Information Technology Reference
In-Depth Information
based analytical tools can be very useful for in-depth study and classification of
data. Artificial neural networks can perform the necessary transformation and
clustering operations automatically and simultaneously. Neural networks have
recently become widely accessible for automatic processing of EEG, thanks to
increased availability of computing facilities. Several studies have presented the
performance of neural network systems when used for the detection and recog-
nition of abnormal EEGs, related to spike detection in epilepsy [1].
EEG has a long tradition as a useful neurophysiological tool in the study of
metabolic encephalopathies, because of non invasiveness, easy performance, wide
availability of implementation in clinical settings, and low cost. Studies from our
laboratory have shown that the EEG of patients with low grade hepatic
encephalopathy have a marked slowdown, as well as significant changes detectable
by linear and nonlinear analysis techniques [2-4]. In the encephalopathy associated
with CRF some EEG findings have also been described, as slowing of background
activity and increases in spectral power in slow frequency bands [5-7].
The aim of the present study was to investigate the potential of applying the
RBF neural network architecture for prospective evaluation of cerebral function
in CRF patients based on relevant characteristic features extracted from EEG.
2 Methods
The proposed approach for the classification of CRF patients involves prepro-
cessing of the EEG signal, extraction of characteristic features and classification
using articial neural networks (ANN) techniques. Using RBF based ANN tech-
niques, the subjects were classified into two classes: normal (control subjects)
and abnormal (renal disease present).
2.1 Patients
We selected 17 patients with CRF without dementia, 7 men and 10 women,
aged between 29 and 82 years (mean 52.86, SD = 17) studied at the Nephrology
Service of Ramn y Cajal Hospital in Madrid, and a control group of 18 age-
matched patients with normal EEG and absence of metabolic illness. The main
demographic, clinical and biochemical criteria of the patients are reported in
table 1.
Table 1.
Clinical description of groups
Parameter
Control
CRF
Cases (n)
18
17
Sex (M/F)
9/9
10/7
Age
57.2
±
17.6
56.0
±
15.1
Time in dialysis (months)
30.9
±
28.5
Glomerular filtration (ml/min)
12.84
±
3.44
Creatinine (mg/dl)
7.63
±
3.00
Search WWH ::
Custom Search