Biomedical Engineering Reference
In-Depth Information
The machine intelligence quotient (MIQ) is a measure of the intelligence level of
machines. The higher a machine's MIQ, the higher the capacity of the machine for auto-
matic reasoning and decision making. The MIQ of a wide variety of machines has risen sig-
nificantly during the past few years. Many computer-based consumer products, industrial
machinery, and biomedical instruments and systems are using more sophisticated artificial
intelligence techniques. Advancements in the development of fuzzy logic, neural networks,
and other soft computing techniques have contributed significantly to the improvement of
the MIQ of many machines.
Soft computing is an alliance of complementary computing methodologies. These meth-
odologies include fuzzy logic, neural networks, probabilistic reasoning, and genetic algorithms.
Various types of soft computing often can be used synergistically to produce superior intelligent
systems. The primary aim of soft computing is to allow for imprecision, since many of the pa-
rameters that machines must evaluate do not have precise numeric values. Parameters of
biological systems can be especially difficult to measure and evaluate precisely.
11.9.1 Fuzzy Logic
Fuzzy logic is based on the concept of using words rather than numbers for computing,
since words tend to be much less precise than numbers. Computing has traditionally
involved calculations that use precise numerical values, while human reasoning generally
uses words. Fuzzy logic attempts to approximate human reasoning by using linguistic vari-
ables. Linguistic variables are words that are used to describe a parameter. For body tem-
perature, linguistic variables that might be used are high fever, above normal, normal,
below normal, and frozen. The linguistic variables are more ambiguous than the number
of degrees Fahrenheit, such as 105.0, 98.9, 98.6, 97.0, and 27.5.
In classical mathematics, numeric sets called crisp sets are defined, while the basic ele-
ments of fuzzy systems are fuzzy sets. An example of a crisp set is A
[0, 20]. Crisp sets
have precisely defined, numeric boundaries. Fuzzy sets do not have sharply defined
bounds. Consider the categorization of people by age. Using crisp sets, the age groups
could be divided as A
¼
[60, 80]. Figure 11.37a shows the char-
acteristic function for the sets A, B, and C. The value of the function is either 0 or 1, depend-
ing on whether the age of a person is within the bounds of set A, B, or C. The scheme using
crisp sets lacks flexibility. If a person is 25 years old or 37 years old, he or she is not
categorized.
If the age groups were instead divided into fuzzy sets, the precise divisions between the
age groups would no longer exist. Linguistic variables, such as young, middle-aged, and
old, could be used to classify the individuals. Figure 11.37b shows the fuzzy sets for age
categorization. Note the overlap between the categories. The words are basic descriptors,
not precise measurements. A 30-year-old woman may seem old to a 6-year-old boy but quite
young to an 80-year-old man. For the fuzzy sets, a value of 1 represents a 100 percent degree
of membership to a set. A value of 0 indicates that there is no membership in the set. All
numbers between 0 and 1 show the degree of membership to a group. A 35-year-old person,
for instance, belongs 50 percent to the young set and 50 percent to the middle-aged set.
As with crisp sets from classical mathematics, operations are also defined for fuzzy sets.
The fuzzy set operation of intersection is shown in Figure 11.38a. Figure 11.38b shows the
¼
[0, 20], B
¼
[30, 50], and C
¼
