You are on page 1of 4

Fuzzy Logic for Child-Pugh classification of

patients with Cirrhosis of Liver


Anu Sebastian Surekha Mariam Varghese
Dept. of CSE Dept. of CSE
Mar Athanasius College of Engineering Mar Athanasius College of Engineering
Kothamangalam Kothamangalam
anoos87@gmail.com surekha.laju@gmail.com

Abstract—Survival analysis is a common procedure in the


medical field. Predicting life expectancy is a highly significant In addition to that the Regression analysis is not capable of
factor in the decision making of both patients and doctors. In processing fuzzy data. Fuzzy logic is an intelligent technique
addition to the correct diagnosis of the ailment, accurate used to model imprecise, non-linear, complex system by
classification and assessment is necessary to commence implementing human experience, knowledge and practice as a
appropriate and effective treatment. The Child-Pugh set of inference rules. In recent years the fuzzy logic has been
classification has been used for many years for the assessment of
used in various systems worldwide. The application of fuzzy
severity and prognosis of chronic liver disease and cirrhosis of
liver. It is considered as a standard for classifying the patients
logic has a marked progress in several fields of science. These
with end–stage liver disease. Fuzzy logic is a clever technique that include automated control: temperature control, speed control,
can be used to model imprecise and complex systems efficiently. auto focus control of video cameras, form recognition,
Intelligent procedure based on Fuzzy logic could be used for the cognitive psychology etc. Fuzzy logic is the procedure of
Child-Pugh classification of the patients suffering from liver mapping with a given set of input variables and output through
cirrhosis. For different classes of cirrhosis patients the life a set of inference rules [9].
expectancy also differs. This paper presents a fuzzy intelligent Fuzzy logic can be used to classify the liver cirrhosis
technique to classify and predict the life expectancy of patients patients based on the Child-Pugh score. The Child-Pugh score
with cirrhosis of liver. consist of five clinical features. Three of which asses the
Index Terms—Fuzzy logic; Child - Pugh classification; End- synthetic function of the liver, it include total bilirubin level,
Stage Liver Disease; Life Expectancy Prediction. serum albumin and international normalization ratio or INR.
The other two factors are based on the clinical assessment, it
include degree of ascites and degree of hepatic
I. INTRODUCTION encephalopathy. Bilirubin is the yellow break down product
caused by the body’s clearance of aged red blood cells.
Liver Cirrhosis is a major health problem worldwide. It is Bilirubin levels elevate as cirrhosis progresses. Serum albumin
the 12th leading cause of death by disease [1]. The disease is the most abundant blood plasma protein and is produced by
progresses gradually. In cirrhosis patients healthy liver tissue liver. Albumin level falls as synthetic function of the liver
is replaced by scar tissue, ultimately stopping liver from declines with worsening cirrhosis.
functioning correctly. Normal liver filters the blood coming International Normalized Ratio (INR) is used to determine
from the digestive tract, before passing it to the rest of the the clotting tendency of blood. Patients with liver cirrhosis
body. It detoxifies chemicals and metabolizes drugs. Liver have an increased blood coagulation time and have a high risk
secretes bile that aids in intestinal function and it is a of internal bleeding [6]. Ascites is the accumulation of fluid
mechanism to remove waste from the body. Liver produces in the body especially in the peritoneal cavity causing
proteins significant for clotting of blood and other functions. It abdominal swelling. Cirrhosis is the most common cause of
regulates several hormones. The major complications of liver ascites [5]. Encephalopathy is the occurrence of confusion,
cirrhosis include ascites, hepatic encephalopathy, portal altered level of consciousness and coma as a result of liver
hypertension, variceal bleeding, less total protein level etc [2]. failure. All these factors are counted for the calculation of the
It can be clinically diagnosed with many tests. In addition to Child-Pugh score. Based on the Child-Pugh score the patients
the correct diagnosis, accurate classification and assessment is can be classified into three categories. Class A, class B and
necessary to commence appropriate and effective treatment. class C. Patients belonging to different classes have different
Child-Turcotte-Pugh (more commonly known as Child-Pugh) chances of survival. Fuzzy implementation of this
scoring system is a risk stratification method for liver cirrhosis classification is an efficient method to categorize new patient
[4]. Child-Pugh classification has been used for many years record.
for the assessment of severity and prognosis of chronic liver
disease and cirrhosis of liver.
II. MEDICAL BACKGROUND
There are different techniques exist to predict an output
variable with different input variables. A commonly used The factors on which the Child-Pugh classification is
technique is regression analysis. Any regression analysis based on and its score is listed in the table1 [4]. Based on the
involves a set of assumptions such as linearity, normality etc. score for each of these factors the classification is made. If the
total score for a patient is 5 to 6 points then he belongs to the
class A.
978-1-5090-1987-8/16/$31.00 ©2016 IEEE

168
If the score is 7 to 9 then he belongs to the class B and The fuzzification and the defuzzification processes are the
belongs to the class C if the score is in the range 10 to 15. The two interfaces between fuzzy logic system and the measured
survival in patients based on the Child-Pugh classification is phenomena. Fuzzification is used for the transformation of
depicted in the fig.1. original data to a fuzzy value using membership functions,
fuzzy set can therefore, unlike a crisp one, be represented by
an infinite number of membership functions. The
TABLE 1 CHILD-TURCOTTE-PUGH CLASSIFICATION defuzzification is the inverse process that transforms the
output fuzzy value (computed by the set of inference rules)
into a crisp real value [11]. Fig.2 displays the overall
Points*
Clinical and Lab Criteria organization of a fuzzy inference system.
1 2 3
Encephalopathy None Mild Severe

Ascites None Mild Severe

Bilirubin(mg/dl) <2 2-3 >3 Fuzzy Rule Base


Albumin(g/dl) >3.5 2.8-3.5 <2.8
International Normalized
<1.7 1.7-2.3 >2.3
Ratio(INR)

*Child-Pugh classification is obtained by adding score for each Input Fuzzif Defuzzi
Parameter Output
Ication fication
Class A = 5 to 6 points(least severe liver disease )
Class B = 7 to 9 points (Moderately severe liver disease)
Class c = 10 to 15 points most severe liver disease

Fuzzy Inference Engine


Cirrhosis patients belonging to different classes have
different chances of survival. The survival percentages of
patients belonging to class A, class B and class C for one year
are 95%, 80% and 40% respectively. The survival percentages Fig. 2. Fuzzy Inference system
of patients belonging to class A, class B and class C for two
years are in the order 90%, 70% and 38%.
Fuzzy inference system maps a given input to an output
using fuzzy logic’s _‘if-then’_ type fuzzy rules. Fuzzy _’if-
then’_ rules are formulated by fuzzy sets and relations. If a set
of conditions are satisfied, a set of consequence can be
inferred. If encephalopathy is severe and INR is between the
ranges 1.7 to 2.3 then the patient belongs to the Child-Pugh
class C is an example for the fuzzy if-then rule. Here the first
part of the rule that starts with _‘if’_ and ends before _’then’_
is referred to as the predicate (premise, antecedent) which
combines the subset of input variables. The second part of the
rule that comes after _’then’_ is called consequent and it
contains subset of fuzzy output based on the premise part. The
most important types of fuzzy inference methods are Mamdani
and Sugeno.
The defuzzification procedure transform the fuzzy output
values of a fuzzy inference system to actual crisp values. Fist a
Fig. 1. Survival in patients based on Child-Pugh Classification typical value is calculated for each term in the linguistic
variable. Then the best compromise is determined by
balancing out the results using different approaches like:
III. FUZZY LOGIC
 x Centre of sums. 
Fuzzy logic brings a paradigm for working with graduation  x Centre of area 
and ambiguity described by linguistic expressions [8].  x Bisector of area 
Although fuzzy logic is relatively young theory the area of its  x Mean of Maxima 
application are very wide. Excellent result and easy  x Random Choice of maximum 
implementation of fuzzy logic have been demonstrated during x Weighted average method 
recent years. By using fuzzy logic human experience can be 
x First (or last) of maxima etc. 
expressed as a set of rules. It is the actual process of mapping
with a set of input variables and output using a number of
inference rules. MATLAB’s fuzzy logic toolbox provides an environment
for analysing, designing and simulating a system based on the

2016 International Conference on Information Science (ICIS)


169
fuzzy logic. Using this toolbox complex systems can be The rule base of the fuzzy Child- Pugh stratification system
modelled by simple logic rules and these rules can be consists of several rules based on different inputs, for correctly
implemented by fuzzy inference system. classifying a new data set.
The system can be used to classify new patients with a new
IV. CHILD-PUGH STRATIFICATION SYSTEM set of record, so that the survival chances for the patient can be
calculated using the figure 1 based on the patient’s cirrhosis
In our work the Child-Pugh classification scheme is severity class. In MATLAB the fuzzy logic tool kit consist of
converted to a fuzzy system using the fuzzy logic toolbox in the rule viewer which can be used to adjust different input
MATLAB. The fuzzy Child-Pugh stratification system has factors – Encephalopathy, Ascites, Bilirubin, Albumin and
five inputs as shown in table1. And the output for the system INR for predicting the severity of liver disease (Child- Pugh
is the child-Pugh class for the corresponding data set, it can be class) for the patient. Fig.5 shows the rule viewer for the
one among the Child-Pugh classes’ class A, class B or class C. system. Using the Rule viewer new patient data can be
Any suitable membership function can be used for classified.
representing the inputs and the output. Triangle membership
function and bell membership functions are used here.
Depending on the minimum error in predicted data,
membership function has been chosen. Triangle membership
function is used for the output and the bell membership
function is used for the input variables. We used mamdani
inference procedure for developing the system. The fuzzy rule
base contains the control rules. It is evaluated by the fuzzy
inference procedure. Fig.3 shows the inputs and output for the
fuzzy Child-Pugh stratification system.

Fig. 5. Rule Viewer for the Child-Pugh stratification system

V. RESULTS

Fuzzy System for the Child-Turcotte-Pugh classification is


thus formed and sample liver cirrhosis patients records are
classified with the system. Table 2 shows a set of sample data
and the Child–Pugh classes predicted by the developed fuzzy
Child-Pugh stratification system.

Fig. 3. Inputs and output for the Child-Pugh stratification system TABLE 2 SAMPLE PATIENT DATA CLASSIFICATION

The Child-Pugh score is calculated and the correct Encepha Bilirubi Albumin INR Chid-
lopathy Ascites n(mg/dl) (g/dl) Pugh
classification is done based on the table using several fuzzy Class
_’if-then’_ rules. Figure 4 shows the snapshot of the rule base
for the proposed system. None None 1 4 1 A

None Mild 3.5 2.4 2.5 C

Mild Mild 1 3.7 1 B

Severe Severe 3.5 2.5 2.7 C

Mild None 2.7 3.6 2 B

The system classified the sample data with 100% accuracy.


All the patients classified as class A, class B and class C
respectively have 95%, 80% and 40% chance of survival for
one year and 90%, 70% and 38% for two years.

Fig. 4. Rule base for the Child-Pugh stratification system

2016 International Conference on Information Science (ICIS)


170
VI. CONCLUSION
[5] van Erpecum KJ; Ascites and spontaneous bacterial peritonitis in
The Child-Turcotte-Pugh classification method for the patients with liver cirrhosis. Scand J Gastroenterol Suppl.
liver cirrhosis patients is implemented as a fuzzy logic system. 2006;(243):79-84.
Using this system a new patient record can be easily stratified [6] Kleinegris MC, Bos MH, Roest M, et al; Cirrhosis patients have a
into the corresponding Child-Pugh classes as class A, class B coagulopathy that is associated with decreased clot formation capacity. J
Thromb Haemost. 2014 Oct;12(10):1647-57. doi: 10.1111/jth.12706.
and class C and the life expectancy of the patient can be Epub 2014 Sep 30.
calculated. As per our knowledge this is the first time an [7] Heidelbaugh JJ, Sherbondy M; Cirrhosis and chronic liver failure: part
attempt is made to use fuzzy logic for the classification of the II. Complications and treatment. Am Fam Physician. 2006 Sep
liver cirrhosis patients. System is found to be accurate in 1;74(5):767-76.
classifying the patient record. [8] L.A. Zadeh. Fuzzy Sets. Information and Control, 8:338-353, 1965.
[9] H.T.Ngyuyen and E.A.Walker. A First course in Fuzzy Logic. Chapman
& Hall/CRC, Third Edition 2006.
REFERENCES
[10] Matlab Software, Accessed at http://www.mathworks.com/
[1] Cirrhosis- https://en.wikipedia.org/wiki/Cirrhosis . [11] Padhy,N.P,Artificial Intelligence and Intelligent Systems,Oxford
[2] Schuppan D, Afdhal NH; Liver cirrhosis. Lancet. 2008 Mar University Press,2006.
8;371(9615):838-51.
[3] Christensen E, Schlichting P, Fauerholdt L, et al. Prognostic value of
Child-Turcotte criteria in medically treated cirrhosis. Hepatology.
1984;4:430–435.
[4] Conn, HO. A peek at the Child-Turcotte classification. Hepatology.
1981;1:673–676.

2016 International Conference on Information Science (ICIS)


171

You might also like