close

Вход

Забыли?

вход по аккаунту

?

Brain electrical activity in patients with presenile and senile dementia of the Alzheimer type.

код для вставкиСкачать
Brain Electrical Activity in
Patients with Presede and Sede
Dementia of the Alzheimer Type
Frank H. Duffy, MD,' Marilyn S. Albert, PhD,t and Gloria McAnulty, PhD*
Neurophysiological and behavioral data obtained from 9 patients with presenile dementia and 10 with senile dementia
of the Alzheimer type were compared with similar data from 25 age- and sex-equivalent controls. Compared with the
healthy controls, both patient groups demonstrated increased background electroencephalographic slowing with a
reduction in fast activity (synchronization). Topographic analyses of data from electroencephalographic and evoked
potential studies indicate that areas of maximal group differences between the presenile patients and their controls
include the right posterior temporal and, to a lesser extent, left midtemporal to anterior temporal areas, whereas the
maximal differences between the senile patient group and their controls involve the midfrontal and anterior frontal
lobes, bilaterally. Moreover, right-sided numerical features derived from topographic maps proved most useful in
differentiating the presenile patients and their age-matched controls, whereas bilateral features were more useful in
separating senile patients from their controls. These topographic dissimilarities between patient groups suggest that an
age-disease interaction exists between patients with presenile and senile dementia of the Alzheimer type. Correlational
analyses between neuropsychological test scores and neurophysiological features indicate that increased slowing and
decreased fast activity were associated with poorer test performance.
Duffy FH, Albert MS, McAnulty G: Brain electrical activity in patients with presenile and senile dementia of
the Alzheimer type. Ann Neurol 16:439-448. 1984
Electroencephalographic (EEG) and evoked potential
(EP) studies have demonstrated characteristic changes
in patients with presenile (PSD) and senile (SD) dementia of the Alzheimer type. Increased background
EEG delta and theta slow activity, accompanied by diminished or absent alpha, are the most commonly reported EEG findings E8, 231. The late components of
the EP show a prolonged latency and a reduced amplitude, especially the P300 event-related component
E22, 37, 381. However, these findings are based
primarily on the examination of moderately to severely
advanced patients.
The prevalence of EEG and EP abnormalities in patients with mild to moderate dementia is unclear. In
fact, the EEGs of patients in the early stages of the
disease are often read as within normal limits for age
c141. It is possible, however, that the basic EEG and EP
measures of such patients with PSD and SD contain
more information than can be easily demonstrated by
simple visual inspection. Because the cognitive abilities
of mildly to moderately impaired patients are clearly
reduced, it seems likely that more detailed EEG and EP
analyses would assist in the demonstration of clinically
meaningful differences.
A topographic mapping method called brain electrical activity mapping (BEAM*) C13, 15, 17-191 has
been designed to enhance the visibility of this complex
information. This method permits the investigator to
summarize EEG and EP data for groups of subjects and
to develop statistical maps ll51 that localize the maximal regional differences between the groups. From
these maps of group difference, numerical features can
be developed to differentiate the groups under study
From the *Seizure Unit and Developmental Neurophysiology Laboratory, Department of Neurology, Children's Hospital and Harvard
Medical School, 300 Longwood Ave, Boston, MA 021 15 and the
tDepartments of Psychiatry and Neurology, Massachusetts General
Hospital, Division on Aging, Harvard Medical School, Boston, MA
02114.
Received June 6, 1983, and in revised form Dec 12 and Dec 27.
Accepted for publication Dec 30, 1983.
cis, 18, 191.
In the present investigation patients with Alzheimer's disease who were otherwise free of systemic illness
were compared with optimally healthy age- and sexequivalent controls. The patients under 65 years of age
were examined separately from those 65 years of age
and over to ensure that age-disease interactions would
not obscure significant differences between the subjects and their controls. It was hoped that topographic
"BEAM is a registered trademark of Braintech, Inc.
Address reprint requests to Dr Duffy,
439
mapping would identify regions of difference between
patients and controls that might be used to assist the
diagnosis of patients with PSD and SD. Because at
present there is no specific noninvasive test that can be
used to diagnose Alzheimer’s disease, such an outcome
could have considerable value.
Methods
The subjects were divided into two age strata, those 40 to 64
and those 65 to 80 years of age. Within each stratum patients
with dementia of the Alzheimer type were compared with
optimally healthy controls.
The younger group consisted of 9 patients with PSD of the
Alzheimer type (4 women and 5 men) and 15 age-matched
controls (7 women and 8 men). O f the 9 patients with PSD, 8
were right handed and one left handed. Thirteen of the normal controls were right handed, and 2 were left handed. The
mean age of the patients with PSD was 58.3 years, and that of
the controls was 57.8 years. There was no significant difference between the ages of the controls and those of the demented patients.
The older group consisted of 10 patients with S D of the
Alzheimer type (8 women and 2 men) and 10 age-matched
controls (6 women and 4 men). The patients with S D and the
controls were all right handed. The mean ages of the patients
with SD and the controls were 70.3 and 70.2 years, respectively, a nonsignificant difference.
The diagnosis of Alzheimer’s disease was based on the
judgment of a staff neurologist, with independent agreement
from the staff psychiatrist and neuropsychologist. Medical
conditions known to produce dementia were excluded. A
large number of laboratory tests (e.g., computed tomographic
scan, lumbar puncture, SMA-12 analysis, total thyroxine, and
folate levels) were conducted to rule out various hydrocephalic, metabolic, neoplastic, infectious, and traumatic causes
for dementia. Patients with a record of severe head trauma,
alcoholism, serious psychiatric illness, learning disabilities,
epilepsy, hypertension, lung disease, kidney disease, coronary artery disease, or cancer were excluded. All patients
with PSD and SD received an ischemic score of 4 or less on
Hachinski’s scale for ruling out multi-infarct dementia [241.
Retrospective studies suggest that approximately 90% of patients classified by these procedures are correctly diagnosed
as having Alzheimer’s disease [ l l , 41).
Only subjects with a score of 30 or below on an activities
of daily living scale {40) were admitted to the study, to prevent the inclusion of patients in moderately advanced stages
of the disease. All patients were studied when they were
living at home.
The normal male controls were members of the Normative Aging Study at the Boston Veterans Administration
Outpatient Clinic. Complete medical records were available
on these individuals for the previous 18 years. Control subjects were screened and excluded from further investigation
based on the same criteria as were used with the patients with
PSD and SD. The female controls were community volunteers who were screened in a like manner. The male and
female controls therefore represented a population of subjects who were as free of systemic disease as were the patients
with PSD and SD.
440
Annals of Neurology
Vol 16 N o
4 October 1984
Neurophysiologicaf Testing
The brain electrical activity data of all 44 subjects were obtained and analyzed in the Seizure Unit of Children’s Hospital, Boston, Massachusetts. EEG data were gathered during
ten different behavioral states. These included resting with
the eyes open (EOP), resting with the eyes closed (ECL),
listening to speech (SPE) and music (MUS), memorizing
(KFI) and being tested (KFT) on the recall of geometric
shapes, reading paragraphs (RTI) and being tested (R‘IT) on
the recall of specific sentences from the paragraphs, and
forming (PAI) and being tested on (PAT) sound-symbol
associations. The activating paradigms were developed for a
study of reading disability 1191 but have proved useful in
many clinical areas, including the studies in our report on
normal aging in this issue [14J The tests are described in
greater detail in that report.
EPs were derived during three states, flash visual evoked
potential (VEP), tone-pip auditory evoked potential (AEP),
and the P300, or event-related potential paradigm using highand low-frequency auditory stimuli (P300). The last of these
employed the Goodin paradigm [2 1,221. These EP states are
further detailed in the normative aging report [14].
Efectrophysiof o g i d Data Analyses
Data were obtained from twenty scalp and two eye electrodes
referenced to linked ears and were recorded on magnetic
tape along with appropriate stimulus and state markers for
subsequent off-line data analysis. Off-line data processing
was controlled via the SIGSYS biomedical software package
(Braintech) and employed spectral analysis for the EEG states
and signal averaging for the EP states. Thus, for each state the
end product was either twenty spectra or twenty EPs (one
from each electrode). Considerable care was taken to minimize artifact at the time of data gathering and to recognize
and eliminate it during data analysis. Details of these procedures have been previously discussed [14, 17).
Sets of twenty spectra or EPs were transformed into a
corresponding series of topographic maps of brain electrical
activity using the BEAM methodology [13, 171. Thus, scalp
areas surrounding the twenty real electrode locations were
assigned numerical values by linear interpolation based on
the values at the three nearest electrodes. The resulting 64 X
64 numerical matrix was maintained for statistical analysis and
used as the basis for the computerized color-coded topographic display (see Fig 1). For EEG data each image represented the topographic distribution of activities in one 4 H z
wide frequency band. Eight such bands, from 0.5 to 32 Hz,
were visualized, beginning with delta (0.5 to 3.75 Hz) and
sequentially passing through theta (4 to 7.75 Hz), alpha (8 to
11.75 Hz), beta 1 (12 to 15.75 Hz), beta2 (16 to 19.75 Hz),
beta 3 (20 to 23.75 Hz), beta 4 (24 to 27.75 Hz), and beta 5
(28 to 31.75 Hz). Spectral values for any given band were
also expressed as a percentage of total energy, a nornialization process. For EP data each image represented the topographic distribution of actual EP voltage during a single 4 ms
epoch following stimulation. Over the 5 12 ms period following stimulation, 128 such images were formed.
To delineate regional differences between groups, the
significance probability mapping (SPM) technique was used
C151. This process involves forming a new image in which the
raw spectral or EP data are replaced by a statistical measure of
distance beween the groups being compared. In the present
investigation such comparisons were made for each state between the PSD and PSDCON groups, and also between the
SD and SDCON groups, via the Student’s t statistic (see Figs
1 and 3).
Last, single numerical measures or features descriptive of
each subject were derived from the SPM-defined regions of
between-group difference. Regions delineated by a t value
above a certain criterion level were used as templates. Single
values for each subject were created by integrating values
underlying a template.
Once features were generated, a series of statistical tests
were applied to them that allowed assessment of their relative
“merit.” Feature generation, selection, and statistical analysis
employed the TICAS software package 12-51. (The TICAS
program names are indicated in capital letters, e.g.,
FMTEST.)
To assess the collective value of several features in multivariate feature space and to identify the features most useful
for this purpose, standard techniques of multivariate analysis
were employed [1-4,9}. As a first step, features achieving or
surpassing the 0.02 level by the nonparametric univariate
Mann-Whitney U test [29] were submitted to a merit value
(MV) analysis (FMTEST). An MV analysis tests the ability of
each individual feature to “detect” subject identity. Two common statistical measures of diagnostic value were used, the
“measure of detectability” [2, 34, 351 and an “ambiguity function” [2, 201. MV analysis also includes a measure of how
uniquely valuable (i.e., nonredundant) each function is, derived by measurement of its average correlation with all other
selected features. Features with low MV value (i.e., low detectability and high average correlation) are generally discarded. This allows selection of features most likely to be of
use in classification procedures and identification of the most
useful template regions.
As a collective test of the value of features in combination,
stepwise discriminant analysis was performed, with subjects
represented by the best features from MV analysis. The degree and significance of the separation between groups was
estimated by Wilks’s lambda [43]. The success of retrospective individual subject classification was determined by
using the discriminant function as a classifier.
4
Neuropsycbological Testing
All subjects were administered a large battery of neuropsychological tests chosen to span the major aspects of cognitive function. The tests included tests of auditory and visual
attention, the Boston Naming Test, the Wechsler Memory
Scale, a word-list learning test, the Block Design subtest of
the Wechsler Adult Intelligence Scale (WAIS), clock drawing, proverb interpretation, competing motor programs, the
Visual-Verbal Test, and a test of verbal fluency. Testing time
for the controls was approximately 3 hours. Because the
length of testing was considerably longer for patients with
Alzheimer’s disease (approximately 4 % hours), test administration was spread over more than one day. The order of
administration was varied to prevent order effects.
To simplify interpretation of the correlational analysis, results of all tests except the Boston Naming Test, proverb
interpretation, and test of verbal fluency were grouped into
four categories on the basis of a priori theoretical judgment as
follows: (1)verbal memory tasks: the word-list learning task,
the Logical Memory subtest of the Wechsler Memory scale,
and the Paired Associate subtest of the same scale; (2) nonverbal memory tasks: the Visual Reproduction subtest of the
Wechsler Memory scale; (3) visuospatial tasks: the Block Design subtest of the WAIS, the copy condition of the Wechsler
Memory Scale Visual Reproduction subtest, and clock drawing; ( 4 ) attention and abstraction abilities: competing motor
programs, the Visual-Verbal Test, and tests of auditory and
visual attention. In all cases higher scores indicate better performance on the neuropsychological task.
Results
Presenile Dementia
TOdelineate regions in which patients with PSD differ
from age-matched controls, these two groups were
compared using the t statistic SPM technique (t-SPM).
Sixty-three cortical areas were identified that showed
sizable and relatively contiguous regions of betweengroup difference. These regions acted as templates or
masks for the development of features. Cortical activity
was summated in the areas underlying the templates in
each subject to produce a single numerical value, which
served as the feature to be used in subsequent analysis.
Because the spectral data were always analyzed in two
ways (raw microvolt values and percentage of total
spectral energy normalized data), there were several
instances in which regional between-group differences
were seen when a given SPM spectral band was viewed
as both raw and normalized data. In those cases, only
the data type showing the larger region andor higher t
value was chosen for further analysis. This ensured that
only operationally independent templates were used
for feature generation.
Forty-eight independent features were generated in
this fashion and subjected to the Mann-Whitney U
test. All forty-eight were significant at the 0.05 level
(two tailed), and twenty-six at the 0.02 level (two
tailed) or better.
Only twenty-six features significant at the 0.02 level
or better were included in subsequent analyses. Although strict operational independence was achieved
by all of the twenty-six features utilized, a number of
templates seemed very closely related. For example,
similar bilateral parietal regions were identified for 4 to
8 Hz raw EEG theta during both MUS and PAL In
addition, a right posterior quadrant region was defined
for 4 to 8 Hz normalized EEG during EOP, ECL, SPE,
and MUS. Accordingly, two new summary features
were created, consisting of the linear combinations of
the features from the topographically analogous regions (AFEA 25C and AFEA 26C). Because of the
addition of these combinations, there were twenty-two
final features.
T h e features will be discussed in terms of (1) the
cortical regions they define, ( 2 ) the behavioral state
Duffy et al: Brain Activity in Dementia
441
Fig I. Seven BEAM images obtained by comparison of the patients with presenile
dementia and their controls. The state from which each t statistic signzjicance
probability mapping ( t S P M )was derived is indicated below each image. The
maximum t value in each image is shown as T M . The t cutoff value used to form
templates for feature development is shown in parentheses. The name of the feature
derived from each t-SPM is also shown. The four best features by final merit value
are, in order, AFEA 26C (AFEA 26C is the linear combination of AFEA 9,
AFEA 10, AFEA 11, andAFEA 12;seetext), BFEA4, BFEA22,and
AFEA 14. Allfour are shown in this figure. Note that the differences between
patients and controls involve the temporal lobes prominently, especially on the right.
Fig 3. Four BEAM images obtained by a comparison of the patients with senile
dementia and their controls. The display convention is the same as in Figure 1.
Thefour best features by final merit value are shown: CFEA 7 , CFEA 1, CFEA
8, and CFEA 24. Note that the differences between patients and controls involve
thefrontal lobes prominently and also the occipital pole.
442
Annals of Neurology
Vol 16
No 4
October 1984
under which they were generated, (3) the frequency
band of the data that were compared, and (4)whether
raw or normalized data were used.
Of the original twenty-six features, seven were exclusively right sided, fourteen were predominantly
right sided, two were equally bilateral, and three were
exclusively left sided. Thus, twenty-one of twenty-six
(80.8p;) were exclusively or predominantly right sided.
The EEG recordings made during activation tasks
provided information beyond that obtained in the simple EOP and ECL states. Representing the results of
spectral analysis as a percentage of total spectral energy
rather than as simple raw spectral values generally produced more significant features. All group differences
in the theta range represented increased theta for the
PSD group. Beta activity from 12 to 28 Hz was decreased for the PSD group (in both raw and normalized
data). No features were developed from the alpha or
delta frequency bands. Fewer features were derived
from EP than from EEG data. Four of the five EP
features were derived from the VEP and one from the
AEP. The latter, however, was the best EP feature by
MV analysis. No between-group differences were seen
for the P300 event-related potential.
T o preselect features on the basis of their potential
value in subsequent multivariate analyses, the rwentytwo final features were submitted to MV analysis. The
intermediate merit value (IMV) represents an independent assessment of each feature considered singly. Features with values of 0.5 or greater are likely to be of
value. The final merit value (FMV) allows a ranking of
the most useful features, taking into account the degree
of intercorrelation among the features.
The best feature by FMV analysis was the fourtemplate combination feature AFEA 26C, which was
derived primarily from the right posterior quadrant for
4 to 8 Hz normalized theta. The second-best feature
was BFEA 4 , derived from the bilateral posterior
parietotemporal regions (right more than left) in the 16
to 20 Hz normalized beta 2 range. The third-best feature, BFEA 22, was derived broadly from the left anterior temporal region of the AEP in the 84 to 104 ms
latency epoch. The fourth-best feature, AFEA 14, was
derived from the left midtemporal region during
geometric figure recall testing in the 4 to 8 Hz normalized theta range. The templates from which these
features were derived are shown in Figure 1.
The ability of these four features to define group
membership was determined by stepwise discriminant
function analysis. The separation between the control
and PSD subjects was assessed by Wilks’s lambda,
which reached 0.294,significant at the p < 0.002 level
by Rao’s approximation {3 31. Retrospective subject
classification by discriminant function analysis demonstrated correct classification of all 15 control subjects
and 8 of the 9 PSD subjects (95.8% correct clas-
0
”I
t
!
.+
!
*
B
?
E
A
19
AF I A
26
Fig 2. Presenile dementia group confidence region plot, illustrating the separation between patients with presenile dementia
(PSD) and their controls (PSDCON) on the basis of two uncoruelatedfeatures.The vertical axis is BFEA 19, a iiisual evoked
responsefeature, and the horizontal axis is AFEA 26C,an electroencephalographicfeature. The centroid of each population ir
shown by a cross. The inner hatched ellipses represent the 93%
confidence region of the meansfor each group. Their nonintwsecting nature rejects a significant group separation. The outer
ellipse represents the 30% tolerance regionr the sector within
which SO% of the population would be expected to fall. The asterisk represent the Bayesian “decisionboundary” (i.e.,the curve
along which the ewor of subject aJsignment to either population is
equalized). This figure was created by TICAS program
BAYPLT. For this plot, data were autoscaLed.
sification). The ability to separate PSD subjects from
their controls on the basis of just two features (BFEA
19 and AFEA 26C) is shown in Figure 2.
Correlational analyses were performed to assess the
relationship between measures of cognitive function in
the patients with PSD and topographic features. Table
1 shows the correlations between the four highestranked features (by FMV) and test scores from the
neuropsychological test battery. The test scores have
been grouped into major cognitive skill areas (i.e., verbal memory, nonverbal memory, visuospatial ability,
and attentiodabstraction skills). Because these general
categories comprise several tests, a number in parentheses has been used to indicate that more than one
test of each skill was significantly correlated with the
designated feature. In all cases in which parenthetical
notations are used, there was directional consistency of
the correlations between the cognitive measure and the
Duffy et al: Brain Activity in Dementia 443
Table 1 . Corvekztion between Neuropsycbological Test Scores and BEAM Features in the Patzents wzth Prelenile Dementia
Verbal Memory
BEAM
Nonverbal Memory
Attention/
Abstraction Skills
Visuospatial Ability
Feature
r
P
r
P
r
P
r
P
AFEA
BFEA
BFEA
AFEA
-0.78
+0.80
...
...
0.0005 (3)"
0.0004 ( 5 )
...
...
-0.70
+0.78
0.002 ( 2 )
0.003
0.01 ( 2 )
0.01
-0.70
0.003 ( 2 )
0.0004 ( 2 )
...
...
-0.67
+0.71
0.006
0.002 (4)
...
0.01
-
26C
4
22
14
-0.68
-0.67
4-0.79
...
...
...
-0.65
"Numbers in parentheses indicate the number of significant correlations in that category. (Significancewas defined as a p value of 0.01 or better.)
In each instance t h e example in the table represents the highest correlation obtained.
neurophysiological feature. For example, BFEA 4 was
significantly correlated with five tests of verbal memory. In each case cognitive performance improved as
the value of the feature increased. Twenty-four correlations between the cognitive measures and neurophysiological features reached significance at the 0.01 level
or better. The two highest-ranked features (AFEA
26C, the combination feature developed from resting
and activated theta, and BFEA 4 , beta speech activation) were correlated with tasks that spanned the cognitive skill areas. In both cases the primary regions involved were in the right posterior temporoparietal
quadrant .
All correlations between the neuropsychological test
scores and the theta features were negative, whereas
the test measures were positively related to the beta
range features. Higher neuropsychological test scores
indicate better performance; thus, more adequate test
performance is associated with less theta and more beta
activity. Performance on two tasks of nonverbal memory was negatively correlated with the AEP feature,
whereas results of tests of nonverbal memory and attentiodabstraction were negatively correlated with the
theta feature.
Senile Dementia
The comparison of the patients with SD to their agematched controls by means of t-SPM identified fiftyone contiguous regions in which group differences exceeded t values of 2.0. As in the analysis of the younger
groups, these data were reduced to represent only
those regions that had been identified by operationally
independent comparisons. Thirty-one regions were selected to serve as templates for feature generation. The
resulting features were analyzed by UTEST. All were
significant at or beyond the 0.02 level (two tailed).
Of the thirty-one regions in which the patients with
SD differed from controls, c) were predominantly right
sided, 12 were equally bilateral, and 10 were predominantly or exclusively left sided. Thus, there was no
hemispheric bias. Only one feature was developed
from 8 to 12 Hz alpha, with thirteen developed from
slow activities (delta and theta, 0 to 8 Hz) and fourteen
444 Annals of Neurology
Vol 16 No 4 O c t o b e r 1984
from fast activities (beta, 12 to 28 Hz). The activation
paradigms provided information beyond that seen for
the EOP and ECL states. In the group differences, slow
activity was increased and fast activity decreased for the
SD group in every instance. This was true whether data
were expressed' as normalized or as raw power.
Three features were derived from the EP data, one
from the AEP and two from the P300. The feature
derived from the midlatency period of the AEP was
increased compared with control values, but the late
P300 waves were decreased for the SD group. In general, the EP features were not as significant as the EEG
features.
The potential discriminating value of the features
was determined by MV analysis. The three best features by FMV (CFEA 7, CFEA 1, and CFEA 8) all
represented increased 0 to 4 or 4 to 8 Hz EEG slowing
in the frontal lobes bilaterally. The fourth-best feature,
CFEA 24, involved increased 4 to 8 Hz theta in the
bilateral occipital region. The templates used to form
these features are shown in Figure 3.
The ability of the four highest-ranked features to
define group membership was evaluated by discriminant function analysis. The separation between the SD
and SD control groups in multivariate feature space
was significant at the 0.0002 level (Wilks's lambda =
0.250). Nine of the 10 subjects in each group were
correctly classified by retrospective discriminant function analysis (90% correct classification).The ability to
separate subjects with SD from their controls on the
basis of just two features (DFEA 22 and CFEA 7) is
shown in Figure 4 (BAYPLT).
To assess the relationship between cognitive performance and regions of neurophysiological difference,
correlational analyses were performed between the
four highest-ranking feature measures (based on MV
analysis) and test scores from the neuropsychological
battery. Twenty-one correlations reached significance
at probability levels of 0.01 or better (Table 2). All
correlations were negative, indicating that better cognitive test performance was demonstrated by individuals
with less delta or theta in the regions delineated by the
SPMs (see Fig 3). For the two highest features, CFEA 1
have EEGs that differ from those of age- and sexmatched controls.
An evaluation of the features derived from the tSPM analysis revealed that the right posterior temporal
region and, to some extent, the left midtemporal to
anterior temporal region were most useful in discriminating the patients with PSD from controls (see
Fig 1). Although these regions are similar to those
identified as important markers of age-related electrophysiological change in the report on normal aging
C141, the nature of the EEG and EP changes is entirely
different. For the PSD group theta slow activity increased and fast activity decreased relative to controls,
whereas the opposite occurred in healthy subjects
across the age range. Thus, the PSD and normal aging studies both reveal significant electrophysiological
changes in the parietal and temporal lobes of both
hemispheres, but normal aging is associated with desynchronization (ie., decreased slowing and increased
fast activity) whereas PSD is associated with synchronization (i.e., increased slowing and decreased fast activity). In this sense the electrophysiological changes of
PSD are in the opposite directLon from those of normal
aging.
The most useful regions for discriminating the SD
group from controls were the midfrontal and anterior
frontal lobes, symmetrically (see Fig 3). In these frontal
areas delta and theta were augmented and beta decreased. It should be noted that Picks disease is extremely uncommon among dementing patients in the
senile age range [26}. Furthermore, patients with extreme personality changes were excluded from the
study. Thus, the predominance of frontal lobe EEG
features among the patients with SD is unlikely to be
the result of the inclusion of patients with Pick‘s disease. Eye blinks are also an unlikely cause of the groupspecific difference, because great care was taken to
eliminate such artifacts. The relative synchronization
found in patients with SD resembles that observed in
patients with PSD of the Alzheimer type. These SD
data, together with the previous PSD findings, fad
to support the notion that dementia represents accelerated normal aging.
* I
*
t
*
**
I
d
+
I,
t
I
t
C P E A ?
Fig 4. Senile dementia group confidence region plot, demonstrating the separation between patients with senile dementia (SD)
and their controls (SDCON) on the basis of two uncorrelated
features. The vertical axis is DFEA 22, a P300feature, and the
horizontal axis is CF E A 7 , an electroencephalographicfeature.
The display convention is identical to that in Figure 2. Note the
total lack of overlap between the mean confidence inner ellipses
and the group tolerance outer ellipses.
and 7, significant correlations were shown to span the
cognitive test categories, which have been grouped as
in Table 1. Verbal memory and visuospatial ability
were negatively correlated with CFEA 8, a measure of
frontal delta, whereas memory, both verbal and nonverbal, was negatively correlated with the occipital
delta feature CFEA 24.
Discussion
Exploratory analysis using topographic mapping techniques for EEG and EP data indicates that subjects in
the early stages of PSD and SD of the Alzheimer type
Table 2. Correlation between Neuropsychological Test Scores and BEAM Features in the Patients with Senile Dementia
Verbal Memory
BEAM
Feature
CFEA
CFEA
CFEA
CFEA
Y
- 0.85
7
1
8
-0.77
- 0.85
-0.77
24
~~~~~
Nonverbal Memory
Attention‘
Abstraction Skills
Visuospatial Ability
P
r
P
r
P
r
P
0.0005
0.003 (2)
0.001 ( 3 )
0.007 (7)
- 0.82
- 0.72
0.003
0.003 (2)
-0.78
-0.78
...
...
...
-0.71
- 0.88
0.002
...
0.003 (2)”
0.002 (2)
0.001
...
...
0.009
...
...
- 0.80
...
...
~
“Numbers in parentheses indicate the number of Significant correlatiotls in that category (Significance was defined as a p value o f 0 01 or better )
In each instance the example in rhe table represents the highest correlation obtamed
Duffy et al: Brain Activity in Dementia
445
The topographic dissimilarities between subjects
with PSD and those with SD suggest important differences between the groups. Although both patient
groups demonstrate synchronization (i.e., increased
slowing), the topographic distribution of these changes
differs. In the older age range, the regions that best
discriminate the demented patients from controls are
in the frontal lobes; in the younger age range, temporal
lobe features are powerful discriminators. Thus, an
age-disease interaction appears to be taking place. One
can speculate that as individuals grow older the frontal
lobes become more vulnerable to the effects of disease.
Therefore, older patients show the greatest difference
from their age-matched controls in frontal regions,
whereas patients in the presenile age range are maximally different in the temporal lobes.
It is more difficult to interpret the meaning of
changes in the EP data than the meaning of those in the
EEG data. Most VEPs and AEPs in the demented
population showed striking alterations in waveform
structure, making localization of standard component
landmarks difficult. Nonetheless, the data do demonstrate regional differences consistent with the EEG
findings and support the notion of broad electrophysiological differences between patients and controls.
Although the specific origin of the EEG slowing described in the present study cannot be established with
certainty, evidence indicates that it relates to changes
in cholinergic activity. It is well known that anricholinergic drugs, such as atropine, produce prominent
EEG slowing 1301. Moreover, Testa and Gloor C391
have postulated that delta activity reflects the relative
cholinergic deafferentation of the cortex. The EEG
changes seen in patients with PSD and with SD may
therefore be the result of decreased regional brain cholinergic activity. Such a relationship would be consistent with the findings reported in patients with PSD
and SD of decreased amounts of choline acetyltransferase in the cortex on postmortem examination [ 10, 111
and decreased synthesis of acetylcholine in cortical
biopsy tissue 136). The nucleus basalis of Meynert, a
region of the basal forebram that contains cholinergicpositive neurons that project to widespread areas of the
cerebral cortex, is also reported to show considerable
neuronal loss in patients with Alzheimer-type dementia 1311. Moreover, there is a parallel between the location of the EEG features that best discriminate patients
with PSD and with SD from controls and the location
of the decreases in cortical choline acetyltransferase
found in these patients postmortem. Among 49- to 68year-old patients, decreases in choline acetyltransferase
are reported to be greatest in temporoparietal regions,
whereas patients aged 76 to 93 years show the greatest
decreases in frontoparietal areas [ 11).
The temporal lobe EEG desynchronization seen
with normal aging 1141 is, by analogy, consistent with
446 Annals of Neurology
Vol 16 No 4
October 1984
cholinergic augmentation. Recent data indicate that
anticholinesterase treatment in monkeys {7, 161 and
humans 116) produces long-lasting temporal lobe desynchronization. It is relevant here that the nucleus
basalis of Meynert is relatively spared during normal
aging {42), whereas there is an age-related loss of
neurons in the temporal cortex 161. The loss of noncholinergic cortical neurons coupled with the relative
preservation of cholinergic neurons (a pattern suggested by the sparing of the nucleus basalis) could produce a relative increase in cholinergic activity in the
temporal lobe with age. The EEG desynchronization
observed may thus result from this increased cholinergic activity.
The finding of a diminished P300 in the SD group is
consistent with similar reports in the literature 12 1,
22). However, the time epoch is late, and the P300
feature did not rank highly by MV analysis. Moreover,
no P300 feature was developed from the PSD study.
This result is surprising, given reports based on previous investigations in this area 122, 37) and the recent
findings that the P300 is probably generated in subcortical structures, especially in the temporal lobe (25,
441. This inconsistency may stem from a fundamental
problem with the use of the P300 methodology in
pathological populations. For P300 results to have
physiological meaning, the underlying task must be
performed well. In our experience subjects with memory deficits, attentional difficulties, or other symptoms
of cognitive dysfunction often have problems with the
P300 paradigm. They may have trouble understanding
what is required, forget the instructions, or let their
attention wander. Each factor would greatly diminish
P300 amplitudes on an artifactual basis and spuriously
suggest group differences in cases in which the results
are better explained by differential adequacy of task
performance. In our situation technicians sat with the
subject and stopped the session whenever necessary to
reinforce previously given instructions. Although the
adequacy of the performance of many demented subjects was questionable even with such help, it was
much better than that obtained without support. This
procedural variatioo may explain why P300 differences
were not as prominent in our study as in those reported
by others.
Results of the correlational analyses between neurophysiologically defined feature regions and cognitive
task performance (see Tables 1 and 2) demonstrated
that all of the most highly ranked neurophysiological
features correlated significantly with performance on
tasks indicative not only of diminished memory functioning, but also of diminished visuospatial and attentiodabstraction abilities. For both the patients with
PSD and those with SD, considered separately, lower
psychological test scores were associated with greater
amounts of EEG slow activity (delta and theta) and
lesser amounts of fast activity (beta). Thus, EEG synchronization was associated with poorer cognitive performance. Although the significance of beta activity is
not always clear, it is generally agreed that the reduction in beta coincident with augmented slow activity is
a sign of underlying abnormality.
Correlations between feature location and cognitive
performance did not always conform to current notions
of structure-function relationships. For example, although feature CFEA l was derived from frontal
regions and correlated significantly with performance
on traditional frontal lobe measures of attention/
abstraction, it also correlated significantly with performance on memory and visuospatial measures (see
Table 2). In addition, three of the four best features
obtained from the comparison of the patients with SD
and their controls involved primarily the frontal lobes,
yet these features were highly correlated (Y = 0.83)
with delayed memory performance, a function that is
known to depend on structures within the temporal
lobe. There may be several reasons for these complicated relationships. As previously suggested { 14, 27,
281, specific cognitive functions may involve more cortical regions than previously appreciated. Alternately,
the answer may lie in the failure of correlational analysis to indicate causality. The behavioral deficits and the
regional neurophysiological abnormalities may be parallel manifestations of the underlying disease state, thus
reflecting both the presence and severity of the dementia. Their statistical correlation, therefore, would be
through a joint association with the disease state. In
patients with Alzheimer’s disease, the location of
neurophysiological features may be markers of the disease in general, not of the specific functional deficits
produced by the disease. Finally, an explanation may
rest with the feature selection process. For example,
features CFEA 7, CFEA 1, and CFEA 8 were derived
from templates in which virtually the entire head demonstrated differences from the control population with
a t value greater than 2.0. The frontal predominance
was determined by the conservative cutoff level of 2.82
used for template formation. It may be that excluded
regions were more causally related to the measured
neuropsychological deficits and the included (frontal)
regions related to some unmeasured cognitive or behavioral deficiency. It is not possible to distinguish
among these possibilities.
The data of the present study demonstrate that
our topographic mapping method delineates neurophysiological features that retrospectively discriminate
healthy controls and age-equivalent patients with PSD
and with SD of the Alzheimer type. A prospective test
of the sensitivity of this method will require cross validation on an independent sample {32). To examine
specificity it will be necessary to determine whether
other neurological, medical, or psychiatric conditions
(e.g., multi-infarct dementia, cardiovascular disease,
and depression) produce EEG changes similar in nature
and location to those seen in patients with PSD and
with SD.
Supported in part by National Institute on Aging program project
AGO2269 (Principal Investigator, Dr Albert), Grant NS 1367 from
the National Institute of Neurological and Communicative Disorders and Stroke (Principal Investigator, Dr Duffy), Grant HD13420
from the National Institute of Child Health and Human Development (Principal Investigator, Dr Duffy), and NICHD program
project HDO6276 (Principal Investigator, Dr C. F. Barlow).
The authors thank Kay McAnulty for manuscript preparation, Alice
Domar and Carol Cordier for conscientious data gathering and analysis, and Hope Heller and Susan LoCastro for careful administration
of all neuropsychological test measures. Drs Michael Jenike, John
Growdon, Harvey Sagar, Marsel Mesulam, John Tellers, and Sandra
Weintraub assisted in the selection of the patients.
References
1. Afiffi AA, Azen SP: Statistical Analysis: A Computer Oriented
Approach. New York, Academic, 1979
2. Bartels PH: Numerical evaluation of cytologic data. 111. Selection of features for discrimination. Anal Quant Cytol 1:153159, 1979
3. Bartels PH: Numerical evaluation of cytologic data. IV. Discrimination and classification. Anal Quant Cytol 2: 19-24, 1980
4. Bartels PH: Numerical evaluation of cytologic data. VI. Multivariate distributions and matrix notation. Anal Quant Cytol
2:155-160, 1980
5. Bartels PH, Subach JA: Automated interpretation of complex
scenes. In Preston K, Onoe M (eds): Digital Processing of Complex Imagery. New York, Academic, 1976, pp 101-114
6. Brody H: Aging of the vertebrate brain. In Rockstein M, Sussman ML (eds): Development and Aging in the Nervous System.
New York, Academic, 1973, pp 121-133
7. Burchfiel JL, Duffy FH, Sim VM: Persistent effects of sarin and
dieldrin upon the primate encephalogram. Toxicol Appl Pharmacol 35~265-279, 1976
8. Constantinidis J, Krassoievitch M, Tissot R: Corrdations entre
les perturbations electroencephalographiques et les lesions
anatomo-histologiques dans les demences. Encephale 58: 19-52,
1969
9. Cooley WW, Lohnes P R Multivariate Data Analysis. New
York, Wiley, 1971
10 Davies P: Neurotransmitter-related enzymes in senile dementia
of Alzheimer type. Brain Res 171:319-327, 1979
11 Davies P, Katz DA, Crystal HA: Choline acetyltransferase,
somatostatin, and substance P in selected cases of Alzheimer’s
disease. In Corkin S, Davis KL, Growdon JH, et al (eds): Alzheimer’s Disease: A Review of Progress. New York, Raven,
1782, PP 9-14
12 Davies P, Maloney AJR Selective loss of central cholinergic
neurons in Alzheimer’s disease. Lancet 2:1403, 1976
I3 Duffy FH: Topographic display of evoked potentials: clinical
applications of brain electrical activity mapping (BEAM). Ann
NY Acad Sci 388:183-196, 1982
14 Duffy FH, Albert MS, McAnulty G, Garvey J: Age-related differences in brain electrical activity in healthy subjects. Ann
Neurol 16:430-438, 1984
15 Duffy FH, Battels PH, Burchfiel J L Significance probability
mapping. an aid in the topographic analysis of brain electrical
activity. Electroencephalogr Clin Neurol 51:455-462, 1981
Duffy et al: Brain Activity in Dementia
447
16. Duffy FH, Burchfiel JL Long term effects of the organophosphate sarin on EEGs in monkeys and humans. Neurotoxicology
1~667-689, 1980
17. Duffy FH, Burchfiel JL, Lombroso CT: Brain electrical activity
mapping (BEAM): a method for extending the clinical utility of
EEG and evoked potential data. Ann Neurol 5:309-321, 1979
18. Duffy FH, Denckla MB, Bartels PH, Sandhi G: Dyslexia: regional differences in brain electrical activity by topographic mapping. Ann Neurol 7:412-420, 1980
19. Duffy FH, Denckla MB, Bartels PH, et al: Dyslexia: automated
diagnosis by computerized classification of brain electrical activity. Ann Neurol 7:421-428, 1980
20. Genchi H , Mori K: Evaluation and feature extraction on automated pattern recognition system. Denki-Tsuchin Gakkai, Part
I, 1965
21. Goodin DS, Squires KC, Henderson BH, Starr A: Age-related
variations in evoked potentials to auditory stimuli in normal
human subjects. Electroencephalogr Clin Neurophysiol 44:
447-458, 1978
22. Goodin DS, Squires KC, Starr A: Long latency event-related
components of the auditory evoked potential in dementia. Brain
101:635-648, 1978
23. Gordon EG, Sim M: The EEG in presenile dementia. J Neurol
Neurosurg Psychiatry 30:285, 1967
24. Hachinski J: Cerebral blood flow differentiation of Alzheimer’s
disease from multi-infarct dementia. In Katzman R, Terry RD,
Bick ICT (eds): Alzheimer’s Disease: Senile Dementia and Related Disorders. New York, Raven, 1978, pp 97-104
25. Halgren E, Squires NK, Wilson CL, et al: Endogenous potentials
generated in the human hippocampal formation and amygdala by
infrequent events. Science 2 105303-805, 1980
26. Heston LL, Mastri AR: Age at onset of Picks and Alzheimer’s
dementia: implications for diagnosis and research. J Gerontol
37:422-424, 1982
27. Larsen B, Skinhoj E, Lassen NA: Variations in regional cortical
blood flow in the right and left hemispheres during automatic
speech. Brain 101:193-210, 1978
28. Lassen NA, Ingvar DH, Skinhoj E: Brain function and blood
flow. Sci Am 239(4):62-71, 1978
29. Mann HB, Whitney DR: On a test of whether one of two random variabies is stochastically larger than the other. Ann Math
Statistics 18:50-60, 1947
30. Neidermeyer E, Lopes da Silva F: Electroencephalography: Basic Principles, Clinical Applications and Related Fields. Baltimore, Munich, Urban and Swarzenberg, 1982
448 Annals o f Neurology Vol 16 No 4 October 1984
31. Price DL, Whitehouse PJ, Struble RG, et al: Basal forebrain
cholinergic system in Alzheimer’s disease and related dementias.
Neuroscience 134-92, 1982
32. Ransohoff DF, Feinstein AR: Problems of spectrum and bias in
evaluating the efficacy of diagnostic tests. N Engl J Med
299:926-930, 1978
33. Rao C R Linear Statistical Methods in Biometric Research. New
York, Wiley, 1952
34. Selin I: Detection Theory. Princeton, NJ, Princeton University
Press, 1965
35. Sherwood EM, Bartels PH, Wied GL Feature selection in cell
image analysis: use of the ROC curve. Acta Cytol 20:254-260,
1976
36. Sims NR, Bowen DM, Davison AN: Acetylcholine synthesis
and carbon dioxide production from glucose by tissue prisms
from human neocortex. Biochem J 196:867-876, 1981
37. Squires K, Goodin D, S t a r A: Event related potentials in development, aging, and dementia. In Lehmann D, Callaway E (eds):
Human Evoked Potentials: Applications and Problems. New
York, Plenum, 1979, pp 383-396
38. Syndulko K, Hansch E, Cohen SN, et al: Long latency event
related potentials in normal aging and dementia. In Courjon J,
Mauguiere F, Revol M (eds): Clinical Application of Evoked
Potentids (Advances in Neurology, Voi 32). New York, Raven,
1982, pp 279-293
39. Testa G, Gloor P Generalized penicillin epilepsy in the cat:
effects of midbrain cooling. Electroencephalogr Clin Neurophysiol 36:517-524, 1974
40. Weintraub S, Baratz R, Mesularn MM: Assessment of daily living activities in Alzheimer’s disease. In Corkin S, Davis KL,
Growdon JH, et al (eds): Alzheimer’s Disease: A Review of
Progress. New York, Raven, 1982, pp 189-192
4 1. Wells CE: Diagnostic evaluation and treatment in dementia. In
Wells CE (ed): Dementia Philadelphia, Davis, 1977, pp 247276
42. Whitehouse P, Parhad I, Hedreen JC, et al: Integrity of the
nucleus basalis of Meynert in normal aging. Paper presented at
the 35th Annual Meeting of the American Academy of Neurology, San Diego, CA, April 1983
43. Wilks SS: Mathematical Statistics. New York, Wiley, 1962
44. Wood CC, Allison T, Goff WR, et al: On the Neural Origin of
P300 in Man. In Kornhuber H H , Deecke L (eds): International
Symposium on Electrical Potentials Related to Motivation,
Motor, and Sensory Processes of the Brain. Amsterdam,
Elsevier, 1980, pp 51-56
Документ
Категория
Без категории
Просмотров
1
Размер файла
2 116 Кб
Теги
patients, dementia, presenilin, activity, electric, typed, senile, brain, alzheimers
1/--страниц
Пожаловаться на содержимое документа