Lung cancer is the second most common cancer in both men and women in the world. The focus of this paper is to design a fuzzy rule-based medical expert system for diagnosis of lung cancer. The proposed system consists of four modules: working memory, knowledge base, inference engine and user interface. The system takes the risk factors and symptoms of lung cancer in a two-step process and stores them as facts of the problem in working memory. Also domain expert knowledge is gathered to generate rules and stored in the rule base. The rule base consists of two different rule sets related to risk factors and symptoms of lung cancer respectively. Finally, type-2 fuzzy inference engine fires relevant rules under appropriate condition and provides the probability of disease as output of the system. The output of the system could act as a second opinion to assist the physicians. Also graphical user interface is presented to facilitate the communication between user and system.
Fuzzy rule based expert system for diagnosis of lung cancer
1. Fuzzy Rule based Expert System
for Diagnosis of Lung Cancer
Mohammad Hossein Fazel Zarandi
Farzad Vasheghani Farahani
Abbas Ahmadi
NAFIPS 2015
Redmond, WA | August 17 - 19
3. Objectives
The objectives we're looking in designing this “Graphical Fuzzy
Rule-based Expert System”:
A system for diagnosis of lung cancer that:
• Helps physicians as a second opinion
• Takes uncertainty and vagueness into account
• Is Fast and helpful
• Has convenience of the users
4. Procedure
A. Domain knowledge gathering
• Interviews with medical experts
• Books
• Websites
B. Patient information acquisition
• Risk factors
• Symptoms
• Other facts
C. Creating fuzzy system (Using fuzzy type-2)
• Calculating membership function
• Creating fuzzy rule bases
• Defuzzification
D. Graphical user interface (GUI)
How to achieve the desired objectives?
7. Background
Early studies of “artificial intelligence in medicine application”
started in the end of the 1960's and led to the emergence of
experimental systems such as:
• MYCIN
• INTERNIST
• ONCOCIN
• CASNET
• EXPERT
A rule-based expert system to
diagnose and recommend treatment
for certain blood infections
8. Background
Early studies of “artificial intelligence in medicine application”
started in the end of the 1960's and led to the emergence of
experimental systems such as:
• MYCIN
• INTERNIST
• ONCOCIN
• CASNET
• EXPERT
A rule-based expert system for the
diagnosis of complex problems in
general internal medicine.
9. Background
Early studies of “artificial intelligence in medicine application”
started in the end of the 1960's and led to the emergence of
experimental systems such as:
• MYCIN
• INTERNIST
• ONCOCIN
• CASNET
• EXPERT
A rule-based medical expert system
for oncology protocol management.
It was designed to assist physicians
in treating cancer patients receiving
chemotherapy.
10. Background
Decision Recommendation Solution
Expert systems (ES): derived from a branch of artificial intelligence
(AI) that provides expert advice such as:
In order to develop an expert system the knowledge should to be
extracted from domain expert and then it should be converted into
a computer program (knowledge representation).
11. Background
Fuzzy Expert Systems in medicine:
The process of diagnosis is difficult and is often associated with
uncertainty.
• For patients, it is often very difficult to explain what kind of pain
they feel
• The diagnostic decision depends upon experience, expertise and
perception of the practitioner
Fuzzy logic presents powerful reasoning methods
that can handle uncertainties and vagueness
14. Step 1
Step 1: The system takes risk factors of lung cancer from user as
crisp inputs. These risk factors include age and some binary
variables.
General
Information
- Age
- Gender
- Skin Color
Life style
- Smoking and secondhand smoke
- Exposure to asbestos or other pollutants
- Exposure to radon
Genetic
background
- Family history
15. Step 2
Step 2: A crisp score between [0,1] produced through binary
variables of risk factors. (All variables except age)
Note: In medical application such as diagnosis, these binary variables
affect the physician’s mind. This step solves the problem of binary input of
risk factors. Indeed, we could consider a score (value) for binary variables
in our system to differentiate patients from each other.
16. Step 3
Step 3: Here, after fuzzification of
age and score obtained in previous
step, the system returns the initial
suspicion to existence of lung cancer
in defuzzified form using centroid
method.
The output of this part is a linguistic
term that expresses the degree of
suspicion to the lung cancer.
17. Step 3 (Cont.)
The first rule set is as follow:
• IF PART:
Score 1 and Age
• THEN PART:
Initial suspicion (first output)
Note: The rule-base of system
consists of two rule sets and the
inference is done twice.
18. Step 3 (Cont.)
Centroid method is used for defuzzification which takes fuzzy set
as input and output is a crisp value. Output of first rule set (Initial
suspicion) is shown below:
19. Step 4
Step 4: Symptoms and signs of the disease must be embedded to
system through user. All of them are binary crisp values.
Primary
symptoms
- Shortness of breath
- Feeling tired
- Infections
- Wheezing
- Loss of appetite
- Consistent cough
- Chest pain
- Hoarseness
- Weight loss
- Coughing up blood
Secondary
symptoms
- Bone pain (like pain in the back or hips)
- Nervous system changes (such as headache,
weakness or numbness of an arm or leg,
dizziness, balance problems, or seizures)
- Yellowing of the skin and eyes (jaundice)
- Lumps near the surface of the body,
(collections of immune system cells)
20. Step 5
Step 5: A crisp score between [0,1] generated using binary variables
of patient’s symptoms like step 2.
Note: According to domain expert knowledge, five linguistic variables (Very low,
Low, Medium, High, Very high) are used based on Gaussian functions for both
scores in the problem about risk factors and symptoms as the same.
21. Step 6
Step 6: Fuzzy inference system activates again to produce
possibility of disease (final result) of the system through Mamdani
inference method.
22. Step 6 (Cont.)
The second rule set is as follow:
• IF PART:
Score 2 and Output of rule set 1
• THEN PART:
Possibility of disease (final output)
Note: The result obtained in the first
rule set (Initial suspicion) used as an
input for the second one.
23. Step 6 (Cont.)
Output of second rule set (Possibility of disease) is shown below.
centroid method is used for defuzzification.
24. Contribution
Apply fuzzy type-2 in order to create inference engine.
Solving the problem of binary variables exist in risk factors and
symptoms of lung cancer.
Provide an interaction between user and system through
graphical user interface (GUI) in MatLab.
25. Conclusion
• In present work , the manner of an expert in diagnosis process is
simulated.
• Fuzzy rules with binary premise and uncertain consequent are
modeled.
• The output of the system is a crisp value and it’s determined the
CT scan should be taken or not.
• Results of this system are close to a physician.
26. References
• “Cancer of the Lung and Bronchus - SEER Stat Fact Sheets.” [Online]. Available:
http://seer.cancer.gov/statfacts/html/lungb.html. [Accessed: 13-Feb-2015]. Additional
supporting data
• “What are the key statistics about lung cancer?” [Online]. Available:
http://www.cancer.org/cancer/lungcancer-non-smallcell/detailedguide/non-small-cell-
lung-cancer-key-statistics. [Accessed: 01-Mar-2015].
• M. H. F. Zarandi, I. B. Turksen, and S. M. M. Hoseini, “Architecture for Supply Chains,”
vol. 15, no. 5, 2008.
• M. H. F. Zarandi, M. Zolnoori, M. Moin, and H. Heidarnejad, “A Fuzzy Rule-Based Expert
System for Diagnosing Asthma,” vol. 17, no. 2, 2010.
• M. A. Ghahazi, “Fuzzy Rule based Expert System for Diagnosis of Multiple Sclerosis,” pp.
1–5, 2014.