# APPLICATIONS OF FUZZY TECHNIQUES IN HEALTH CARE

# APPLICATIONS OF FUZZY TECHNIQUES IN HEALTH CARE

Fuzzy set model was established by Zadeh [1], the possible medical entities are defined as fuzzy sets. It provide a linguistic method which denotes an outstanding estimate to medical texts. Fuzzy logic provide cognitive approaches capable of building suitable references. The details recommend to improve the computerized diagnosis system and the fuzzy set theory is base for the system. Present advancement & applications of certain medical expert systems are based on fuzzy set theory and fuzzy logic [2]. This blog presents an idea as to how fuzzy technique is made use in health care problems.

A. General Technique

In the field of medicine, imprecision and ambiguity play a significant role, which is why the medical field has become one of the fruitful fields for applying fuzzy set theory and proof theory. With the increased amount of information available to physicians, the process of classifying different symptom sets under one name and determining a therapeutic action is becoming increasingly complex at different phases. The doctors typically collect medical information from the history of the case, physical examination, laboratory test results and other forensic techniques such as EEG, ECG, X-Rays and ultraviolet photographs. The knowledge provided by each of these sources carries various degrees of uncertainty with it. Hence the exact natural and pathological borderline state is always uncertain. Therefore a doctor who has only a small degree of accuracy will know the patient’s condition. The desire to better understand and conduct the medical diagnosis has led designers to model the system based on fuzzy sets.

B. METHODOLOGY

Generally, the limits of the fuzzy set are not defined clearly, for the case, clinicians classify a patient’s condition in indefinite terms very extreme, extreme, minor, subnormal, and regular, middle pathology, etc. Such words can be dignified as fuzzy sets specified in terms of language, so a patient’s condition may belong to a category or may not belong to that category or may fit into that category to a definite extent. The membership function of fuzzy set values ranges from 0 to 1. When the membership feature close to 1 specifies that the patient is still listed in the specified diagnostic group with a high degree of reliability, while a membership function value near to 0 specifies a low degree of reliability [3].

C. DEFINITION OF PROBLEM ELEMENTS

The problem elements are separated into two sets.

- The set S of patient states, which is the range between normal and pathology and
- The set E of experimental constraints of the analysis such as cerebral blood flow, EEG signal parameters etc.

The challenge is to fit the elements of set E, which the physician typically casts in inaccurate or fuzzy terms, to the elements of set S. In that end, the set S is split into imprecise subsets, namely fuzzy subsets. In the case study, the elements of set S are normal, neuropathic, diabetic and acute, which corresponds to the usual way of linguistic terms classification of patients [4]. Succeeding the features of set E must be clear. These features denote the experimental values of detected parameters, for example cerebral blood flow.

D.REPRESENTATION OF MEDICAL KNOWLEDGE

In any medical diagnosis expert system, the representation of medical knowledge plays a vital role. According to Adlassnig [1]. CADIAG-2 deliberates 4 classes of medical units

- Indications, signs, test results and discoveries (Si)
- Diseases and diagnoses ( Dj)
- Intermediary combinations (Ick)
- Signs combinations (Sci)

The succeeding relationships between medical entities are considered in CADIAG-2;

- Symptoms- Disease relations (Si Dj)
- Symptom mixture- Disease association ( Sci Dj)
- Symptom- Symptom relations (Si Sj)
- Disease –Disease relations ( Di Dj)

These relationships are considered by two parameters

- Frequency of incidence (o)
- Strength of Validation (c)

For a connection between medical entities X and Y (where X and Y may be symptoms, illnesses, or combinations of symptoms), the incidence frequency defines how often X happens when Y is present. Likewise, confirmatory power represents the degree to which X’s presence implies Y’s presence. Likewise, confirmatory power represents the degree to which X’s presence implies Y’s presence. The relationships among medical entities are provided in the form of relationship levels associated with tuples of relationships. The formalization of these laws in general is

If (antecedent) THEN (Consequence) with (o,c)

The relationship tuples (o,c) comprise either numerical values m_{o} and m_{c }or linguistic fuzzy standards l_{o} and l_{c} or both.

E.FUZZY EXPERT SYSTEM

Across an increasing number of application fields, Fuzzy expert systems are used successfully; usually, the structures are characterized using linguistic rules. A more suitable method for complex system problems is a rule-based system anywhere a mathematical explanation of the system is very difficult but not impossible. While constructing a fuzzy system one of the most significant considerations is the generation of fuzzy rules as well as the membership function for each set[5]. Experts in this field produce fuzzy rules with an increase in the number of variables.

Fuzzy modeling is the task of finding the parameters of a fuzzy inference system so that a preferred behavior is attained. With the direct method fuzzy model is built using information from a human expert. This task becomes tough when the existing knowledge is inadequate or when the problem space is very huge, thus inspiring the use of automatic methods to fuzzy modeling. One of the major difficulties in fuzzy modeling is the case of dimensionality, meaning that the computation requirements raise exponentially with the number of variables. The usually used membership functions are 3, 5, 7 or 9 fuzzy sets for every fuzzy variable.

F .CASE STUDY

Three different forms of case studies are selected to illustrate the fuzzy logic in medical diagnosis. The succeeding section of the paper discusses the case studies.

G “FUZZY BASED CLASSIFICATION OF PATIENT STATE IN DIABETIC NEUROPATHY USING CEREBRAL BLOOD FLOW”.

The case study chosen to exemplify the approach deals with the treatment of cerebral blood flow in diabetic neuropathy. The daily distribution of blood over the adult brain tissue averages 38 ml to 45 ml per 100 grams of brain per minute. This is 750 to 900 ml/min for the entire brain, or 15 percent of the total cardiac output remaining [6]. Due to higher glucose concentration in the bloodstream the blood flow increases and the severity of the disease increases. The neurons get damaged due to the higher concentration of glucose molecules. 4 initial linguistic classes are defined by a doctor for a diabetic neuropathy patient namely, normal, neuropathy, diabetic, and acute [8]. In our tactic, the fuzzy logic model has 2 inputs such as cerebral blood flow in the frontal region with and without activation. The output obtained is the classified patient state. A rule base of sixteen rules is developed. Mamdani triangular membership function is adopted. Centroid technique of defuzzification is used to find the crisp output. This fuzzy logic model is standardized using 50 diabetic cases and validated using 40 different cases of diabetic patients. The outcomes match very closely to the doctor’s analysis.

H.” FUZZY BASED DIAGNOSIS OF EPILEPSY FROM EEG SIGNALS”

This case study provides an idea of the Fuzzy-based classification of EEG signal parameters for defining the high-risk group of patients with epilepsy. The work consists of three stages of data gathering, extraction and level detection using fuzzy techniques. Three parametric features are take out from EEG signals namely energy of the epoch, the number of positive and negative peaks, and sharp spikes on all the sixteen bipolar channels of EEG recorder. The normal and abnormal cases can easily be classified by the feature extraction method. Fuzzy technique can be used to classify the level of risk associated with the grandmal or petimal type of epilepsy. A group of ten patients is selected with known clinical findings of grandmal and petitmal epilepsy. The Mamdani method which employs fuzzy technique is used to obtain the classification of epilepsy such as normal, low risk and high risk patients in the grandmal and petitmal types.

The Mamdani process is used for epilepsy classification, with two inputs and one output. The energy of EEG signal which is represented in the three linguistic levels of low, medium and high and the positive and negative peaks with the same linguistic levels are the two inputs. A rule base is developed with five rules. The same fuzzy technique is repeated for the sharp waveforms as one input and energy as the other input. The results of these approaches are compared. This classification performs poorly in the focal epilepsy cases because of the inconsistent EEG parameters.

I.DISCUSSIONS AND CONCLUSION

This research explores a analysis of the fuzzy logic of medical diagnosis in the sense of classifying a diabetic patient’s epilepsy risk level. One arbitrary choice is the number of linguistic groups carefully nominated in each case study. The performance of this fuzzy logic model is more than 75 percent and the other cases are constantly identified in the near cases, resulting in a false alarm, reduced false alarm rate, optimization methods are to be created.

REFERENCES

[1] K.P.Aldassnig, “ Fuzzy Set theory in Medical Diagnosis”, IEEE Transactions on Syst Man Cybern Vol.16, pp 260-265, March 1986.

[2] Yuhui Shi etal, “ Implementation of Evolutionary Fuzzy Systems”, IEEE Transaction on Fuzzy systems, vol 7, no:2, pp 109-119,April 1999

[3] Alison A.Dingle etal, “ A multi stage system to detect epileptic form activity in the EEG”, IEEE Transaction on Bio-Medical Engineering, vol 40, no:12, pp 1260-1268,December 1993

[4] Lucien Duck Stein etal, “ Fuzzy Classification of patient state with application to electro diagnosis of peripheral neuropathy”, IEEE Transaction on Bio-Medical Engineering, vol 42, no:8, pp 786-791, August 1995.

[5] Donna.l.Hudson, “Fuzzy Logic in Medical Expert Systems”, IEEE EMB Magazine 13(6), pp 693-698, November/December 1994.

[6] Arthur.C.Gayton, “Text Book on Medical Physiology”, Prism Books Pvt.Ltd, Bangalore, 9^{th} edition1996.

[7] G.Beliakov and John Warren, “Appropriate choice of aggregation operators in Fuzzy Decision systems”, IEEE Transaction on Fuzzy Systems, vol 9, No:6,pp 773-784, December 2001.

[8] Leo.P.Karall Md, “Joslin Diabetes Manual”. LEA & FEBIGER, Philadelphia London 1998.