• Nenhum resultado encontrado

EFFECTIVE ANALYSIS AND DIAGNOSIS OF LUNG CANCER USING FUZZY RULES

N/A
N/A
Protected

Academic year: 2017

Share "EFFECTIVE ANALYSIS AND DIAGNOSIS OF LUNG CANCER USING FUZZY RULES"

Copied!
7
0
0

Texto

(1)

EFFECTIVE ANALYSIS AND DIAGNOSIS OF LUNG CANCER

USING FUZZY RULES

1

M.A.Saleem Durai 2 N.Ch.S.N.Iyengar 1

Assistant Professor (SG), School of Computing Sciences and Engineering, VIT University, Vellore-14 2

Senior Professor, School of Computing Sciences and Engineering, VIT University, Vellore-14 Email: masaleemdurai@vit.ac.in, nchsniyengar48@gmail.com

ABSTRACT

In this paper, the structure of a lung cancer analysis system is presented. The main focus for the development of the system is on the architecture and the algorithm used to find the probable disease, stage and the appropriate treatment of cancer a patient may have. The disease is determined by using a rule base, populated by rules made for different types of lung cancer. The algorithm uses the output of the rule base (i.e. the disease name) and the symptoms entered by the user to determine the stage of cancer the patient is in. Both these results (disease name and stage) help the diagnostic logic to determine the treatment for the patient with accuracy. Our diagnosis does a complex analysis of all the information gathered about our symptoms. In this paper, we have evolved a method of choosing the best treatment for cancer using fuzzy decision making techniques.

Keywords: Lung Diseases, Diagnostic System, Fuzzy Rule Base, Threshold Value and Membership Value.

INTRODUCTION

Lung diseases are the second most dangerous disease in the world (for example: lung abscess, lobe pneumonia, bronchitis...). However there are many effective medicines which can treat these lung diseases, the danger is that the patients always ignore some initial symptoms of lung diseases and they are only afraid when those symptoms are likely clear. But at that time, it seems to be more difficult to treat those diseases.

In diagnosing lung diseases, doctors have to cope with many difficulties, the patient’s symptoms are usually not clear; the similarities in some lung disease’s symptoms are difficult to distinguish. Doctors always have to test many times before making a decision. So the diagnosis result depends on not only patient’s symptoms but also the doctors' experiences. Wrong decision means wrong treatment and the patient would suffer more.

PROPOSED WORK

In this paper, we will describe the technique used to determine the disease name, stage and the diagnostic treatment using the algorithms and a rule base (developed specifically for a disease).

Here the physician will take the symptoms of a patient as input and based on his symptoms proper diagnostic treatment will be prescribed to a patient. Each symptom has its own priority value and physician will assign membership values to the linguistic variables i.e. slight, low, medium, high and critical.

The algorithm developed uses the rule base to determine the disease name based on the symptoms entered by the user. This rule base consists of rules developed for a particular type of cancer (in this case lung cancer) and determines the type of lung cancer a patient has.

Based on the severity of the symptoms entered the algorithm calculates the total membership of the symptoms and the threshold value for this total membership. Then it creates the ranges which represent the different stages of cancer such as probable, low, medium, high and critical.

(2)

for different type of lung cancers. These rules take into account the disease and the cancer stage determined by the algorithm to prescribe the treatment.

Architecture 1:

Sample View of the Table

Patient ID

Name S1 S2 S3 S4 .. .. .. .. .. .. .. S30 S31 Output

001 Bob .5 .25 1 .75 .. .. .. .. .. .. .. - - Bronchitis(L)

002 Rob .25 .25 .75 1 .. .. .. .. .. .. .. .5 .25 Pleural

Tumor(P) User 

Query 

Fuzzy Logic Decision 

Threshold  Calculation 

Fuzzy  Rule 

Output 

Disease 

(3)

Architecture 2:

User   Interface 

        Input 

         

         Find Rule Match 

      

      Output   

Fuzzy Rule Base 

  

Priority Database 

       

(4)

Combined Architecture:

Algorithm 1:

1.) Take input from user as symptoms and their severity.

2.) Analyze symptoms using IF-Then rules (consisting of sets of symptoms) for possible disease (individual rule for separate diseases.)

Example:-

 IF (Cold Like Symptoms, Fever, Chills, Throat Pain, Pneumonia, Wheezing) THEN Disease = Bronchitis.

 IF (Dry Cough, Wheezing, Shallow Breath, Chest Pain) THEN Disease = Bronchial Asthma.

3.) Calculate threshold value on basis of symptoms entered. User 

Query 

 Logic Decision         

       Disease   

       

         Stage    Threshold 

Calculation 

(Algorithm) 

Output  IF‐THEN 

Rules 

Data  Base  Logic for 

(5)

Threshold U = No. of Symptoms * Total No. of Stages. Lower Value L= 0.

4.) Calculate total membership as:

mv = (mv * mv)/(No. of Stages/4)

Where mv = sum of severities of all symptoms.

5.) Calculate disease stage using threshold value and the severity value of the symptoms (using the intervals created based on the symptoms.)

t0 = (l + ((u) / 14)); // No Cancer t1 = (l + ((u) / 12)); // Probable Stage t2 = (l + ((u) / 10)); // Low Stage t3 = (l + ((u) / 8)); // Medium Stage t4 = (l + ((u) / 6)); // High Stage t5 = (l + ((u) / 4)); // Critical Stage

6.) Store the input and the result (disease and stage) into database and display result to the user.

Algorithm 2: STEP1:

Select n number of symptoms i.e. Population or Operational data Assign some priorities values (weight) to the symptoms

For eg …… psymptom i=pi , where i=1 to n STEP2:

Now choose the severity for symptoms and assign some membership values to the linguistic variables : Very low=0, low=0.25, medium=0.5, high=0.75, very high=1

Symptom i= very low||low||medium||high||very high where i=1 to n STEP3:

Calculate the total value i.e support

Total=

(6)

Highvalue

Lowvalue STEP5:

Now calculating the membership function i.e fuzzy value(FV) FV=Total value/Highvalue(total weight)

(Or) we can say Support(Highvalue)=FV STEP6:

Now dividing into N ranges between high and low values B (i+1) = lowvalue + (k * highvalue / N)

where k = 0 to N and I =0 to N. STEP7:

Now Check

For each i=0 to N-1

If(total >= b(I + 1) && total < b(I + 2)) Then Stage=”Sj” where j=1 to N+1

Else if (total==b(i+2)) Then stage=”N+2” STEP8:

Now check

For each j=1 to N+1 if(stage.equals("Sj"))

{ cancertype="type j"; treatment="treatment j"; }

else{

cancertype="type N+2"; treatment="treatment N+2"; }

ANALYSIS

(7)

are highly accurate and provide information in a wide range of formats (text, images, clusters etc.) but are very complex and difficult to implement. The system proposed in this paper is very simple and the implementation is very easy.

This system is compatible with other techniques as the database can be used with case based reasoning and other mining methods. This capability of the system increases the scope and age of the system.

Computational study in medical diagnosis is still in its infancy stage and simpler but accurate systems are the need of the day.

CONCLUSION / FUTURE WORK

This system is useful to the physician as well as the user for determining the type of lung cancer and the stage of the cancer. This system is based on a simple algorithm and hence the accuracy is low. The accuracy can be increased by implementing more analysis techniques on the same database used in this system along with the current algorithm. The capability of this system for easy modification and continuous up gradation of the database is a plus point and increases the scope and life of the software.

REFERENCES

[1] LDDS – A Fuzzy Rule Based Lung Diseases Diagnostic System Combining Positive and Negative Knowledge. Nguyen Hoang Phuonga, Le Huu Nghia, Hune Chob,Yun Sik Kwakb

[2] Discovering human understandable fuzzy diagnostic rules from medical data Giovanna Castellano, Anna Maria Fanelli and Corrado Mencar Department of Computer Science University of Bari.

[3] Armato,S.G.,Giger,M.L.,&Macmahon,H.(2001).Automated detection of lung nodules in CT scans.Medical Physics,28,1552-1561. [4] Data Mining: Medical and Engineering Case Studies. ByA. Kausiak, K.H. Kernstine, K.A. McLaughlin and T.L. Tseng

[5] Abe,H.,Ashizawa,K.,Katsuragawa,S.,MacMohon,H.,&Doi,K.(2002).Use of an artificial neural network to determine the diagnostic value of specific clinical and radiologic parameters in the diagnosis of interstitial lung disease on chest radiographs.Academic Radiology,9,13-17.

Referências

Documentos relacionados

Initially, the number of symptoms increased with disease progression (Increasing Symptoms Curve), followed by a constant presentation of symptoms (Symptoms Plateau) and a

Figueiredo, Abel (2009) „The Combat Sports in Physical Education Classes – A Basic Perspective‟, in: Cynarski, Wojciech (ed.) Martial Arts and Combat Sports – Humanistic

Since we are using generic product data we will compare mature and emerging market positive association rules results using the main indicators of the rules instead of the rule

aimed to analyze lower urinary tract symptoms (LUTS) in a restricted cohort establishing a possible inluence of depressive symptoms and psychoactive drugs (namely

Foi instalado no ano de 1970, em Pelotas e Passo Fundo, Rio Grande do Sul, um ensaio em que se comparou o comportamento de 15 linhagens de triticale, introduzidas do México, e

Como objetivos principais para uma gestão e sustentabilidade financeira do setor destacam-se a melhoria da qualidade do abastecimento de água e saneamento de

Numa última análise, a partir dos resultados do coeficiente de correlação r de Pearson das variáveis maus tratos (risco de abuso e os indicadores de abuso nos idosos),

relating disease duration with severity of motor symptoms, balance impairment and func -..