first one in which the anterior commissure was first present. On the most posterior slices, a straight line was drawn from the bottom of the lateral fissure (ventral insular sulcus) to the medially located choroidal fissure in order to separate the temporal lobe from the frontal lobe. The volumes of the lateral ventricles were also measured and consequently subtracted from the volume of the slice. The volume of each slice was multiplied by three, and thereafter the slice volumes were summed up (I). Examples of frontal lobe delineation are presented in figure 3.
Figure 3. Examples on delineation of the
frontal lobes.
First slice on top left, last on bottom right.
4.3.3. MEASUREMENT OF THE HIPPOCAMPAL AREA
The hippocampal area was measured from the slice in which the anterior commissure was first present when proceeding from anterior to posterior (IV).
4.3.4. MEASUREMENT OF THE INTERUNCAL DISTANCE
The IUD in the study is defined as a shortest possible distance between the hippocampal heads (pes) on a coronal slice at the level where the anterior commissure was first present (II). Measurement of the IUD is presented in figure II. 1
4.3.5. MRI TECHNIQUE FOR T2 RELAXOMETRY
The method used for T2 relaxometry was similar to that described by Jackson et al. (1993). T2 maps were calculated in each of three oblique coronal 8 mm sections from 16 images obtained at echo times of 22 to 262 using a Carr-Purcell-Meiboom-Gill sequence. The interslice gap was 2.0 mm. The tilting angle was perpendicular to the long axis of the hippocampus.
The T2 maps were generated by a computer program that fitted a single exponential to the signal intensity data of corresponding pixels from all 16 echoes after ensuring that no motion artifacts were visible in the source images. The T2 relaxation time was thus calculated for each pixel, and an image was then constructed in which pixel intensity corresponded to the calculated T2. The T2 images thus generated were magnified by a factor of 2.3-2.5.
In this study T2 was measured in the head, body and tail of the hippocampus, in the temporal and parietal white matter at the corresponding levels, as well as in the amygdala and the thalamus. These variables are identical to those of Kirsch et al. (1992) except for the hippocampus being measured at three sites instead of one.
Mean hippocampal T2 was measured within the anatomic boundaries of the hippocampus by selecting the largest possible circular ROI with minimum 8, but typically 30-50 pixels (40-60 mm3), within the anterior, middle and posterior sections corresponding to head, body and tail of the hippocampus, respectively. Boundaries where partial volume effects might occur were carefully avoided. ROIs were placed similarly in the amygdala (100 pixels, 125 mm3, shown in the anterior section in 98 % of cases), the thalamus (100-150 pixels, 125-190 mm3, in the posterior section in 83 % of cases) and the temporal (50-70 pixels, 63-88 mm3) and parietal (200 pixels, 250 mm3) white matter in each of the three sections. Since the values basically do not represent any distinct anatomic locations, the T2 of temporal and parietal white matter is presented as an average of the three sections (III; figures III.1.B.C.)
4.3.5. NORMALIZATION PROCEDURES
In order to exclude the effect of individual head and brain sizes, various methods for normalization were tried in the study. The normalized values were used in the statistical analyses. The methods used for normalization differ between the variables and the substudies, and are presented individually when used.
ICA refers to intracranial area. Two different ICAs were measured. ICA1 (figure 4.) refers to a coronal ICA measured at the level of the anterior commissure, and ICA2 (figure 5.) to a sagittal ICA measured in the midsagittal scout image (II, IV-VI).
Brain area (BA) (figure 6.) refers to the area of the brain measured at the level of the anterior commissure with the lateral and ventricular spaces excluded (I, VI).
Intracranial width (ICW) is a variable that was used in normalization of the IUD (figure II.1). ICW is a straight line through the inner cranium measured horizontally at the level of the head of the hippocampus. In cases where the heads were not on the same horizontal level, the intracranial line was tilted horizontally at the midpoint of the IUD (II).
Figure 4. The coronal intracranial area. A control subject.
Figure 5. The sagittal intracranial area. A control subject.
Figure 6. The brain area. A control subject.
4.4.1. THE VOLUME MEASUREMENTS
The intrarater agreement in volumetry of the hippocampus and the amygdala has been reported earlier (Soininen et al. 1994). The interrater reproducibility between two raters was tested in 16 subjects. The differences between the volumes obtained by two raters compared to the mean of these two measurements were 4.1 % for the right hippocampus, 1.6 % for the left hippocampus, 8.7 % for the right amygdala and 3.7 % for the left amygdala, respectively (I).
4.4.2. THE LINEAR MEASUREMENTS
The interrater reliability for IUD was tested between two raters in 16 subjects. The mean of the measurements was 26.4 for rater 1 and 26.3 for rater 2. The intraclass correlation coefficient was 0.82, ANOVA F(1,15) = 0.014, p = 0.907. Furthermore, in 16 subjects IUD was measured 1 3 slices (max. 4.5 6.0 mm) posterior in the same patient. The correlation coefficient between these two measurements was 0.64, p<0.01 (II).
4.4.3. THE RELAXOMETRY
Intra- or interrater agreements were not evaluated for the measurement of T2. However, in order to assess the stability of T2, five repeated measurements of the same YC volunteer were performed within a 12-month period. The mean coefficient of variation in different locations of the hippocampus and the amygdala was 2.6 % (range from 1.3 to 3.7 %), 6.6 % (6.4 - 6.8 %) in the thalamus, 6.8 % (6.3-7.2 %) in the temporal and 3.4 % (3.0-3.8 %) in the parietal white matter.
The data were analyzed utilizing SPSS-PC+ V.4.1 software (SPSS Inc., Chicago,
IL). Analysis of variance (ANOVA) was used to compare the means over the
study groups. The Duncan method was applied in post hoc analysis to detect
which groups differed significantly. For T2 times that were significantly
different across the study groups, ANOVA adjusted for presence of hypertension
and coronary heart disease was performed. In the analysis of psychometric
test scores, education was included as a covariate (ANCOVA). In the study
of the amygdala, multivariate analysis of variance (MANOVA) was used for
repeated measures by side (right, left) x diagnostic group x gender, for
raw and normalized volumes.
Correlations were calculated using a two-tailed Pearson's correlation test. The categorial data, such as gender and ApoE distribution were analyzed by a Chi-square test. To test the accuracy of the measurements in distinguishing AD patients from controls, stepwise discriminant function analysis (Wilk's method) was used. The results are expressed as mean ± SD. The level of statistical significance of differences is p<0.05.
For discrimination analyses, the IUD and the volumes of hippocampus and amygdala were multiplied times 10 or 100 to produce reasonable numbers for the statistics.
For this analysis the study group was rearranged into four groups: AD patients
correctly classified (ADCC), AD incorrectly classified (misclassified;
ADMC), controls correctly (CCC) and incorrectly classified (CMC). The control
group includes only the AAMI and the OC groups. Thus, and due to some missing
data, the size of the study group varies between 137 and 140. The size
of the individual study groups varies as follows: ADCC 46-49, ADMC 6-9,
CCC 76-79, CMC 6-9. The compared variables in the analysis were basic demographic
data, such as age and gender; assessment of clinical severity with different
rating scales; assessment of accompanying symptoms other than mnemonic,
and the ApoE genotype. Finally, the clinical data of the incorrectly classified
AD patients was individually reappraised (IV).
Samples of venous blood were collected in EDTA-tubes and ApoE genotype
was determined from blood leukocytes. DNA was extracted by the standard
phenol-chloroform extraction. ApoE genotypes were analyzed using polymerase
chain reaction, HhaI digestion and polyacrylamide gel electrophoresis as
described earlier (Hixon and Vernier 1990; Tsukamoto et al. 1993; Lehtovirta
et al. 1995 b.) with slight modifications.
The AD patients and the nondemented elderly controls did not differ significantly in age. Among the demented groups ANOVA showed a significant difference in age, the VaD patients being older than AD or PD patients [F(4,107)=2.7, p<0.05]. The groups were sex-matched except for the AAMI group in which women were overrepresented. Duration of dementia did not differ between the demented patient groups. Duration of PD was slightly longer in the PDD group than in the PD group, but not significantly (IV-VI).
By definition, the MMSE scores were lower for the demented patients than for the age matched study groups (p<0.0001) (IV). The OC and PD groups performed better in the MMSE than the AD, VaD and PDD groups (p<0.05); the VaD group had lower scores than the AD and PDD groups (p<0.05) (VI).
The YC group had had a longer education than the OC, AAMI or AD groups [F(3, 139)=44.5, p<0.0001]. All the groups differed from each other with a significance of p<0.05. Significance of cognitive performance did not change with education as a covariate in the analyses. Compared to the AD group, education was lower in the VaD, PD and PDD groups (p<0.05).
The ischemic scores differed significantly between the groups (p<0.0001), the VaD and PDD groups having higher scores than AD, PD and OC groups (p<0.05). However, none of the PDD patients received an ischemic score higher than 4. The Webster scores were significantly higher in PD and PDD than in AD and VaD groups (ANOVA/Duncan p<0.05). VaD patients had a history of cerebral infarcts. At the time of the neurological examination, right hemiparesis was present in one VaD patient and left hemiparesis in another. Two AD patients also had had a history of cerebral infarct. The frequency of coronary heart disease was 19/50 (38 %) in AD, 6/9 (67 %) in VaD, 3/8 (38 %) in PDD, 2/12 (17 %) in PD patients and 1/34 (3 %) in controls. The frequency of hypertension was 24 % in AD, 100 % in VaD, 0 % in PDD, 8 % in PD and 12 % in OC (VI).
5.2.1. THE HIPPOCAMPUS
Mean hippocampal volumes for the study groups are presented in table 4, and their distribution in figures 7 and 8. In general, men had larger hippocampi than women. In the OC group, hippocampal volumes correlated strongly to the coronal and sagittal intracranial areas (p<0.001). Differences in volumes by gender vanished after normalization. In the nondemented groups, age was not significantly related with the right (r= 0.19) nor the left (r= 0.19) hippocampal volume. The same was true when the analysis was done by gender. In women the r values were -0.13 and -0.20; and in men -0.12 and -0.04, respectively. Normalization did not change the pattern (IV).
The AD patients showed significantly smaller raw and normalized volumes of both hippocampi (ANOVA, p<0.0001) compared to nondemented controls (I, IV). The volumes did not differ significantly when the AD patients were divided into two groups, one group having had the disease less for than and the other for 36 months (the average duration) or more. The absolute volumes were smaller in patients with a longer duration. In the duration less than 36 months group, the volumes of hippocampus were 2394 mm3 on the right and 2141 mm3 on the left. In the 36 months or more group the volumes were 2223 mm3 and 1952 mm3, p=0.38 and p=0.22, respectively. When the patients were grouped into those having had the disease for a year or less vs. the remaining patients, the hippocampal volumes displayed a nonsignificant trend of being larger in the patients with the longer duration. In the less than a year group the volume of the hippocampus on the right was 2294 mm3 and 2036 mm3 on the left. In the duration more than a year group, the corresponding volumes were 2323 mm3 and 2058 mm3, respectively, p=0.91.
Hippocampal volumes did not differ significantly between OC and AAMI groups (I, IV). These groups were therefore combined into one large control group for further analyses. The combined OC/AAMI group consists of 86 subjects matched for age and gender with the AD patients.
The value of volume measurements to differentiate AD patients from controls was tested in five discriminant function analyses (IV). The analyses included 1) the raw volumes of the hippocampus; 2) the raw volumes and gender; 3) the volumes normalized for the coronal ICA1; 4) the volumes normalized for the sagittal ICA2, and 5) the volumes normalized to brain area/ICA1. The young controls were excluded from the discrimination analyses.
The analysis including the volumes of the right and left hippocampus yielded a sensitivity of 83.6 % and a specificity of 89.4 % (Chi square 104.8, df 2, Wilks' lambda 0.46, p<0.0001). The analysis explained 54 % of the variance between groups. The volume of the left hippocampus alone accounted for 53 %. The analyses using normalized volumes resulted in similar sensitivity and specificity. The highest sensitivity (94.4 %) was achieved in the analysis in which the hippocampal volumes were normalized taking into account both the coronal intracranial area and brain area. The results of the analyses are presented in table IV 3.
When the hippocampal volumes of the OC, AD, VaD, PD, and PDD groups were compared, ANOVA showed significant differences over the study groups both on the right and on the left (p<0.0001) (V table 2, figure 1 A and B ). The OC group showed larger volumes on both sides than the other study groups (p<0.05). The right hippocampal volumes of VaD patients were also larger than those of AD, PD and PDD patients (p<0.05). On the left side, the AD patients had significantly smaller volumes than VaD and PD patients. The raw volumes in the PDD group, were even smaller than in the AD group, though not significantly. In the VaD group, two out of the nine patients had no atrophy at all, four had bilateral atrophy and three had unilateral atrophy. These patterns of atrophy could not be explained by other findings in the MRI such as number, size or strategic location of the infarcts (V).
In the AD group, the volume of the left hippocampus significantly correlated with MMSE score (r=0.42, p=0.029), immediate story recall (r=0.39, p=0.029) and delayed story recall (r=0.50, p=0.003) (I, table 5.).
Table 4. Raw hippocampal volumes (mm3)
+/- standard deviation by group.
The distribution of the volumes is graphically presented in figures 7 and
8.
| AD | VaD | PDD | PD | AAMI | OC | YC | |
|---|---|---|---|---|---|---|---|
| Right Hippocampus | 2337 +/- 704 | 2975 +/- 572 | 2152 +/- 391 | 2557 +/- 360 | 3319 +/- 447 | 3394 +/- 519 | 3554 +/- 607 |
| Left Hippocampus | 2070 +/- 562 | 2406 +/- 448 | 2042 +/- 215 | 2401 +/- 380 | 3100 +/- 399 | 3435 +/- 536 | 3478 +/- 466 |
Figure 7. The distribution of the unnormalized volumes
of the left hippocampus of the study groups.
Both genders are included.
Figure 8. The distribution of the unnormalized
volumes of the right hippocampus by the study group.
Both genders are included.
5.2.2. THE AMYGDALA
Amygdaloid volumes were determined for the AD, AAMI, OC and YC groups. Table VI 2 presents the mean raw and normalized amygdaloid volumes for these groups, and mean volumes separately for men and women. Figure VI 1 displays the scatterplots of the volumes per study group.
The volumes were significantly smaller for AD patients than for all other groups (p<0.05). The AAMI and the OC groups did not differ significantly from each other, and were combined into a one large OC/AAMI group. The YC group differed from AD, AAMI and OC in the left amygdaloid volumes and from AD and OC in the right amygdaloid volumes. Accordingly, the normalized volumes differed across the study groups by group [F(3,138)=21.8, p<0.0001] and side [F(1,138)=26.4, p<0.0001] but not by gender (F=0.9, p>0.05) (VI).
In the study groups, age correlated significantly (p<0.001) with raw (Figure VI 2) (r=-0.31 on the right and r=-0.41on the left) and normalized (r=-0.28 on the right and r=-0.37 on the left) amygdaloid volumes. When the AD group was excluded, age still correlated significantly with raw (r=-0.29, p=0.003 on the right and r=-0.37, p<0.0001 on the left) and normalized (r=-0.24, p=0.018 on the right and r=-0.30, p=0.002 on the left) volumes. In the AD group, only the volume of the left amygdala correlated with age (r=-0.28 , p<0.05). On the other hand, the normalized right amygdala was strongly independent of age (r=-0.04, p=0.79). Interestingly, in the young control group, raw (r=-0.63, p=0.003 on the right and r=-0.80, p<0.0001 on the left) and normalized (r=-0.63, p=0.003 on the right and r=-0.75, p<0.0001 on the left) volumes correlated highly significantly with age. In this group the raw and normalized volumes of the left amygdala also correlated with education (p<0.05) (VI).
MANOVA for repeated measures by side (right, left) x group x gender for raw volumes showed significant effects of group [F(3,139)=20.1, p<0.0001), side [F(1,139)=26.8, p<0.0001) and gender [F(1,139)=10.8, p<0.001), but no interactions (p>0.05) group x side, group x gender or gender x side. Irrespective of the diagnostic group, the unnormalized volumes were larger for men and on the left side (VI).
Three discrimination function analyses were performed, including 1. the raw volumes, 2. the normalized volumes, and 3. the normalized volumes and age. Since the volume of amygdala did not differ significantly between the OC and AAMI groups, these groups were combined again, and the YC group was excluded.
The analysis including the normalized volumes produced the best correct classification: 41/54 (75.9 %) of AD patients and 52/72 (72.2 %) of the combined AAMI and OC group, and the overall correct classification 73.8 % (Chi-square 40.2, df 1, Wilks' lambda 0.72, p<0.0001). In all the analyses only the volume of the left amygdala entered the model. Thus, adding the normalized volume of the right amygdala or age, or both, did not improve the diagnostic sensitivity or specificity.
In the AD patients, there was no significant correlation between clinical severity assessed by MMSE and raw or normalized amygdaloid volumes (r=-0.09 on the right and r=0.14 on the left, p>0.05). The duration of the disease did not correlate with the volumes of the amygdala either (VI).
5.2.3. THE FRONTAL LOBES
The volumes of the frontal lobes were determined for AD, AAMI, OC and YC groups. Compared to the YC group, volumes were diminished in all other groups (right unnormalized, F(3, 146)=13.6, p<0.0001, left unnormalized F(3, 145)=12.9, p<0.0001, right normalized F(3, 146)=25.0, p<0.0001, left normalized F(3,145)=23.2, p<0.0001). Again, volumes in the OC and AAMI groups did not differ from each other, but differed from the YC group (p<0.05). Compared to age-matched controls, the volumes of the AD group were diminished in 10.5 % on the right and 12.8 % on the left when unnormalized, and 14.5 % on the right and 15.3 % on the left when normalized.
In the discriminant function analysis the frontal volumes produced the following overall classification accuracies: right unnormalized 68 %, left unnormalized 68 %, right normalized 68 %, and left normalized 71 %.
The volume of the left frontal lobe correlated significantly with the Trail-Making A test (r=-0.44, p=0.013): the longer time spent in the test, the smaller the volume of the left frontal lobe. The number of errors in the Trail-Making A test also correlated with the volume of the left frontal lobe (r=-0.44, p=0.014) (I).
The hippocampal area measured at the level of the anterior commissure was
significantly smaller on the right and left for AD patients and AAMI subjects
than for older and younger controls (p<0.05). On the left AD patients
also differed significantly from AAMI subjects (p<0.05). The areas on
both sides correlated to the actual volume on the right (r=0.70, N=125,
p<0.0001) and on the left (r=0.71, p<0.0001). The discriminant function
analysis between the AD and the OC group produced a sensitivity of 76.9
% and specificity of 72.4 %, resulting in an overall correct classification
of 75.3 % (Chi square 36.1, df 2, Wilks' lambda 0.63, p<0.0001). The
hippocampal areas were unnormalized (IV).
The IUD data are summarized in Table II 2. IUD was measured for AD, AAMI,
OC and YC groups. ANOVA over the study groups showed a significant difference
in standard IUD [F(3,137)=11.4, p<0.0001]. The YC group showed significantly
shorter IUD compared to the three older groups (p<0.01), but the OC,
AAMI and AD groups did not differ from each other. The scatterplots (Figure
II 2 a) demonstrate the overlap of IUD values across the study groups.
IUD exceeded the proposed pathological limit of 30 mm in 21/54 (37 %) of AD patients, in 13/40 (33 %) of AAMI subjects, in 6/27 (22 %) of OC, and in 1/20 (5 %) of YC group. The cut-off point of 30 mm resulted in 37 % sensitivity and 72 % specificity to separate AD from combined AAMI and OC groups. The positive predictive value was 53 % and the negative predictive value was 59 %.
There was also a significant difference in IUD/ICW (p<0.0001). The YC group differed from all the other groups (p<0.01), whereas the values of the AD group were comparable with those of the age-matched older groups. The IUD/ICA and IUD/BA values were also significantly smaller for YC compared to the OC, AAMI and AD groups (p<0.01). In addition, the AD group differed significantly from OC and AAMI groups in IUD/ICA (p<0.05) and IUD/BA (p<0.01). Despite these significant differences, the variables overlapped across the age-matched study groups (Figure II 2 b).
In the whole study population age correlated significantly (p<0.0001) with IUD (r=0.41), IUD/ICW (r=0.41), IUD/ICA (r=0.45), IUD/BA (r=0.55) as well as with brain area (r=-0.59). Similar significant correlations (p<0.0001) with age were observed when only the nondemented subjects (young and old controls and AAMI subjects) were included in the analysis. Within the AD group standard and adjusted IUD was not related to age, but brain area was (r=-0.41, p<0.01). Brain area also correlated significantly with age for AAMI subjects (-0.48, p<0.01) and old controls (r=-0.39, p<=0.05).
In the first stepwise discriminant function analysis including IUD, IUD/ICW, IUD/ICA, IUD/BA, age and sex, 72 % of AD patients and 79 % of age-matched nondemented elderly subjects were correctly classified (Wilk's lambda 0.69, Chi-square 42.5, df 5, p<0.00001). This model explained only 31 % of the variance between groups. The best distinguishing variable, IUD/BA, explained 13 % of the variance, and the further contribution of other variables was less (IUD/ICW 8%, sex 4%, age 3% and IUD/ICA 2%). In the second analysis, including IUD/BA, age and sex, 63% of AD patients and 72% of nondemented elderly were correctly classified (Wilks' lambda 0.82, Chi-square 22.8, df 3, p<0.00001).
The mean IUD for AD patients with CDR questionable, mild and moderate dementia was 28.9 ± 1.7 mm, 28.7 ± 4.6 mm and 30.9 ±5.7 mm, respectively. The difference in standard or normalized IUD was not significant between the AD groups with differing clinical severity. Within the AD group, there was no significant correlation between MMSE test scores and IUD either (II).
Table III 1 presents the results of the relaxometry for AD, AAMI, OC and
YC groups. ANOVA showed significant differences in T2 of the right hippocampal
head (p<0.01) and tail (p<0.05) between AD and nondemented study
groups; AD patients differed from OC (head) and from OC and AAMI (tail).
For T2 of the right hippocampal head, which showed the most clear significant
difference, the 95 % confidence intervals of the means were 95.7-101.2
ms for AD, 90.6-96.8 ms for AAMI, 87.4-95.9 ms for OC, and 92.9-99.3 ms
for YC groups. The T2 of the amygdala did not differ between groups.
In AD patients, T2 of the left hippocampal head correlated significantly with MMSE scores (r=-0.44, p<0.01) (Figure III. 2). Table III. 2 and Figure III. 3 show that T2 in the hippocampal head and tail differed significantly between AD patients at CDR stages 0.5, 1 and 2, as well as OC. AD patients with moderate dementia differed both from OC and AD patients with mild disease. In AD patients, T2 did not correlate with memory test scores (data not shown).
The T2 of temporal white matter on the left was significantly prolonged both in AD and OC groups (p<0.05) and the parietal relaxation times in all elderly groups were prolonged compared to young controls (p<0.05). The T2 of the thalamus on both sides was also significantly longer in the AD and AAMI groups (p<0.05) than in the YC group. In AAMI, the T2 of the structures studied did not differ significantly from elderly controls.
The differences in T2 remained significant in ANOVA adjusted for a history of hypertension and coronary heart disease. No significant correlations between age and hippocampal T2 were found (Figure III 4). Thus, the prolongation of T2 in the hippocampal head and tail of AD patients was not explained by either age or by the presence of vascular diseases. In contrast, T2 in temporal and parietal white matter was significantly related to age in the whole study population, including nondemented subjects (AAMI, OC, YC), with r ranging from 0.27 to 0.45 (p<0.01) (Figure III 5). Therefore, the increased T2 in cortical white matter is explained by age rather than by the diagnostic category or by a history of hypertension or coronary heart disease (III).
In VaD, the T2 was prolonged in the left hippocampal head (106 ms, p<0.001) and in both the right and left tail (106 ms, p<0.05 compared to AD). T2 values in PD and PDD did not differ from nondemented controls.
The average T2 of cerebrospinal fluid (CSF) from fourteen measurements of the lateral ventricles was 2214 ± 544 ms.
The analysis of the incorrectly classified is based on the classification
of AD patients versus nondemented subjects by hippocampal volume. The main
results of the analysis are presented in Table IV 4. Correctly and incorrectly
classified AD patients did not differ significantly in age, sex, age at
onset, duration of disease, education, extrapyramidal signs assessed by
the Webster scale, depressive symptoms evaluated by the Hamilton scale,
ischemic scores, or frequency of a positive family history. As expected,
hippocampal volumes were significantly larger in the ADMC than in the ADCC
group (p<0.0001). The ADMC patients also had significantly higher MMSE
scores (p<0.05) as well as less severe memory impairment in tests assessing
delayed recall of the Heaton Visual Reproduction test (p<0.05) and the
story (p=0.0002) than ADCC patients. Otherwise the profile of cognitive
deficits was similar in both groups.
The incorrectly classified controls were significantly older than those correctly classified. They also showed smaller hippocampal volumes (p<0.0001), and more severe memory decline evident from the Buschke Selective test total (p<0.01) and long term scores (p<0.001). Performance in other cognitive domains was equivalent for these two groups.
Individual evaluation of follow up data on the nine incorrectly classified AD patients showed that only two of the patients could be considered to show merely typical features of AD. Of the other seven patients, one patient had experienced a vascular insult and possibly had VaD, one patient had strong extrapyramidal signs and may therefore represent atypical PD or Lewy body disease or variant of AD, two patients had a strong family history for AD but had very mild disease, scoring 26 and 27 points in the MMSE examination and showing no deterioration during the follow-up. Three patients exhibited frontal features. No follow up data for the older controls or AAMI subjects are available yet.
This study focused on MRI of the hippocampus, and its use in the diagnosis
of early AD. Hippocampal atrophy appears not to be a finding specific to
AD, but a sensitive finding in dementias with temporal lobe involvement.
The lack of hippocampal atrophy, however, appears to be highly accurate
for exclusion of AD, and perhaps other dementing conditions. Therefore
it can be concluded that this study provides strong evidence on behalf
of using hippocampal volumetry in the characterization of memory impairment.
Several approaches for the evaluation of hippocampal involvement were chosen, starting with volumetry, for which thin, contiguous, optimally oriented T1-weighted slices were used. This method is easily applicable, relatively quick to learn and use (approximately 10 minutes to obtain the volumes bilaterally), and provides substantial intra- and interrater accuracy. The importance of using thin slices cannot be overemphasized, since even with slices as thin as 3 mm, partial volume effect may cause a 30 % error in the volume of the hippocampus when coronal slices are used (Cook et al. 1992). Furthermore, various methods for normalization were tried. All measurements were performed blindly by a single rater and in a similar manner. The study subjects were well documented and their number is representative - by far the largest in a study of this method. Second, the sensitivity of volumetry was compared to that of a single measurement of hippocampal area, and a linear measurement, the IUD. Third, MR T2-relaxation times for head, body and tail of the hippocampus were measured. Moreover, sensitivity and specificity of hippocampal imaging were compared to those for the adjacent amygdala and frontal lobes.
The hippocampus was chosen as a ROI since it is known to be important in memory functioning (Squire and Zola Morgan 1991) and known to be affected and atrophied early in the course of AD (Hyman et al. 1984; 1990; 1995; Ball et al. 1985). By contrast, the volume of the hippocampus seems to remain practically unaffected by normal aging (Bhatia et al. 1993; DeCarli et al. 1994). Imaging of the hippocampus has been found to be reliable (Squire et al. 1990; Jack et al. 1992 a; b). Nevertheless, MRI studies of the hippocampus have produced mixed results in diagnosing AD.
All previous major findings were supported by this study: the hippocampus is atrophied early in AD; performance in several memory functions that are considered to be mediated by the hippocampus declined and correlated with hippocampal volume, whereas no correlations between hippocampal volume and tests assessing non-mnemonic performance were found. In nondemented subjects there was no significant correlation between hippocampal volume and age. This was the case despite the fact that the YC group can easily be referred to as "supernormal" control group; they were students of medicine and staff members, some having a Ph.D in medicine. It was also found that a short-cut in methodology, such as measurement of the hippocampal area or IUD on a single slice, is likely to produce an inaccurate result.
The hippocampus is atrophied already at an incipient clinical AD. The best
discriminative value of the volumetric hippocampal atrophy found in this
study, the overall accuracy of 87.1 92.0 %, demonstrates that MRI of the
hippocampus provides entirely additional data to support the clinical diagnosis
of AD versus nondemented subjects. Particularly, the specificity was high,
92.9 % at best. The sensitivity for the AD group, probably bargaining at
the expense of specificity, could be raised up to 94.4 % by taking into
account the overall brain atrophy. This analysis also yielded the best
overall accuracy of 92.0 %. It is important to notice that the hippocampal
volumes did not correlate to duration of AD but were atrophied in the AD
group having the disease last less than a year.
The measurement of the hippocampal area produced a correct overall classification of only 75 %. The IUD, using a cut-off point of 30 mm suggested in the literature, produced only a sensitivity of 37 % and a specificity of 72 %. The T2 was significantly prolonged in the right hippocampal head and tail and the prolongation was not explained by high age or presence of vascular disorders. The actual diagnostic value, however, was compromised by substantial overlap between the study groups.
It does seem that the use of volumetry is the one and only common denominator found in the studies that have ended up in a good discriminating accuracy (Kesslak et al. 1991; Jack et al. 1992 a.; Killiany et al. 1993; Lehéricy et al. 1994). No contradictory findings have been reported. Instead many simpler linear and planimetric measurements have not been as successful (LeMay et al. 1986; Cuénod et al. 1993; Early et al. 1993; Erkinjuntti et al. 1993; Howieson et al. 1993 b.). This trend is also supported by the experience of the CT studies (LeMay et al. 1986; DeCarli et al. 1990). Occasional studies, such as that of Seab et al. (1988), have provided better accuracy. Still, the average accuracy of linear measurements in distinguishing AD patients from controls seems to range from 65 to 75 %, as presented in this study as well.
It is presumable that simple measurements are far too much subject to individual variability, to variability caused by rotation of the head in every possible dimension, and are also more easily affected by various aberrations, such as artifacts or volume averaging. These sources of error are avoided only by measuring the whole ROI, preferably by using thin slices.
Normalization methods used in this study did not notably improve the accuracy. This is not completely unexpected, because the study subjects were well matched and their number was large enough to allow much of individual variance to vanish. The significance of proper normalization must, however, not be underestimated, but be considered as an important procedure for exclusion of normal variation and particularly differences due to gender. In the OC group, the volume of the hippocampus correlated to cranial area, and men do tend to have larger head size.
Normalization variables have varied a plenty in the literature. An often proclaimed normalization variable is intracranial volume, which logically seems most authoritative. Still, the demarcation of the intracranial volume is often merely stated and seldom defined in detail. Also, delineation of a bony structure is manditorially somewhat arbitrary, since compact bone is not properly visualized by MRI. This is true in this study as well, since at least the coronal intracranial area is drawn along the most clear signal, which is caused by subcutaneus adipose tissue. Therefore, the method includes calvarium, and the proper name for the variable might be the cranial area. However, the normalization is performed similarly for each and every study subject using not too simple, and therefore vulnerable, a variable. The sagittal ICA (ICA 2) has been previously considered as a valid method for normalization (Free et al. 1995). Judging by this study, the coronal ICA (ICA 1) is a valid method as well. The intracranial width emerges to be too vulnerable a method for accurate normalization. Normalization to BA deserves further attention. This method can not be regarded as a reliable variable for normalization of atrophic changes, since it is vulnerable to physiological, and neurodegenerative, atrophy in itself. Therefore an index between two atrophying, or preferably diminishing, structures might end up as a normal result or at least worse accuracy. Oppositely, normalization of atrophic changes in which there occurs dilatation, or growth, by BA might improve the accuracy. Still, normalization to BA might provide some information of the nature or amount of the atrophy of a certain structure compared to more generalized overall atrophy. As for the future, in order to obtain a vastly accepted imaging protocol for pooling comparable data from multicenter studies, and to define normal and pathological ranges for the hippocampal volumes, a commonly accepted imaging protocol, including proper normalization, is needed.
Analysis of the incorrectly classified is based on evaluation of study
subjects who were incorrectly classified by hippocampal volumetry. Curiosity
about the analysis of the incorrectly classified arose from the fact, that
any current clinical criteria can not be regarded as water-proof either
in separating AD from normal impairment along aging, not to mention non-AD
dementias (Erkinjuntti et al. 1986; Tierney et al. 1988; Kukull et al.
1990; Risse et al. 1990; Welsh et al. 1992; Almkvist and Bäckman 1993;
Blacker et al. 1994). The sensitivity and specificity of the NINCDS ADRDA
criteria have been reported to vary between kappa 0.83 0.92 and 0.65 0.84,
respectively (Kukull et al. 1990; Blacker et al. 1994). Thus, it is probable
that despite careful clinical diagnosis by the same raters, the study sample
includes patients grouped as AD but who are not demented, or then suffer
from a non AD dementia. Vice versa, it is also possible that the control
group includes subjects with subclinical dementia, or subjects who are
later bound to develop dementia. It is also possible that some patients
would represent heterogeneity of AD, in which the hippocampus remains relatively
unaffected by the disease.
To reexamine the presence of dementia, results of basic cognitive tests and tests assessing delayed recall were compared. As stated previously, the tests assessing delayed recall have been documented to be sensitive to dementia and spared by aging (Petersen et al. 1992; Welsh et al. 1992; Howieson et al. 1993 a.). In addition, the hippocampal volume has shown to correlate with performance in tests assessing delayed recall (Scheltens et al. 1992; Golomb et al. 1994 b.; Deweer et al. 1995). To examine the presence of possible subtypes of AD or non AD dementia, such as Pick's disease, frontal lobe dementia, Lewy body disease, or a dementia syndrome, some symptoms that may appear with AD but more typically are associated with other forms of dementia were compared (Gibb et al. 1985; Mayeux et al. 1985; Friedland et al. 1988; Neary et al. 1988; Reichman and Cummings 1990; Friedland 1993; Mendez et al. 1993; McKeith et al. 1994). All the AD patients that were incorrectly classified using volumetry were individually reappraised from follow up data. Of course, this analysis does not reveal or exclude mistakes in volumetry possibly made by "rater asleep at the console".
The ADMC group performed significantly better in the MMSE and tests assessing delayed recall, but their deficits in other cognitive domains were similar to those of ADCC. Individual reevaluation showed that seven out of nine AD patients to have features that relate to other forms of dementia or to have very mild AD. Thus, the true discriminative accuracy may be even higher than reported. This is particularly true in regard to false positives. All the epsilon 4 homozygote AD patients were correctly classified. Among the nondemented elderly there were seven epsilon 4 homozygotes, and they all were correctly classified as well. Follow-up data on controls and AAMI subjects was not available.
The raw and the normalized volumes of the right and the left amygdala in
AD patients compared to the combined OC/AAMI group were diminished 18.9/21.9
% and 21.0/23.6 %, respectively. Using these volumes a correct overall
classification of 74 % of these study subjects was achieved. No significant
correlation between the clinical severity assessed by MMSE and amygdaloid
volumes was found.
The MRI of the amygdala has also been a matter of controversy. The discrepancy in the magnitude of volume loss of the amygdala in this and previous studies may be explained by the less severe dementia with a relatively short duration, and narrowness of the clinical severity of the study group as well as probably by methodological differences in outlining the amygdala. Since the amygdala is a heterogeneous structure composed of numerous nuclei in both cortical and subcortical areas, its strict anatomic boundaries in MRIs have been found difficult to determine, and the demarcation of these borders has been somewhat arbitrary, often resulting in compromises (Pearlson et al. 1992; Bartzokis et al. 1993; Cuénod et al. 1993; Killiany et al. 1993; Lehéricy et al. 1994). This, in turn, may result in exclusion of nuclei responsible for changes in the volume. An exact delineation of the amygdala was considered difficult in this study as well: the reproducibility of the volume of hippocampus was found to be more reliable than that of the amygdala. Demarcation of the amygdala was considered particularly difficult in its anterior parts.
Histopathologic studies have indeniably demonstrated that the amygdala is damaged in AD, with pathological changes occurring mostly in the nuclei that receive or give rise to hippocampal or entorhinal projections (Kromer-Vogt et al. 1990; Scott et al. 1991; Mann 1992; Vereecken et al. 1994). The results of these post-mortem studies are not readily comparable with current findings. It is likely that the trend towards atrophy reflects the course of the disease process, proceeding from more affected regions, such as the entorhinal cortex and the hippocampus, towards the amygdala. Thus, judging by volumetric atrophy, the amygdala is not one of the primary sites of AD. Correlations between the amygdala and severity of the disease or memory functions were not found, supporting the current concept according to which the amygdala is not directly involved in memory functioning.
The volumes of the frontal lobes compared to nondemented controls were
diminished. Aging also contributed to the size. The decrease of the volumes
in percentage was a little less than that of amygdala, but the discriminating
accuracy was about similar. Thus, volumes of frontal lobes apparently reflect
merely enhanced and unspecific overall atrophy of AD brain. Also, the factor
contributing to the widespread volume loss appears to be present already
at the early stage of the disease, and the AD, at the level of the novel
symptoms, has already proceeded beyond the temporal pole. On the other
hand, aging itself does affect the volume of the frontal lobes more than
it does the temporal lobe (Cowell et al. 1994; DeCarli et al. 1994).
A significant prolongation of T2 in the right hippocampal head and tail
was found which could not be explained by either age or presence of vascular
disorders. In AD, the global clinical severity, but not the memory test
scores, correlated with the T2 of the left hippocampal head. The T2 was
prolonged in AD patients with moderate disease, whereas values for those
in the mild stage were comparable to those of controls. The T2 of amygdala
did not differ between the groups. In the neocortical white matter, significant
differences of the T2 were not related to AD, but were explained by high
age.
Even though differences in the T2 were found, the diagnostic value was compromised by major overlap in the study groups. No significant side-to-side differences were discovered, nor was there any segmental pattern of T2 values in the three anatomical regions of the hippocampus. The considered pathological limit of about 110 ms was exceeded in only 3 patients of the AD group and none of the controls. Those few could by statistics represent a rare dementia with hippocampal sclerosis as a primary pathologic finding (Zweig et al. 1989; Jellinger 1994). The mean T2 of all the regions of the hippocampus in normal controls groups was comparable to those reported by Kirsch et al. (1992) and Jackson et al. (1993).
In addition to the prolongation of T2 in the hippocampus reported by Kirsch et al., a visually increased hippocampal T2 signal in AD has been reported (Fazekas et al. 1987). The visually detectable high signal, however, is probably due to hippocampal sulcus remnant present in 39 % of people without temporal lobe pathology (Sasaki et al. 1993), or due to dilatation of choroidal and hippocampal fissures (George et al. 1990). When measured at 7 T in vitro, the T2 did not indicate the presence or severity of AD or vary between hippocampal subfields (Huesgen et al. 1993). T1 values measured at 0.08 T were reported to be similar for presenile AD patients and controls in the study of Christie et al. (1988).
The T2 (spin-spin or transverse relaxation) is dependent on numerous factors, such as observation frequency, temperature, mobility of observed spin, and presence of large molecules, paramagnetic ions and molecules, or other outside interference. In most tissues one component, usually water (or CSF), dominates the relaxation behavior. In the presence of two components with different relaxation properties, the quantitative interpretation is even more complicated (Rinck 1993 a.). Many pathologic conditions are known to produce alterations in factors that account for the observed relaxation behavior of tissues. The most important of these is the increased presence of water, in the region. There are many possible reasons for this such as hippocampal fissure and/or uncal sulcus, developmental cyst (Bronen and Cheung 1991 b,c; Sasaki et al. 1993), increased CSF in atrophying region, lacunae, and oedema.
Another common cause of prolonged T2 is considered to be gliosis or glioma (Bronen et al. 1991). In AD, the chronic inflammation resulting from the accumulation of amyloid is known to produce microgliosis and astrocytosis (McGeer et al. 1989; Delacourte 1990; Mandybur and Chuirazzi 1990; Ohgami et al. 1991). The T2 arising from both microgliosis and astrocytosis might be difficult to interpret: the signal of astrocytosis might increase the overall signal whereas T2 might be shortened due to iron in microglia, NFTs or SPs in the hippocampus (Leveugle et al. 1994; Antonini et al. 1993; Drayer 1988 a, b).
Further possibilities are normal variations between subjects and within one individual, and variations caused by anatomical position or machine drift (Harvey et al. 1991), pharmaceuticals (Karlik et al. 1986), hamartoma, vascular malformations, nonspecific calcification, scarring (Bronen et al. 1991), flow artifacts, volume averaging, observer variation, poor image quality, or differences in imaging parameters. It is also noteworthy that T2 of the hippocampal body did not show differences between the groups. This might be due to the fact that, when using proper tilting, the body does not curve. It may be possible, that when measuring more irregularly shaped atrophied head or tail, partial volume averaging of CSF might occur. However, since the underlying sources of the T2 are not yet determined, it is also possible that the T2 signal arises from factors that are entirely independent of those producing the atrophic changes.
The contradictory results of this study compared to the study of Kirsch et al. might be partly explained by larger sample size and the fact that most of the patients in the present study had mild dementia. Prolongation of the hippocampal T2 was not seen until in moderate dementia showing that the measurement of the hippocampal T2 does not help in diagnosing mild AD. Also, in the present study a 1.5 T imager was used compared to a 0.04 T imager used in the study by Kirsch et al. A 1.5 T imager has a better signal-to-noise ratio. Thus, it is possible that a low signal-to-noise ratio would result in noisy T2 maps. Consequently, it might be easier to avoid CSF in the measured ROI at high field. The field strength should not affect the interpretation of T2 as much as it would affect that of T1 (Rinck 1993 a.). Methodologically, the Carr-Purcell-Meiboom-Gill sequence used in the study is adequate (Rinck 1993 b.).
In this study, the overlap between the groups was considerable and it is concluded that the T2 relaxometry of the hippocampus or other structures studied here are of no use in diagnosing AD at its early stage.
Hippocampal atrophy is not a phenomenon completely specific for AD. There
are some conditions that must be considered when judging the atrophy.
Patients with Down's syndrome (DS) constantly develop dementia after 40 years of age. As in AD, CT studies have revealed overall brain atrophy with affection of the temporal lobe (Lai and Williams 1989; Schapiro et al. 1989). Also, MRI study has reported hippocampal atrophy in DS, with the atrophy being more profound in demented patients. The same studies also reported an unexplained enlargement of the parahippocampal gyrus in DS (Kesslak et al. 1994; Raz et al. 1995). Even though differentiation of dementia from a possible mental retardation is the cause of uncertainty, the dementia overrepresented in DS, and hippocampal atrophy, bears no dilemma for differential diagnosis, since it has been considered to be AD, and has even been mentioned in subtypes of AD in the NINCDS-ADRDA research criteria (McKhann et al. 1984). Patients with DS show similar clinical symptoms as do patients with AD (Brugge et al. 1994); in both diseases there are similar neuropathological (NTs, SPs, loss of hippocampal area) (Mann et al. 1990) and neurochemical findings (Godridge et al. 1987). Further, DS is due to trisomy of chromosome 21, which also harbors genes for Aß precursor and familial AD, and may therefore lead to increased Aß burden (Tanzi et al. 1987; St. George-Hyslop et al. 1987). Therefore, the hippocampal atrophy in DS can in fact be weighed as data that mostly supports the findings of the AD studies.
Recent study has also proposed the hippocampus being diminished in another genetic disorder, Turner's syndrome (Reiss et al. 1993). By contrast, in fragile X syndrome the hippocampus has been reported to be enlarged (Reiss et al. 1994). In that study the control group, however, consisted of subjects suffering from other developmental disorders, such as DS patients, and the average hippocampal volume was smaller than what is proposed in this study.
Hippocampal atrophy has also been reported in amnestic conditions due to temporal lobe pathology (Press et al. 1989; Kritchevsky and Squire 1993). On the other hand, the volume of the hippocampus has been reported to remain unaffected in non temporal amnesias, for example in Wernicke Korsakoff syndrome (Valenstein et al. 1987; Squire et al. 1990), and even in traumatic amnesia with anterior temporal lobe pathology (Kapur et al. 1992). Infections or inflammations, such as Herpes simplex encephalitis and paraneoplastic limbic encephalitis, that may cause amnesia or dementia by affecting the temporal lobe may also influence the hippocampal volume (Lacomis et al. 1990; Kapur et al. 1994; Yoneda et al. 1994). This in turn might have an effect on the hippocampal volumes in HIV dementia. Amnesias must be regarded as a differentially important category, since amnesia may initially be misdiagnosed as AD, or the first symptom for AD may be amnesia (Katzman 1986; Kritchevsky and Squire 1993).
The symptomology of schizophrenia may also include memory dysfunction, resembling that of classic amnestic syndrome, that is unattributable to motivation, co operation or side effects of the medication (Duffy and O'Carrol 1994). Also the volume of hippocampus may be affected by the pathology of schizophrenia. In the twin study of Suddath et al. (1990) 14 out of 15 monozygotic twins affected by schizophrenia had bilaterally diminished volumes of the hippocampus. Bogerts et al. (1993) found correlation between diminished volume of the hippocampus amygdala complex and positive psychotic symptoms. Despite significant atrophy of the complex, the study groups presented overlap in 75 % of the cases. Zipursky et al. (1994), on the other hand, found no differences in the hippocampal volumes between schizophrenics and controls. Even schitzophrenia may be associated with entrohinal pathology (Arnold et al. 1991). The volume of hippocampus has been reported to be spared in another differentially important neuropsychiatric concept, pseudo dementia caused by depression (Coffey et al. 1993; O'Brien et al. 1994). In dementia associated with multiple sclerosis, hippocampal plaques and demyelinization visible in MRI has been reported (Fontaine et al 1994; Tsolaki et al. 1994)
Apart from memory disorders, the best known condition in which hippocampal atrophy occurs, is temporal lobe epilepsy (TLE). In TLE, MRI volumetry has already found a place both in research and clinical decision making, in which detection of atrophy and signal abnormalities of the hippocampus have been found to be useful tools in the lateralization of the epileptogenic focus. In TLE the hippocampus is mostly atrophied unilaterally, and often accompanied with changes in morphological appearance (Cook et al. 1992; Jackson et al. 1993).
Further review for MRI of the hippocampus in cognitive dysfunction is restricted, since no data appears to be available. But, what about the other functions of the hippocampus? Does pathology in the olfactory system or of adequate response to stimuli affect the hippocampal volume? In Kallman's syndrome (hypogonadotrophic hypogonadism with congenital anosmia) hippocampal-amygdaloid volume has been reported to remain the same as with controls (Yousem et al. 1993). The stress, and the stress response, modulated by corticosteroids is a matter of controversy. Corticoid excess is known to cause cytotoxic hippocampal damage (Stein-Behrens et al. 1994). That excess has been reported to cause modest hippocampal atrophy detected by MRI (Starkman et al. 1992; Bremner et al. 1995), but an another study could not replicate the finding (Axelson et al. 1993).
6.6.1. MAPPING THE AGE-ASSOCIATED MEMORY IMPAIRMENT
On the basis of this study, AAMI can be separated from dementia by the hippocampal volumetry. In a previous study the volume of hippocampus in AAMI was spared, but the normal right left asymmetry was diminished (Soininen et al. 1994). The diminished asymmetry may imply that in some of the study subjects, the pathologic process of AD has begun. This sparing of the volume of the hippocampus in general, however, strongly supports the assumption that AAMI, by NIMH criteria, is related rather to normal aging than to dementia and cannot be located into the cognitive continuum. Besides the volume of the hippocampus, the OC and the AAMI groups did not significantly differ from each other in the volumes of the amygdala or the frontal lobes, in the length of the IUD or the hippocampal or amygdaloid T2. These findings gathered it is concluded that non-dementia memory impairment can be diagnosed and reliably differentiated from dementia by MRI methods, particularly by hippocampal volumetry. Given the high prevalence of AAMI (Barker et al. 1995; Koivisto et al. 1995) and substantial validity of the criteria (Hänninen et al. 1995), AAMI in terms of prognosis and differential diagnosis, in relation to AD, is more important than VaD.
6.6.2. VASCULAR DEMENTIAS
This study indicates that hippocampal atrophy appears not to be a specific finding to AD, but it also occurs in VaD and PDD as well as, to a lesser degree, in PD without dementia.
The patterns of atrophy in the VaD group are most interesting, but more difficult to explain. Four patients had bilateral hippocampal atrophy, two had no atrophy at all, and three had unilateral atrophy. In three out of the four patients with unilateral atrophy, the right hippocampus was larger. Other findings in T2-weighted axial MR images could not clarify the nature of these findings. No large or strategic infarcts or other lesions were found in the temporal lobes or in the area of the posterior cerebral artery, responsible for the circulation to the hippocampus. However, it is likely that these patterns of atrophy represent the results of different vascular lesions or etiologies. One might assume, that in those two VaD patients without atrophy the lesions causing the dementia do not affect the size of hippocampus at all. Likewise, in cases with unilateral atrophy, the lesions affect the hippocampus on only one side. The cases with bilateral atrophy have vascular lesions affecting the hippocampus bilaterally, or alternatively they may well represent mixed dementia, that is, the combination of AD and VaD. In a previous autopsy study, the presence of mixed dementia was surprisingly high: three out of four patients considered to have VaD had also AD, and 5/16 patients considered to have mixed dementia had AD alone (Wade et al. 1987).
These findings need definitely to be further surveyed, and the background of this asymmetry explained by a pathologic confirmation, since VaD is the differentially most important type of dementia for AD, in terms of both numbers and because there is potential for prevention and intervention for VaD.
6.3.3. PARKINSON'S DISEASE AND DEMENTIA
Major atrophy of the hippocampus in PDD and even PD is a bit of a surprise, and difficult to explain as well. Temporal lobe pathology plays apparently more important role in PD than previously emphasized. No temporal atrophy in MRI of the PD brain has been previously reported. Even though markers of AD pathology are known to be found in PD brains, hippocampus and entorhinal cortex (Jellinger 1987; Braak et al. 1993), the hippocampal atrophy was not primarily supposed to reach, and exceed, the atrophy observed in AD. This is because AD changes may be observed in nondemented aging as well (Bouras et al. 1993; Langui et al. 1995), but the MRI of hippocampus reveals practically no atrophy, not even sporadically, in nondemented elderly, several out of whom by statistics would be harboring landmarks of AD pathology under the microscope. This raises a question whether the volumetric atrophy is caused by factors other than the classical AD landmarks. The T2 of the hippocampus did not differ between the PD groups and the nondemented controls.
The number of PD subjects in this study is restricted, and a possible coexistence of AD cannot be excluded in PDD patients, but still, the trend towards atrophy in PD patients indicates profound pathologic process to take place in the temporal lobe. The pattern of the hallmarks of AD in the PD hippocampus might be different from that in AD. The amygdala is known to be affected in PD as well, but the pattern of affection and atrophy is not similar to that observed in AD (Braak et al. 1994). More sophisticated MRI techniques than plain volumetry would be probably required to reveal a possibly distinct pattern of atrophy.
1. Hippocampal atrophy is a highly sensitive indicator of
incipient AD. It's evaluation by MRI volumetry provides entirely additional
data to support the diagnosis. In contrast, simpler measurements or T2
relaxometry provided only little diagnostic aid. The volume of the hippocampus
correlated significantly with clinical severity assessed by MMSE, and with
tests assessing delayed recall, which is considered to be affected early
in dementia and may reflect hippocampal dysfunction.
2. The volume of the hippocampus is not significantly affected by normal aging.
3. AAMI can be differentiated from dementia by hippocampal volumetry. Judging by hippocampal volumes AAMI is a phenomenon of normal aging rather than an intermediate in the continuum from normal cognition to dementia.
4. Hippocampal atrophy may not be a phenomenon specific to dementia in AD, but a common feature of dementias with temporal lobe pathology. Hippocampal atrophy was also found in VaD and PD as well. In PD, temporal lobe pathology appears to be present more often and more severe than usually described. The hippocampus was atrophied bilaterally in some, but not all, VaD cases, and displayed unilateral pattern in some. Since VaD is most important type of dementia to be differentiated from AD, this phenomenon will need some further attention in future studies.
5. Volumetry of the amygdala or frontal lobes provide little substantial data to help with MRI diagnosis.
All and all, hippocampal volumetry is useful in excluding AD and separating it from benign, (non-parkinsonian) age-associated memory impairment.
|
___________
|