Early Detection of Mild Cognitive Impairment and Mild Alzheimer’s
Disease in Elderly using CBF Activation during Verbally-based
Cognitive Tests
Shohei Kato
1
, Hidetoshi Endo
2
, Risako Nagata
2
, Takuto Sakuma
1
and Keita Watanabe
1
1
Dept. of Computer Science and Engineering, Nagoya Institute of Technology
Gokiso-cho, Showa-ku, Nagoya 466-8555, Japan
2
Dept. of Comprehensive Geriatric Medicine, National Center for Geriatrics and Gerontology
35 Gengo, Morioka-machi, Obu, Aichi 474-8511, Japan
Keywords:
Verbally-based Cognitive Training to Prevent Cognitive Impairment from Elderlies, Early Detection of
Dementia, Cerebral Blood Flow Activation during a Cognitive Task, Functional near Infrared Spectroscopy
(fNIRS).
Abstract:
With the goal of promoting a fruitful and healthy longevity society, this paper presents a verbally-based cog-
nitive task and an early detection method of dementia and mild cognitive impairment for elderly. As designed
with conscious of daily conversation, the task is done by verbally responding to questionnaire. An elderly
firstly talks about the topics of favorite season, travel, gourmet, and daily life, and then he/she does three cog-
nitive tasks of reminiscence, category recall, and working memory. With the use of the functional near-infrared
spectroscopy (fNIRS), which can measure cerebral blood flow activation non-invasively, we had collected 42
CHs fNIRS signals on frontal and right and left temporal areas from 22 elderly participants (7 males and 15
females between ages of 64 to 89) during cognitive tests in a specialized medical institute. All participates are
classified into three clinical groups: elderly individuals with cognitively normal controls (CN), patients with
mild cognitive impairment (MCI), and mild Alzheimer’s disease (AD). In this paper, we report a task effect
measurement of the verbally-based cognitive task by the statistical tests of fNIRS signals, and then report the
examination of the detection performance by cross-validation using proposed Bayesian classifier, which can
discriminate among elderly individuals with three clinical groups: CN, MCI, and AD. Consequently, empirical
result indicated that total accuracy rate is more than 95% and the result suggests that proposed approach is
adequate practical to screen the elderly with cognitive impairment.
1 INTRODUCTION
It is no secret that dementia is one of the most press-
ing challenges in developed countries. Current esti-
mates indicate 35.6 million people worldwide are liv-
ing with dementia but with the world’s populations
aging, the World Health Organization estimates that
number will nearly double every 20 years, to an es-
timated 65.7 million in 2030, and 115.4 million in
2050. With this serious situation, the first G8 sum-
mit on dementia was held in London in December
2013 in order to start global action against dementia,
and policy papers of G8 dementia summit agreements
(RDD/10495, 2013) were published.
In Japan, the Japanese Ministry of Health, Labour
and Welfare (MHLW) has begun projects to improve
dementia treatment and quality of life from 2008, and
in 2012, MHLW announced five-year measures for
dementia care,
Orange Plan
, in which hospitals and
geriatric facilities centered dementia care should be
shift to home care support and community centered
dementia care, for sustaining the elderly’s activities of
daily life and quality of life. The plan includes pro-
vision of various support such as watch over, health
care and rehabilitation to community life support ser-
vice (Nakanishi and Nakashima, 2013).
On the other hand, it is also important to de-
velop a diagnostic tool which can early detect elderly
with dementia. To screen for dementia and cogni-
tive impairment, a questionnaire test such as Mini-
Mental State Examination (MMSE) (Folstein et al.,
1975), Revised Hasegawa’s Dementia Scale (HDS-
R) (Imai and Hasegawa, 1994), Clinical Dementia
Rating (CDR) (Morris, 1993), and Memory Impair-
ment Screen (MIS) (Buschke et al., 1999), is com-
366
Kato S., Endo H., Nagata R., Sakuma T. and Watanabe K..
Early Detection of Mild Cognitive Impairment and Mild Alzheimer’s Disease in Elderly using CBF Activation during Verbally-based Cognitive Tests.
DOI: 10.5220/0004821603660373
In Proceedings of the International Conference on Health Informatics (HEALTHINF-2014), pages 366-373
ISBN: 978-989-758-010-9
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
monly used. These are simple and well-formed tests,
but there are several problems in addition to time-
consuming consultation with an expert in psycholog-
ical testing.
• Score may include subjectivity.
• Score may be influenced by education, social
class, and gender deference.
MMSE is in widespread use as a practical and clin-
ically useful test for practitioners. However, even if
it takes about 10 minutes, patient load goes over the
capacity of a general practitioner.
In our previous study, we have studied novel ap-
proaches for the early detection of cognitive impair-
ment in the elderly, in which we focused on the
prosodic features of speech sound during the sub-
ject’s answers to the questionnaire; the first was to
detect signal and prosodic signs of cognitive impair-
ment (Kato et al., 2011), the second was to take
a measurement of cerebral blood flow (CBF) (Kato
et al., 2012). We then have developed a prototype of
prosody-CBF hybrid screening system and discussed
the cost-effectiveness and the discrimination perfor-
mance (Kato et al., 2013).
In such situation, this paper provides a new dis-
criminative method of MCI and mild AD without
score of questionnaire-based test, with the use of
just speech voice and fNIRS signals during daily
conversation. In Section 3, we firstly propose a
block-designed verbally-based cognitive training task
for elderly with mild cognitive impairment. As de-
signed with conscious of daily conversation, the task
is done by oral answering some questionnaire. An
elderly firstly talks about the topics of favorite sea-
son, travel, gourmet, and daily life, and then he/she
does three cognitive tasks of reminiscence task, cate-
gory recall, and working memory task. With the use
of the functional near-infrared spectroscopy (fNIRS),
which can measure cerebral blood flow activation
non-invasively, we had collected fNIRS signals on
frontal and right and left temporal areas from elderly
participants during cognitive tests in a specialized
medical institute. In Section 5, we report a task ef-
fect measurement by the statistical tests of fNIRS sig-
nals. In Section 6, we also report the examination of
the detection performance by cross-validation using
proposed Bayesian classifier, which can discriminate
among elderly individuals with three clinical groups:
CN, MCI, and AD.
2 WHY FOCUS ON THE
VERBALLY-BASED
COGNITIVE ACTIVITIES
The Nun Study is known as a continuing longitudi-
nal study to examine the onset of Alzheimer’s Dis-
ease focusing on the linguistic ability. Snowdon et
al. (Snowdon et al., 1996) found that an essay’s lack
of linguistic density (e.g., complexity, vivacity, flu-
ency) functioned as a significant predictor of its au-
thor’s risk for developing Alzheimer’s disease in old
age. How about the verbal performance?
Relatively infrequent in the literature, but there
are several studies that suggest associations with
Alzheimer’s disease and verbal performance. Taler
et al. have reported language performance (Taler
and Phillips, 2007), and especially comprehen-
sion of grammatical and emotional prosody (Taler
et al., 2008) are impaired in elderly patients with
Alzheimer’s disease (AD). Hoyte et al. (Hoyte
et al., 2009) reported that the components of speech
prosody are useful for detecting the syntactic struc-
ture of speech. P
´
erez Trullen and Modrego Pardo
(Trullen and Pardo, 1996) investigated aprosody in
AD and multi-infarct dementia (MID), and reported
that aprosody is more frequent and severe in AD. Mc-
Dowd et al. (McDowd et al., 2011) reported some
interesting findings about verbal fluency in healthy
aging, Alzheimer’s disease, and Parkinson’s disease.
Carter et al. (Carter et al., 2012) suggested that as-
sessment of patients with mild cognitive impairment
(MCI) should include language and semantic mem-
ory tests in addition to typical episodic memory tests,
as changes within this domain might be a sensitive
indication of incipient AD, on the basis of a cross-
sectional investigation compared cognitively normal
age-matched controls with patients with mild AD and
MCI using a detailed neuropsychological assessment.
Cognitive models of voice processing by Belin et al.
(Belin et al., 2004) propose that there is parallel pro-
cessing of identity information and emotional content
from voices. Hailstone et al. have provided intrigu-
ing reports about voice processing (Hailstone et al.,
2011) and accent processing (Hailstone et al., 2012)
in dementia, on the basis of neuropsychological and
neuroanatomical analysis of voice processing.
These reports suggest the possibility of using
prosodic feature extracted from elderly speech to
screen for dementia. In our previous studies (Kato
et al., 2011), (Kato et al., 2012), (Kato et al., 2013),
we have focused on the prosodic feature of elderly’s
speech sounds during orally answering some ques-
tionnaire, for early detection of mild cognitive impair-
ment (MCI) and mild Alzheimer’s disease (AD) in the
EarlyDetectionofMildCognitiveImpairmentandMildAlzheimer'sDiseaseinElderlyusingCBFActivationduring
Verbally-basedCognitiveTests
367
Table 1: A Breakdown List of Participants (N=22).
Age 64-70 71-75 76-80 81-85 86-90 91-93 Total
Male 0(0,0,0) 2(2,0,0) 1(1,0,0) 1(0,0,1) 2(0,2,0) 1(0,1,0) 7(3, 3,1)
Female 4(1,3,0) 2(1,1,0) 4(1,2,1) 3(1,0,2) 2(0,1,1) 0(0,0,0) 15(4, 7,4)
Subtotal 4(1,3,0) 4(3,1,0) 5(2,2,1) 4(1,0,3) 4(0,3,1) 1(0,1,0) 22(7,10,5)
Value in bracket means the number of subjects in CN, MCI, AD clinical groups.
Season (60sec)
30sec
Rest
Daily conversation
Rest
(10sec)
Travel (60sec) Gourmet (60sec)
270sec
Life (60sec)
Rest
(10sec)
Rest
(10sec)
00:00 05:00
60sec
Reminiscence1: Talking
Birthdate
Birthplace
Elementary School
60sec
Rest
60sec
Rest
60sec
Reminiscence2: Watching
Kitchen
Old Things
60sec
Rest
05:00 10:00
60sec
Working Memory1
Category Recall
Animals
Fruits
60sec
Rest
60sec
Working Memory2
30sec
Rest
10:00
13:30
Letter-Number Sequencing (WAIS-III)
: fNIRS measurement
: speech recording
Figure 1: Block Design Task of Cognitive Tests.
elderly.
From the results of the studies, we have designed a
verbally-based cognitive training task for elderly with
MCI and mild AD. In this paper, we firstly evaluate
comparatively the CBF activation during the tasks in
cognitively normal controls (CN) and patients with
MCI and mild AD. Secondly we propose a Bayesian-
based discriminative method of CN, MCI, and AD.
3 VERBALLY-BASED
COGNITIVE TASK
3.1 Participants
Twenty two Japanese subjects (7 males and 15 fe-
males between the ages of 64 and 93 years) par-
ticipated in this study. Table 1 shows the break-
down list of participants. In this study, all partic-
ipants are classified into three clinical groups: el-
derly individuals with cognitively normal controls
(CN), patients with mild cognitive impairment (MCI),
and mild Alzheimer’s disease (AD). All participants
in disease groups are clinically conditioned that the
CDR of a participant in MCI group and AD group
corresponds to CDR0.5 and CDR1, respectively. The
MMSE scores were distributed with 29.30 ± 0.96
(CN), 28.63 ± 1.85 (MCI), and 23.40 ± 2.51 (AD).
3.2 Cognitive Tasks
To measure brain function of an elderly during vari-
ous cognitive tests including HDS-R, we have made
a block designed task shown in Fig. 1, and then con-
ducted simultaneous voice-fNIRS measurement dur-
ing cognitive tests.
Firstly a participant talks with a clinical psy-
chotherapist (they meet for the first time) about fa-
vorite season, travel, gourmet, and daily life. And
then, he/she does two reminiscence tasks (1. talking
about birthdate and birthplace, 2. watching old-style
kitchen and old things), category recall (animals and
fruits), and working memory tasks. Under the cate-
gory recall task, the participant says the name of ani-
mals and fruits as many as he/she knows. As the the
working memory task, we have adopted letter-number
sequencing, a working memory index from Wech-
sler Adult Intelligence Scale-Third Edition (WAIS-
III(Wechsler, 1997)). Under the task, the participant
is told three numbers and
hiragana
, a Japanese syl-
labary, characters and then replies them with sorted in
ascending and dictionary order, respectively.
These eight tasks are done for 60 seconds after rest
gazing at a single point on the display for 60 seconds
interval, and it takes thirteen minutes in the total.
4 fNIRS MEASUREMENT
Functional near-infrared spectroscopy (fNIRS) can
measure neural activity of the cerebral cortex using
HEALTHINF2014-InternationalConferenceonHealthInformatics
368
fNIRS
indicator
Display
speaker
mic
partition
seated position
Figure 2: Snapshot of fNIRS measurement of an elderly
participant having a cognitive test.
Figure 3: Channel arrangement of fNIRS measurement.
infrared rays that are safe to living organisms (Vill-
ringer and Chance, 1997). fNIRS monitors regional
relative changes of oxy/deoxygenated hemoglobin
concentration to measure cortical activation utiliz-
ing the tight coupling between neural activity and
regional cerebral blood flow (Villringer and Firnafl,
1995). This measurement method requires only com-
pact experimental systems and can eliminate physical
restraint from a subject by non-invasive procedures.
Fig. 2 shows a snapshot of fNIRS measurement of
an elderly participant having a cognitive test. As the
figure, an elderly participant, with 32 fNIRS probes
mounted on his/her head, seats in front of the LCD
display and microphone. While measurement, he/she
does casual talks and cognitive tasks as mentioned
above.
We used the fNIRS topography system FOIRE-
3000 Near-Infrared Brain Function Imaging System
(Shimadzu, Kyoto, Japan), which uses near-infrared
light with wavelengths of 780, 805, and 830 nm. We
set 16 illuminators and 15 detectors in lattice pattern
to form 42 channels (CHs) (22 CHs on frontal lobe,
10 CHs on right parietal and temporal lobe, 10 CHs
on left parietal and temporal lobe) shown in Fig. 3.
5 TASK EFFECT ASSESSMENT
Toward personalized cognitive training, we assess
a task effect. We have conduct statistical tests of
between-group significant differences using fNIRS
signals of oxy-Hb during cognitive task of
• Daily Conversation: talking about favorite season,
travel, gourmet, and daily life;
• Reminiscence Task 1: talking about birthdate;
• Reminiscence Task 2: watching old things;
• Category Recall: answering the name of fruits as
many as he/she knows;
• working memory tasks.
We used Welch’s t-test with significance level of
(P< 0.001) after applying Bonferroni’s adjustment
(1/42). Fig. 4 and 5 show the results of t-test for sig-
nificant differences in channel-wise fNIRS signals be-
tween any single pair from CN, MCI, and AD groups.
The CHs that exhibited significant oxy-Hb increase
are colored according to the t-values, as shown in the
color bar, while those below the threshold are indi-
cated in gray. The higher absolute value of a t-value
indicates the larger difference between two groups.
In Fig. 4 D-(a), for example, red colored mapping
on frontal lobe indicates that the brain function in
frontal lobe of elderlies with cognitively normal is
significantly more active than that of elderly patients
with MCI. In Fig. 4 D-(c), for example, blue colored
mapping on left temporal lobe indicates that the brain
function in left temporal lobe of elderly patients with
MCI is significantly more calm than that of elderly
patients with mild AD.
5.1 Results
5.1.1 Daily Conversation
Fig. 4 indicates the significant difference of fNIRS
signals between normal group and disease groups for
all topics of the conversation. Speech contains non-
verbal elements known as paralanguage, including
voice quality, rate, pitch, volume, and speaking style,
as well as prosodic features such as rhythm, intona-
tion, and stress. Conversation, even if it is just a ca-
sual conversation about daily living, requires lot of
interpersonal communication skills of not only lin-
guistic competence but also some nonverbal compe-
tencies so as to recognize partner’s paralanguage and
to control self paralanguage appropriately. For this
reason, the results indicate that elderly people with
cognitively normal use their brain function more ac-
tively than those who are with MCI and AD group.
EarlyDetectionofMildCognitiveImpairmentandMildAlzheimer'sDiseaseinElderlyusingCBFActivationduring
Verbally-basedCognitiveTests
369
(a) NC group - MCI group (b) NC group - AD group (c) MCI group - AD group
A: talking about season
(a) NC group - MCI group (b) NC group - AD group (c) MCI group - AD group
B: talking about travel
(a) NC group - MCI group (b) NC group - AD group (c) MCI group - AD group
C: talking about gourmet
(a) NC group - MCI group (b) NC group - AD group (c) MCI group - AD group
D: talking about daily life
t-value
-50 500 25-25
Figure 4: Results of t-test for significant differences in channel-wise fNIRS signals between any single pair from CN, MCI,
and AD groups (daily conversation).
Maybe it is because some participants talked heated
gradually, the difference becomes larger as a conver-
sation progresses.
5.1.2 Reminiscence Talk, Category Recall, and
Working Memory
Fig. 5 indicates the significant but slight difference
of fNIRS signals between normal group and disease
groups for the later three tasks. With respect to the
reminiscence task 1, for this particular participants,
regardless of whether they have cognitive impairment
or not, answering his/her birthdate is very easy; it
is no need to use their brain function actively. With
respect to the category recall and working memory
tasks, for an elderly participant, it is a little too hard
to say the name of fruits as many as he/she knows for
long time (30 seconds), and is also a little too diffi-
cult to listen numbers and
hiragana
characters and sort
them in ascending and dictionary order, at once. With
respect to the reminiscence task 2, we failed to col-
lect fNIRS signals for adequate analysis because the
visual stimulation was not enough to prompt a verbal
response from participants.
6 CLASSIFICATION OF CN, MCI,
AD GROUPS USING fNIRS
SIGNALS
The section describes a Bayesian classifier using
fNIRS signals of elderlies during cognitive tests,
which can discriminate among elderly individuals
with three clinical groups: cognitively normal con-
trols (CN), patients with mild cognitive impairment
(MCI), and Alzheimer’s disease (AD). To design an
algorithm for computer-aided diagnosis of cognitive
impairment in the elderly, we consider the screening
process by a specialist in geriatrics. We thus propose a
two-phase Bayesian classifier shown in Fig. 6 on the
assumption of screening process, that firstly checks
the suspicion of the cognitive impairment (CI) or not
(CN) from given fNIRS signals; if any, and then sec-
ondly judges the degree of the impairment: MCI or
AD.
HEALTHINF2014-InternationalConferenceonHealthInformatics
370
(a) NC group - MCI group (b) NC group - AD group (c) MCI group - AD group
E: reminiscence 1: birthdate
(a) NC group - MCI group (b) NC group - AD group (c) MCI group - AD group
F: category recall: fruits
(a) NC group - MCI group (b) NC group - AD group (c) MCI group - AD group
G: working memory task
t-value
-50 500 25-25
Figure 5: Results of t-test for significant differences in channel-wise fNIRS signals between any single pair from CN, MCI,
and AD groups (later three tasks).
fNIRS signals
NB Classifier (1st phase)
NB Classifier (2nd phase)
CN / CI
MCI / AD
CN MCI AD
(42CH oxy-Hb)
cognitive impairment (CI)
Feature Extraction
spectral, statistics 77 features
Primitive Analysis
CH-wise low-path and/or difference filters,
domain averaging
Figure 6: Classification of CN/MCI/AD by two-phase
Bayesian Classifier.
6.1 Primitive Analysis of fNIRS Signals
In advance of Bayesian classification, we make a
primitive signal processing fNIRS signals. Firstly,
we make five fNIRS signals every channels such that
noise is reduced by channel-wise smoothing through
three low-pass filters (F1: 1.92 [Hz], F2: 0.96 [Hz],
F3: 0.48 [Hz]) and difference filters (F1-3 and F2-3).
Secondly, we segregate 42 CHs into the seven brain
areas (see Fr, Fc, Fl, Rf, Rr, Lf, and Lr in Fig. 3) and
then make signal averaging that integrates fNIRS sig-
nals within each of the areas.
Table 2: fNIRS Feature Candidates.
fNIRS filtered Feature / Statistics
Filter 1 (F1) Mean value (mean)
Fundamental Frequency (f0)
Centroidal Frequency (fc)
Filter 3 (F3) Maximum value (max)
Minimum value (min)
Variance (var)
Mean value (mean)
Fundamental Frequency (f0)
Gradient of the linear regression
line (gr)
Filter1-3 (F1-3) Variance (var)
Filter2-3 (F2-3) Variance (var)
6.2 Extraction of fNIRS Features
We enumerate features that represent fluctuations of
regional cerebral blood flow if it is the slightest effec-
tive in detection of cognitive impairment, and extract
11 features shown in Table 2 from fNIRS signals in
each of the seven brain areas.
6.3 Bayesian Classifier
In this paper, we adopted naive Bayes classifier (NB),
(Langley et al., 1992) which is a simple Bayesian
classifier with strong independence assumption of
attributes. We construct two classifiers: NB
CN/CI
,
which checks the suspicion of the cognitive impair-
EarlyDetectionofMildCognitiveImpairmentandMildAlzheimer'sDiseaseinElderlyusingCBFActivationduring
Verbally-basedCognitiveTests
371
Table 3: Selected fNIRS Features.
Classifier Selected Features
NB
CN/CI
Lr F3 gr, Rf F1 mean, Rf F3 mean, Fl F3 f0
NB
MCI/AD
Rr F3 min, Fr F3 f0, Rr F1-3 var, Lf F3 gr,
Fc F3 gr
Table 4: Classification Results.
P
P
P
P
P
P
clinical
detection
CN MCI AD accuracy
CN 7 0 0 100%
MCI 1 9 0 90%
AD 0 0 5 100%
predictive value 87.5% 100% 100% 95.5%
ment (CI) or not (CN) at the first phase, if any suspi-
cion, and NB
MCI/AD
, which judges the degree of the
impairment (MCI or AD) at the second phase.
6.4 Classification Assessment
We have examined discrimination performance by
modeling two-phase Bayesian classifiers for discrim-
inating among elderly individuals with NL, MCI, and
AD, by using fNIRS signals of oxy-Hb during daily
conversation (travel) (see Fig. 1) collected from 22
participants (see Table 1). Table 3 shows the selected
fNIRS features by each of NB classifiers. To evalu-
ate detection performance, we adopted leave-one-out
cross-validation.
Table 4 shows the confusion matrices and the
statistics of classification results using two-phase
classifiers consist of NB
CN/CI
and NB
MCI/AD
. The re-
sults indicate that total accuracy rate is 95.5% and the
accuracy rate of AD and the predictive value of CN
are both 100%. Table 4 shows that only one subject
in MCI groups are misclassified into CN group and
other all subjects are classified correctly. This sug-
gests that proposed approach is sufficient practical to
screen the elderly with cognitive impairment.
7 CONCLUSIONS
With the goal of promoting a fruitful and healthy
longevity society, this paper presented a verbally-
based cognitive task and an early detection method of
dementia and mild cognitive impairment for elderly.
With conscious of daily conversation, we firstly
designed eight cognitive tasks, under which an elderly
individual talks about some topics and orally answer-
ing some questionnaire. With the use of the func-
tional near-infrared spectroscopy (fNIRS), we evalu-
ated comparatively the cerebral blood flow (CBF) ac-
tivation during the tasks in cognitively normal con-
trols (CN) and patients with mild cognitive impair-
ment (MCI), and mild Alzheimer’s disease (AD), by
the statistical tests of fNIRS signals between any sin-
gle pair from CN, MCI, and AD. Then we confirmed
the significant difference of CBF activation between
normal group and disease groups for casual conver-
sation. Consequently, the results suggested that pro-
posed cognitive tasks, especially casual conversation,
are adequate practical to training of brain function ac-
tivation for elderly people.
We secondly proposed a two-phase Bayesian clas-
sifier using fNIRS signals, which can discriminate
among elderly individuals with three clinical groups:
CN, MCI, and AD. Then we examined the detection
performance with total accuracy rate of 95.5% was re-
ported by cross-validation. Consequently, the empir-
ical results suggested that proposed approach is suf-
ficient practical to screen the elderly with cognitive
impairment. However, the results are limited in the
small number of participants. Cognitive functioning
are continuous within each category, there is a range
of severity and functional ability. In future work, we
will dedicate to much more clinical trials to validate
the practicality of proposed method. We will also
dedicate to division of MCI group with clinical sub-
types of amnestic MCI and non-amnestic MCI.
ACKNOWLEDGEMENTS
We are grateful to National Center for Geriatrics and
Gerontology and Ifcom Co. Ltd. for clinical data col-
lection environment and data collection, respectively.
This work was supported in part by the Ministry of
Education, Science, Sports and Culture, Grant–in–
Aid for Scientific Research under grant #25280100
and #25540146, and part by Adaptable and Seamless
Technology Transfer Program through target-driven
R&D, JST, and part by the Program to Promote Tech-
nology Transfer and Innovation through Collabora-
tion between Universities, JST, and part by Suzuken
Memorial Foundation.
REFERENCES
Belin, P., FecteauS, S., and B
´
edard, C. (2004). Thinking the
voice: neural correlates of voice perception. Trends in
Cognitive Sciences, 8:129–135.
Buschke, H., Kuslansky, G., Katz, M., Stewart, W. F., Sli-
winski, M. J., Eckholdt, H. M., and Lipton, R. B.
(1999). Screening for dementia with the Memory Im-
pairment Screen. Neurology, 52(2):231–238.
Carter, S. F., Caine, D., Burns, A., Herholz, K., and Ralph,
M. A. L. (2012). Staging of the cognitive decline
HEALTHINF2014-InternationalConferenceonHealthInformatics
372
in Alzheimer’s disease: insights from a detailed neu-
ropsychological investigation of mild cognitive im-
pairment and mild Alzheimer’s disease. International
Journal of Geriatric Psychiatry, 27(4):423–432.
Folstein, M. F., Folstein, S. E., and McHugh, P. R. (1975).
“Mini-Mental State”: A practical method for grading
the cognitive state of patients for the clinician. J. Psy-
chiat. Res, 12(3):189–198.
Hailstone, J. C., Ridgway, G. R., Bartlett, J. W., Goll, J. C.,
Buckley, A. H., Crutch, S. J., and Warren, J. D. (2011).
Voice processing in dementia: a neuropsychological
and neuroanatomical analysis. Brain, 134:2535–2547.
Hailstone, J. C., Ridgway, G. R., Bartlett, J. W., Goll,
J. C., Crutch, S. J., and Warren, J. D. (2012). Accent
processing in dementia. Neuropsychologia, 50:2233–
2244.
Hoyte, K., Brownell, H., and Wingfield, A. (2009). Com-
ponents of Speech Prosody and their Use in Detection
of Syntactic Structure by Older Adults. Experimental
Aging Research, 35(1):129–151.
Imai, Y. and Hasegawa, K. (1994). The revised Hasegawa’s
Dementia Scale (HDS-R): evaluation of its usefulness
as a screening test for dementia. J. Hong Kong Coll.
Psychiatr., 4(SP2):20–24.
Kato, S., Endo, H., Homma, A., Sakuma, T., and Watan-
abe, K. (2013). Early Detection of Cognitive Impair-
ment in the Elderly Based on Bayesian Mining Using
Speech Prosody and Cerebral Blood Flow Activation.
In Proc. of the 35th Annual International Conference
of the IEEE Engineering in Medicine and Biology So-
ciety (EMBC’13), pages 5183–5186.
Kato, S., Endo, H., and Suzuki, Y. (2012). Bayesian-Based
Early Detection of Cognitive Impairment in Elderly
Using fNIRS Signals during Cognitive Tests. In Proc.
of 6th International Joint Conference on Biomedi-
cal Engineering Systems and Technologies (BIOSTEC
2012), pages 118–124.
Kato, S., Suzuki, Y., Kobayashi, A., Kojima, T., Itoh, H.,
and Homma, A. (2011). Statistical analysis of the
signal and prosodic sign of cognitive impairment in
elderly-speech: a preliminary study. In Proc. of 5th
International Joint Conference on Biomedical Engi-
neering Systems and Technologies (BIOSTEC 2011),
pages 322–327.
Langley, P., Iba, W., and Thompson, K. (1992). An anal-
ysis of Bayesian classifiers. In Proc. of The Tenth
National Conference on Artificial Intelligence (AAAI-
92), pages 223–228.
McDowd, J., Hoffman, L., Rozek, E., Lyons, K., Pahwa,
R., Burns, J., and Kemper, S. (2011). Understanding
verbal fluency in healthy aging, Alzheimer’s disease,
and Parkinson’s disease. Neuropsychologia, 25, 210-
225., 25(2):210–225.
Morris, J. C. (1993). The Clinical Dementia Rating
(CDR): Current version and scoring rules. Neurology,
43(11):2412–2414.
Nakanishi, M. and Nakashima, T. (2013). Features of
the Japanese national dementia strategy in compari-
son with international dementia policies: How should
a national dementia policy interact with the public
health- and social-care systems? Alzheimer’s & De-
mentia. (in press, corrected proof, available online).
RDD/10495 (2013). G8 Dementia Summit Declara-
tion and G8 Dementia Summit Communique. De-
partment of Health and Prime Minister’s Office,
UK. http://dementiachallenge.dh.gov.uk/category/g8-
dementia-summit/.
Snowdon, D. A., Kemper, S. J., Mortimer, J. A., Greiner,
L. H., Wekstein, D. R., and Markesbery, W. R. (1996).
Linguistic Ability in Early Life and Cognitive Func-
tion and Alzheimer’s Disease in Late Life: Findings
From the Nun Study. Journal of the American Medi-
cal Association, 275(7):528–532.
Taler, V., Baum, S. R., Chertkow, H., and Saumier, D.
(2008). Comprehension of grammatical and emo-
tional prosody is impaired in Alzheimer’s disease.
Neuropsychology, 22(2):188–195.
Taler, V. and Phillips, N. (2007). Language performance in
Alzheimer’s disease and mild cognitive impairment:
A comparative review. Journal of Clinical and Exper-
imental Neuropsychology, 30(5):501–556.
Trullen, J. M. P. and Pardo, P. J. M. (1996). Comparative
study of aprosody in Alzheimer’s disease and in multi-
infarct dementia. Dementia, 7(2):59–62.
Villringer, A. and Chance, B. (1997). Non-invasive optical
spectroscopy and imaging of human brain function.
Trends Neurosci., 20:435–442.
Villringer, A. and Firnafl, U. (1995). Coupling of brain ac-
tivity and cerebral blood flow: basis of functional neu-
roimaging. Cerebrovasc. Brain Metab. Rev., 7:240–
276.
Wechsler, D. (1997). Wechsler Adult Intelligence Scale–
Third Edition. San Antonio, TX: The Psychological
Corporation.
EarlyDetectionofMildCognitiveImpairmentandMildAlzheimer'sDiseaseinElderlyusingCBFActivationduring
Verbally-basedCognitiveTests
373