Biomedical Engineering Reference
In-Depth Information
ANNs are well suited for a variety of biosignal processing applications and may be used
as a tool for nonlinear statistical analysis. They are often used for pattern recognition and
classification. In addition, ANNs have been shown to perform faster and more accurately
than conventional methods for signals that are highly complex or contain high levels of
noise. ANNs also have the ability to solve problems that have no algorithmic solution—in
other words, problems for which a conventional computer program cannot be written.
Since ANNs learn, algorithms are not required to solve problems.
As advances are made in artificial intelligence techniques, ANNs are being used more
extensively in biosignal processing and biomedical instrumentation. The viability of ANNs
for applications ranging from the analysis of ECG and EEG signals to the interpretation of
medical images and the diagnosis of a variety of diseases has been investigated. In neurol-
ogy, research has been conducted by using ANNs to characterize brain defects that occur in
disorders such as epilepsy, Parkinson's disease, and Alzheimer's disease. ANNs have also
been used to characterize and classify ECG signals of cardiac arrhythmias. One study used
an ANN in the emergency room to diagnose heart attacks. The results of the study showed
that, overall, the ANN was better at diagnosing heart attacks than the emergency room
physicians were. ANNs have the advantage of not being affected by fatigue, distractions,
or emotional stress. As artificial intelligence technologies advance, ANNs may provide a
superior tool for many biosignal processing tasks.
EXAMPLE PROBLEM 11.29
A neuron in a neural network has three inputs and uses a sigmoid function to calculate the
output of the neuron. The three values of the inputs are 0.1, 0.9, and 0.1. The weights associated
with these three inputs are 0.39, 0.72, and 0.26, and the bias weight is 0.48 after training. What
is the output of the neuron?
Solution
Using Eq. (11.32) to calculate the relative sum of the inputs gives
x
¼ð
Input 1
Weight 1 Þþð
Input 2
Weight 2 Þþð
Input 3
Weight 3 Þþ
Bias Weight
¼ð
0
:
1
Þ
0
:
39
þð
0
:
9
Þ
0
:
72
þð
0
:
1
Þ
0
:
24
þ
0
:
48
¼
1
:
19
The output of the neuron is calculated using Eq. (11.60):
e x
e 1:19
y ¼
1
1
þ
Þ¼
1
1
þ
Þ¼
0
:
77
11.10 EXERCISES
1. What types of biosignals would the nerves in your legs produce during a sprint across the
street?
2. What types of biosignals can be recorded with an EEG? Describe in terms of both origins and
characteristics of the signal.
Continued
Search WWH ::




Custom Search