Biomedical Engineering Reference
In-Depth Information
controls. These CNV data combined with the voltage response values were
then used as data sources for Kohonen unsupervised ANNs and ART in an
attempt to classify the groups. The recognition results of HD versus normal
healthy controls were reported to be similar for the two classifiers.
6.2.1.3
Alzheimer's Disease
AD is a common form of dementia and the causes are uncertain although it
is thought that genetic factors are highly related. The most serious effect is
on memory and long-term locomotive and lingual skills. Alzheimer's patients
usually demonstrate a steady decline in such abilities due to the neurologi-
cal consequences, which has no cure at present. Treatment usually includes
acetylcholinesterase inhibitors, psychosocial interventions, various therapies,
and vaccines. With early diagnosis essential to the ecient treatment of
Alzheimer's patients, EEG techniques are being developed to provide early
detection.
Owing to the increasing prevalence of Alzheimer's, as indicated earlier,
there is a need for techniques to predict onset so that appropriate care can
be planned. In addition to methods such as MRI, EEG is regarded as being
potentially effective in identifying the development of this disease. Some bands
of the EEG waveform are affected by the onset of AD, for example, the devel-
opment of AD was found to correlate well with an increase in the delta and
theta power band (Kamath et al., 2006). Bennys et al. (2001) experimented
with several spectral indices calculated from the EEG to discover whether
these indices are affected in the AD population, their study included EEG
recordings from 35 AD patients and 35 controls with no history of neurologi-
cal disorders. The spectral indices were calculated from the quantitative EEG
recordings and statistical analysis was conducted to evaluate the discrimi-
natory power of these indices between the two groups. The high-frequency
(alpha + beta)/low-frequency (delta + theta) ratio was found to differentiate
well between the control and AD patients suggesting that an increase in delta
and theta power in AD patients provides a good discriminating variable.
The early detection of AD is critical for effective treatment and this can
slow the rate of progression. Yagneswaran et al. (2002), for example, compared
the power frequency and wavelet characteristics of an EEG segment between
patients with and without Alzheimer's; LVQ-based NNs were then trained to
distinguish between individuals with and without AD.
Although the detection of Alzheimer's from an EEG signal is generally
suitable, various “abnormalities”within the symptoms of Alzheimer's patients
can cause such models to be less ecient. Petrosian et al. (1999), therefore,
focused on the development of a computerized method for the recognition of
Alzheimer's patients from EEG where they applied trained recurrent neural
networks (RNNs) pooled with wavelet processing for the determination of
early-Alzheimer's patients as opposed to non-Alzheimer's. Cho et al. (2003)
investigated an Alzheimer's recognition method involving ANNs and GA. The
EEG from early AD patients and non-Alzheimer's patients were analyzed to
Search WWH ::
Custom Search