Diabetes Advisor - A Medical Expert System For Diabetes Management

Diabetes Advisor – A Medical Expert System
for Diabetes Management
Audrey Mbogho1 , Joel Dave2 , and Kulani Makhubele

1
Department of Computer Science, University of Cape Town
audrey.mbogho@uct.ac.za
2
Division of Endocrinology and Diabetic Medicine, University of Cape Town
joel.dave@uct.ac.za
Abstract. Access to medical services in rural communities, especially

in the developing world, is extremely limited. Medical expert systems
can play a significant role in alleviating this problem by providing deci-
sion support in the giving of advice on diagnosis, treatment and disease
management. This study built a prototype for diabetes, a chronic illness
affecting millions across the globe. Preliminary evaluation suggests that
such a system could be useful for expanding medical services in rural
communities and as an educational tool for unskilled medical staff.
Keywords: expert systems, diabetes, decision support
1 Introduction
Diabetes is a serious, life-threatening, chronic disease affecting over 300 million
people [1]. It is estimated that this figure will reach 366 million by 2030 [2] with
81% of these diabetics being in developing countries,where medical care remains
severely limited.
Fortunately, diabetes can be managed very effectively through healthy lifestyle
choices, primarily diet and exercise. Furthermore, people who are at high risk can
avoid developing the disease also through diet and exercise. This paper reports
on a prototype expert system for the provision of medical advice on diabetes
which can be used at home or at a rural clinic.
2 Expert Systems
An expert system is a computer program that emulates the decision making abil-
ity of a human expert [3]. The expert system developed in this project, Diabetes
Advisor (DA), is designed to give advice to patients with diabetes. DA can also
be used to educate people without diabetes in order to raise awareness about the
disease and to enable people to assess their own risk and take appropriate action.
In addition, DA can be used as an educational tool by unqualified medical staff
to learn more about diabetes and become better prepared to attend to patients
safely and effectively.
2 Audrey Mbogho, Joel Dave, and Kulani Makhubele
Expert systems can provide decision support or can be used as decision mak-
ers [4]. In the decision support role, expert systems assist the clinician. This can
reduce the cost of healthcare for the patient because the clinician can consult
with the expert system in the same visit instead of referring the patient to a
specialist. In rural communities, an expert system can provide decision support
or can be the decision maker if medical staff are unqualified or completely ab-
sent. In the decision making role, expert systems would be used in place of the
clinician. While this might be seen as risky, the risk from getting no medical
attention at all is greater than that posed by getting it from a computer.
2.1 Expert Systems for Diabetes
In [5], Hernando et al. described DIABNET, an expert system whose purpose

was to assist clinicians in treating patients with gestational diabetes. As this
form of diabetes tends to be temporary, patients do not have enough time with
the disease to learn how to manage it, and frequent visits with clinicians are
necessary. DIABNET aimed to reduce the frequency of such visits by giving
recommendations on insulin dose adjustment and meal planning. In addition,
DIABNET could monitor patient data continuously and alert clinicians when
unusual readings appeared, allowing clinicians to attend to patients remotely.
Rudi et al. proposed a similar expert system [6] for giving diabetes management
advice as well as storing clinical data and enabling doctors to remotely monitor
patients. In [7], Garcia et al. described ESDIABETES, an expert system for
helping patients monitor and control their blood glucose level. These systems
suggest that diabetes management is particularly amenable to expert system
intervention.
3 Development of DA
DA is a medical expert system prototype for medical advice provision on diabetes

built using the JESS expert system shell [8]. DA is designed as a menu based
interactive system and the language is simplified so that novice users can easily
understand it. It gives advice at various points in the course of the consultation
to avoid giving a large amount of advice at the end of the consultation as this
could overwhelm and confuse the user. People might not have the patience to
read so much information at once. The system gives some information for each
symptom to the user during the consultation process and thus helps the user
to gain more understanding of diabetes. This is an advantage over consulting
a human because doctors do not normally have time to explain their reasoning
to each patient. The final output of a DA session is a summary of the main
recommendations that a patient will need to take note of and may tell the
patient to visit a clinic or to see a doctor.
Diabetes Advisor 3
3.1 Knowledge Engineering

Acquiring high quality knowledge and validating it can be extremely time-
consuming. Converting it into a form suitable for a computer program (knowl-
edge representation) adds another layer of complexity. In this project, knowledge
acquisition occurred through meetings with the diabetes expert in which notes
were written by hand and conversations were audio-recorded. This knowledge
was converted to IF-THEN rules as required by JESS.
3.2 System Development Methodology

An iterative system development approach was used. After the first meeting with
the diabetes expert, the first prototype was designed. This prototype gave advice
on exercise plans, eating plans and monitoring sugar level. The system first asked
the patient about the symptoms or warning signs they were experiencing. A
second meeting with the expert revealed certain flaws and what improvements
were needed, as follows: the system was not able to consider all aspects relevant
for decision making; advice was only provided at the end of the consultation;
the system was not giving enough advice to patients; the language used was not
simple enough for people with poor literacy; the system asked for unnecessary
information that was never used in formulating advice; the system assumed only
people with diabetes would use the system (this assumption was removed in the
final system). The expert also provided magazines, leaflets and other documents
to aid in the implementation of the improvements identified in the first prototype.
A second prototype was designed incorporating the suggestions made. Below are
some of the questions the system asks the user.
Do you experience increased thirst? (y/n)
Have you experienced unexplained weight loss? (y/n)
How often do you eat fruits and vegetables? (0-9)
How many meals do you eat per day? (0-9)
Do you smoke? (y/n)
Do you consume alcohol? (y/n)
Do you monitor your blood sugar level? (y/n)
Do you experience unexplained weakness and exhaustion? (y/n)
The final system had a simple text-based user interface which provided entry
into the following use cases.
1. Receive diabetes advice
2. Learn more about diabetes
The next section gives two scenarios of the sytem in use.
3.3 Use Case for Diabetes Advice

Patient A, who has diabetes, provides information to DA. This information is
represented as facts within DA. These facts, then, cause the appropriate rules
to fire, and these rules in turn generate output for the user which is tailored to
her needs. Different ouput would be generated for someone without diabetes.
4 Audrey Mbogho, Joel Dave, and Kulani Makhubele
Patient A is a 20 year old woman, who has been living with diabetes for 2
years. She does not know enough about diabetes management and has been
taking insulin to treat her diabetes. DA responds with various recommendations
as the interaction proceeds. Some of the recommendations given to patient A are
shown in the listing below. (Intervening prompts by the system are not shown.)
Advice for Patient A: Now that you have diabetes, you must make sure that
you manage your diabetes properly. Please remember to take your medication
as your doctor tells you. Remember that you must not use the same place for
injection every time because your medication can stop working. It is good that
you have a healthy eating plan. Remember to eat your meals in small portions.
3 meals a day with snacks in between your meals is advisable. High fibre food:
Eat 2 to 3 portions of this type per meal.
3.4 Use Case for Diabetes Information

If the user selected to learn more about diabetes, the system displays a menu of
eight topics covering what diabetes is, its differnt forms, its causes, symptoms,
treatment, prevention, medication and complications. We consider this to be
an equally, if not more, important feature of the system because getting people
educated about the disease can help them keep it at bay.
4 Evaluation
The diabetes expert used in developing the system checked the knowledge and
behaviour of the first prototype and gave feedback, including suggestions for
improvement. It was also important, as recommended in [5], to use other evalu-
ators who had not taken part in the development of DA. Therefore four medical
students nearing completion of their degree were used as additional domain ex-
perts. They were asked to interact with the final prototype and to give feedback
via a questionnaire. The questionnaire contained desirable properties in such a
system, and the respondents were required to use a Likert scale to rank how
strongly these were present in DA. The feedback was largely positive; respon-
dents’ median response was Agree to 7 of the 9 statements below and Strongly
Agree to Statements 7 and 8; these were the two highest ranks, the others being
Neutral, Disagree, Strongly Disagree and Not Applicable. This indicated that
the system was useful and met its intended goals.
1. A person with no computer skills can be able to use this system.

2. The clinics or hospitals can use this system to learn more about the required
expertise.
3. This system can free physicians from boring routine tasks.
4. This system can be very useful to physicians or nurses in rural areas.
5. This system does look at all branches needed to be considered by a physician
while giving advice.
Diabetes Advisor 5
6. Even though I am a medical student I can rely on this system for advice
instead of going to a specialised doctor.
7. I would recommend this system to my diabetic patient if necessary.
8. The advice provided by the system is correct and useful.
9. The advice provided by the system can be understood by people with poor
literacy.
5 Conclusion and Future Work

In this study, a prototype expert system for the provision of advice on diabetes
was constructed. It was evaluated and found to be useful. As this was only a pre-
liminary phase of a planned larger study, the evaluation was not very extensive.
A larger number of domain experts are needed to more fully evaluate the system
and to confirm the positive results seen here. Secondly, field testing is critical
before actual deployment. That is, the system needs to be tested with a random
sample of actual rural patients so that its efficacy in its intended environment
can be confirmed.
Another important improvement that is needed is in the user interface. The
DA prototype has a basic textual interface, which falls short in terms of usability.
A graphical user interface or a speech-based interface would be more desirable,
and this is included in the plan for future work. In addition, a system like this
one could be made accessible to more people if deployed as a Web application.
This is also a planned extension to the work in the near future.
References
1. International Diabetes Federation, http://www.idf.org
2. World Health Organisation. Diabetes Program, http://www.who.int/diabetes
3. Jackson, P.: Introduction to Expert Systems. Addison Wesley (1999)
4. Metaxiotis, K., Samouilidis, E.: Expert Systems in Medicine: Academic Illusion or
Real Power? Inf. Manag. Comp. Sec. 8(2), 75–79 (2000)
5. Hernando, M., Gómez, E., Corcoy, R., del Pozo, F.: Evaluation of DIABNET, a
Decision Support System for Therapy Planning in Gestational Diabetes. Comp.
Meth. Prog. Biomed. 65, 235–248 (2000)
6. Rudi, R., Celler, B.: Design and Implementaiton of Expert-Telemedicine System
for Diabetes Management at Home. In International Conference on Biomedical and
Pharmaceutical Engineering, pp. 595–599. Research Publishing Services (2006)
7. Garcia, M., Gandhi, M., Singh, T., Duarte, L., Shen, R., Dantu, M., Ponder, S.,
Ramirez, H.: Esdiabetes (An Expert System in Diabetes). In: Meinke, J. (ed.) J.
Comp. in Small Coll. 16(3), 166–175 (2001)
8. Friedman-Hill, E.: Jess in Action: Java Rule-Based Systems. Manning Publications,
Greenwich, CT (2003)

Diabetes Advisor - A Medical Expert System For Diabetes Management

Uploaded by

Copyright:

Available Formats

Diabetes Advisor - A Medical Expert System For Diabetes Management

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Diabetes Advisor - A Medical Expert System For Diabetes Management

Uploaded by

Copyright:

Available Formats

Diabetes Advisor – A Medical Expert System

for Diabetes Management

Audrey Mbogho1 , Joel Dave2 , and Kulani Makhubele

Abstract. Access to medical services in rural communities, especially

Keywords: expert systems, diabetes, decision support

2.1 Expert Systems for Diabetes

In [5], Hernando et al. described DIABNET, an expert system whose purpose

DA is a medical expert system prototype for medical advice provision on diabetes

3.1 Knowledge Engineering

3.2 System Development Methodology

3.3 Use Case for Diabetes Advice

3.4 Use Case for Diabetes Information

1. A person with no computer skills can be able to use this system.

5 Conclusion and Future Work

You might also like