Page 46 - Biomedical Engineering and Design Handbook Volume 1, Fundamentals

MODELING OF BIOMEDICAL SYSTEMS 23

Crisp parameter input

Fuzzification using membership functions

Membership values

Rule base

Output membership
values

Defuzzification

Crisp output
FIGURE 1.13 The overall scheme of fuzzy logic. The
measured crisp parameters are first fuzzified with the aid
of membership functions to obtain the membership value
to various subdomains. The membership values are then
subjected to a predefined rule base. The output of the rule
base is in the form of output membership values to various
subdomains. This fuzzy output is then defuzzified to
obtain a crisp output.

1.6.1 Fuzzy Classification of Risk for Aspiration in Dysphagic Patients
Suryanarayanan et al. (1995) developed a fuzzy logic diagnosis system to classify the dysphagic patient
into “normal, mild, moderate, and severe dysphagia” based on several parameters measured from the
dysphagic subject. Dysphagia denotes dysfunction of the swallowing mechanism and presents a
major problem in the rehabilitation of stroke and head injury patients. Dysfunction of the pharyngeal
phase of swallow can lead to aspiration, chocking, and even death. Consequently, the assessment of
risk for aspiration is important from a clinical point of view. Reddy et al. (1990, 1991, 1994) have
identified and developed instrumentation and techniques to noninvasively quantify various biome-
chanical parameters that characterize the dysphagic patient and clinically evaluated the technique by
correlating with the videofluorography examination (Reddy et al., 2000). For the assessment of the
pharyngeal phase, they have placed an ultra miniature accelerometer at the throat at the level of thy-
roid cartilage and asked the patient to elicit a swallow (Reddy et al., 1991, 1994). Swallowing in nor-
mal subjects gave rise to a characteristic acceleration pattern which was distorted or absent in
dysphagic individuals. In addition to the acceleration measurements, they measured swallow suction
pressure (with a catheter placed toward the posterior aspect of the tongue), and the number of
attempts to swallow before eliciting a swallow response, etc. Suryanarayanan et al. (1995) fuzzified
these measurements by defining membership functions for each of these parameters (magnitude of
acceleration, swallow pressure, and number of attempts to swallow) which defined the four subdo-
mains (severe risk, moderate risk, mild risk for aspiration, and normal) for each of these parameters
(Fig. 1.14). Membership functions were constructed using the function
μ= 1/(exp(αx +β) (1.61)
where μ is the membership value, x is the measured parameter, and α and β are constants. The slope
of the sigmoid function can be changed by changing the value of α.