Main Project 1 To 3 Edited in Church
Main Project 1 To 3 Edited in Church
Main Project 1 To 3 Edited in Church
1.1. Background
Diabetes mellitus is the commonest endocrine-metabolic disorder characterized by
chronic hyperglycaemia giving rise to the risk of microvascular (retinopathy,
nephropathy, and neuropathy) and macrovascular (ischaemic heart disease, stroke and
peripheral vascular disease) damage, with associated reduced life expectancy and
diminished quality of life. Diabetes mellitus may present with characteristic symptoms
such as thirst, polyuria, blurring of vision, and weight loss. In its most severe forms,
ketoacidosis or a nonketotic hyperosmolar state may develop and lead to stupor, coma
and, in absence of effective treatment, death (WHO, 1999). People with diabetes are at
increased risk of cardiovascular, peripheral vascular and cerebrovascular disease.
Several pathogenetic processes are involved in the development of diabetes. These
include processes, which destroy the beta cells of the pancreas with consequent insulin
deficiency, and others that result in resistance to insulin action. The abnormalities of
carbohydrate, fat and protein metabolism are due to deficient action of insulin on target
tissues resulting from insensitivity or lack of insulin (WHO, 1999). The prevalence of
diabetes is increasing rapidly worldwide and the World Health Organization (2003) has
predicted that by 2030 the number of adults with diabetes would have almost doubled
worldwide, from 177 million in 2000 to 370 million. Experts project that the incidence
of diabetes is set to soar by 64% by 2025 meaning that the disease will affect a
staggering 53.1 million citizens (Rowley and Bezold, 2012). The estimated worldwide
prevalence of diabetes among adults in 2010 was 285 million (6.4%) and this value is
predicted to rise to around 439 million (7.7%) by 2030 (Shaw et al., 2010). Recent
estimates indicate there were 171 million people in the world with diabetes in the year
2000 and this is projected to increase to 366 million by 2030. This increase in
The
prevalence in Nigeria varies from 0.65% in rural Mangu (North) to 11% in urban Lagos
(South) and data from the World Health Organization (WHO) suggests that Nigeria has
the greatest number of people living with diabetes in Africa (Wild S et al., 2004). It is
pertinent to note that in our setting, clinical criteria are often used to classify patients
with DM into Type 1 and Type 2 Diabetes Mellitus (T2DM). These criteria include a
cut off age of thirty years and insulin requirements or usage since diagnosis. For T2DM
additional clinical criteria for diagnosis, include history of usage of oral hypoglycaemic
agents or usage of combination of insulin and the oral hypoglycaemic agents (Ogbera et
al., 2014).
Type 2 Diabetes Mellitus (T2DM) risk factors allow for a prediction of an individuals
predisposition to developing T2DM disease. The presence of multiple risk factors
increases an individuals chance of being affected by T2DM in an exponential and not
additive manner (Blessey, 1985). Risk factors could be modifiable or non-modifiable.
Non-modifiable risk factors include age, sex, race and a positive family history.
Modifiable risk factors include smoking, excess alcohol use, unhealthy diet, obesity,
hyperlipidaemia, sedentary or physical inactivity (Ellis, 1948; Dawber and Kannel,
1958; Blessey, 1985; Rosengren et al, 2004; Yusuf et al, 2004; Stein et al, 2005; Anad et
al, 2008; Mayosi et al., 2009). Type 2 diabetes is due primarily to lifestyle factors and
genetics. A number of lifestyle factors are known to be important to the development of
type 2 diabetes, including obesity (defined by a body mass index of greater than thirty),
lack of physical activity, poor diet, stress, sedentary, and urbanization. Dietary factors
As questionnaire shows, the system was accepted in general, and this shows that the
role of information system in the health sector cannot be overlooked.
Towards reducing the burden of DM (majorly T2DM) in Nigeria, there is need for
concerted efforts by healthcare professionals and stakeholders in the health industry to
put in place preventative measures, a better functioning health insurance scheme and a
structured T2DM program. Therefore, earlier detection, public awareness and peoples
education seem to be the way out as human behaviour may slow down progress in the
eradication of diseases. People are eating less fruits and vegetables, more sugar, salt and
saturated fat. This together with decreases level of physical activity and other unhealthy
habits has resulted in more cases of T2DM and other diseases such as Diabetes mellitus
disorder (Humink et al., 1997). Due to scarce resources and inadequate health
provision, given the difficulty of long-term drug treatment in low-income countries,
primary prevention assumes a greater public health importance (Adedoyin et al., 2008).
This is motivated by the need for the provision of a decision support system, which
helps medical experts easily monitor Diabetes Mellitus (DM) diseases risk among
patients with a view of early detection of the likely occurrence.
1.2 Statement of the Problem
There is compelling data to show an increasing incidence and prevalence of DM in the
continent. The estimated prevalence of diabetes in Africa is 1% in rural areas, and
ranges from 5% to 7% in urban sub-Saharan Africa (Ogbera et al, 2014). Healthcare
providers require an accurate estimate of the Diabetes mellitus risk in patients to plan
the best possible allocation of finite resources to the core elements of DM control:
primary prevention, screening and early diagnosis, treatment, rehabilitation and
palliative care. Diabetes mellitus diseases affect individuals in their peak, early and
mid-life years disrupting the future of the families dependent on them and undermining
the development of the nations by depriving them of valuable human resources in their
most productive years (WHO, 2002). This is because Diabetes mellitus diseases could
eventually lead to disabilities such as stroke and thus, scarce family and societal
resources are directed to the costly and prolonged medical care of such ones (WHO,
2002). Therefore, the challenge of this project is to understand the risk factors or
variables that are responsible for the likelihood of Type 2 Diabetes Mellitus (T2DM)
disease occurrence and evaluate the likelihood of T2DM disease based on these
variables.
1.3 Scope of the Problem
This project is limited in scope by the development of a predictive model for Type 2
Diabetes Mellitus risk using Fuzzy Logic model.
1.4 Aim and Objectives of the study
The aim of this study is to develop a model for prediction of T2DM disease using the
Fuzzy Logic Model.
The specific objective of this study is to:
(i)
identify variables required for predicting T2DM disease risk.
(ii)
simulate the model.
(iii)
validate the model
1.5 Methodology
In order to achieve the aforementioned objectives, the methodology approach will be as
follows:
(i)
(ii)
(iii)
(iv)
variables and aggregation of the output variables, the software used for the
implementation of the system. Chapter four gives detailed information about the system
design, implementation and the tools used in the development of the system. It also
gives a description of the user interface, which the user uses in interacting with the
system. Finally, chapter five concludes the work by stating the summary, conclusion
and recommendation of the work done.
CHAPTER TWO
LITERATURE REVIEW
2.1 Introduction
According to Detmer (1997), epidemiology is the study of the distribution and
patterns of health-events, health-characteristics and their causes or influences in welldefined populations. It is the cornerstone method of public health research and
practice, and helps inform policy decisions and evidence-based medicine by
identifying risk factors for diseases and targets for preventive medicine and public
policies. Epidemiologists are involved in the design of studies, collection and
statistical analysis of data, and interpretation and dissemination of results (including
peer review and occasional systematic review). Over the past 30 years, epidemiology
has significantly contributed to improve methods used in clinical research and, to a
lesser extent, basic (microbiological, genetic) research (Jankowski, 1999). Major
areas of epidemiological study include bio monitoring, and comparisons of treatment
effects such as in clinical trials, outbreak investigation, diseases surveillance and
screening (medicine). Epidemiologists rely on a number of other scientific disciplines
such as Biology (to better understand diseases processes), Biostatistics (to make
efficient use of the data and draw appropriate conclusions), and Exposure assessment
and Social science disciplines (to better understand proximate and distal risk factors,
and their measurement (Bourlas et al, 1999).
The advancement in computer technology has encouraged the researchers to develop
software for assisting doctors in making decision without consulting the specialists
directly. The software development exploits the potential of human intelligence such
as reasoning, making decision, learning (by experiencing) and many others. Artificial
intelligence is not a new concept, yet it has been accepted as a new technology in
computer science. It has been applied in many areas such as education, business,
medical and manufacturing. This project explores the potential of artificial
intelligence techniques in determining the likelihood of Diabetes mellitus diseases in
an individual given a number of associated risk factors.
2.2 Diabetes Mellitus Diseases
Type 2 diabetes mellitus (T2DM) is the commonest form of diabetes affecting more
than 90% of the diabetic population worldwide. There is a rapid upsurge in the
number of diabetic patients and this explosive growth is noted in both urban and rural
areas. Wild et al. estimated the number of T2DM patients in the year 2000 at 174
million and predicted it to increase to 366 million in 2030. Diabetes mellitus (DM) is
a serious condition with potentially devastating complications that affects all age
groups worldwide. In 1985, an estimated 30 million people around the world were
diagnosed with diabetes; in 2000, that figure rose to over 150 million; and, in 2012,
the International Diabetes Federation (IDF) estimated that 371 million people had
diabetes. That number is projected to rise to 552 million (or 1 in 10 adults) by 2030,
which equates to three new cases per second (Sonny C. et al., 2011). This increase in
prevalence is expected to be more in the Middle Eastern crescent, Sub-Saharan Africa
and India. In Africa, the estimated prevalence of diabetes is 1% in rural areas, up to
7% in urban sub-Sahara Africa, and between 8-13% in more developed areas such as
South Africa and in population of Indian origin. The prevalence in Nigeria varies from
0.65% in rural Mangu (North) to 11% in urban Lagos (South) and data from the
World Health Organization (WHO) suggests that Nigeria has the greatest number of
people living with diabetes in Africa (Sonny C. et al., 2011).
10
2.2.1
11
12
The effects of risk factors is multiplicative rather than additive, thus people with a
combination of risk factors (for example, smoking, obesity and hypertension) have the
greatest risk of developing heart diseases. It is important to distinguish between
relative risk (the proportional increase in risk) and absolute risk (the actual chance of
an event). Thus, a 35 year old man with a plasma cholesterol of 10mmol/litre who
smokes 40 cigarettes a day is relatively much more likely to die from coronary
diseases within the next decade than a non-smoking woman of the same age with a
normal cholesterol, but the absolute likelihood of his dying during this time is still
small (high relative risk, low absolute risk) (Blessey, 1985).
Proximal risks for T2DM include those associated with consumption patterns (mainly
linked to diets, tobacco and alcohol use), activity patterns, and health service use as
well as biological risk factors such as increased cholesterol, blood pressure, blood
glucose, and clinical diseases. The Framingham Study first centred attention on the
concept of risk factors associated with T2DM, and most recently reported
substantial 30-years risk data showing the accumulation of risk over time (Pencina et
al, 2009). Importantly, risk factors for the incidence of T2DM and those associated
with T2DM severity or mortality are not synonymous. Risk factors for incidence
become important starting very early in life and accumulate with behavioural, social,
and economic factors over the life course to culminate in biological risks for T2DM
such as increased blood pressure, blood glucose, and clinical diseases. Over the past
few decades, the effectiveness of early screening and long-term treatment for
biological risks or early diseases has contributed to the sharp declines in DM
mortality seen in many countries (Humink et al, 1997). The American Diabetes
Association Guide to Diabetes Medical Nutrition Therapy and Education (Ross,
Boucher, and O'Connell, 2005) listed the major risk factors for type 2 diabetes
13
mellitus as: age 45 years, ethnicity, family history, habitual physical inactivity,
overweight (BMI 25 kg/m2), hypertension ( I40/90 mm Hg in adults), and
previously diagnosed impaired fasting glucose or impaired glucose tolerance, HDL
cholesterol < 35mg/dl) and/or triglyceride level (>250 mg/dl), polycystic ovary
syndrome, and history of vascular disease.
The recent WHO Global Health Risks Report of 2009 (Lopez et al, 2006) and the
earlier World Health Report of 2002 provide comparable and robust estimates of the
contribution of risks to total mortality and measures of disability (Mathers et al, 2003;
WHO, 2002, 2009b). Relatively few major behavioural and biological risk factors
account for T2DM incidence around the world. Tobacco use, diet (including alcohol,
total calorie intake, and specific nutrients) and physical inactivity serve as the three
major behavioural risks. Between them, they account for a significant proportion of
cancer, cardiovascular disease, and chronic respiratory diseases incidence in addition
to DM (Hu et al, 2001; WHO, 2002; Yach et al, 2004, 2005; Van Dam et al, 2008).
Concerted action focused on these behavioural risks, along with biological risks such
as high blood pressure, high blood lipids, and high blood glucose, would have a wide
impact on the global incidence and burden of diseases (WHO, 2009b). High blood
pressure, tobacco use, elevated blood glucose, physical inactivity, and overweight and
obesity are the five leading factors globally. In middle income countries, alcohol
replaces high blood glucose in the top five; in low income countries, a lack of safe
water, unsafe sex, and under nutrition are important. These latter points are related
to both the role of early childhood nutrition in the later onset of cardiovascular disease
and DM as well as the need to integrate the management of HIV/AIDS more closely
with DM in low-income countries (WHO, 2009b).
14
2.3.1
Overweight and obesity are defined as abnormal or excessive fat accumulation that
presents a risk to health. A crude population measure of obesity is the body mass
index (BMI), a persons weight (in kilograms) divided by the square of his or her
height (in metres). A person with a BMI of 30 or more is generally considered obese.
A person with a BMI equal to or more than 25 is considered overweight (WHO,
2015). Overweight and obesity are major risk factors for a number of chronic
diseases, including diabetes, cardiovascular diseases and cancer. Once considered a
problem only in high-income countries, overweight and obesity are now dramatically
on the rise in low- and middle-income countries, particularly in urban settings (WHO,
2015). According to Lebovitz (2004), overweight and obesity is a risk factor for
developing type 2 diabetes. The best measure of overweight and obesity is the body
mass index (BMI). Overweight status, a BMI of equal to or greater than 25 kg/m2,
and obesity, a BMI greater than or equal to 30 kg/m2, have become a problem
throughout the World. BMI levels of this proportion cause an increased risk of
developing many types of chronic diseases, including type 2 diabetes mellitus. In fact,
the term "diabesity" has been used to demonstrate the close link between type 2
diabetes mellitus and obesity.
2.3.2
Smoking
Nakanishi, Nakamura, Matsuo, Suzuki and Tatara (2000) have linked smoking to
diabetes. In this 5-year study of 1266 male Japanese office workers, 87 developed
impaired fasting glucose and 54 men developed type 2 diabetes mellitus. They found
that the number of cigarettes smoked per day as well as exposure to second hand
smoke was associated with development of the disease in these men. Therefore,
15
Wannamethee, Shaper, and Perry (2001) concluded that smoking is considered a risk
factor for developing type 2 diabetes mellitus.
2.3.3
Age
Physical Activity
WHO and FAO highlighted the importance of physical activity as a key determinant
of obesity, CVD, and diabetes (Joint WHO/FAO Expert Consultation, 2003).
Physical activity is defined as any bodily movement produced by skeletal muscles that
require energy expenditure. It has been identified as the fourth leading risk factor for
global mortality causing an estimated 3.2 million deaths globally. Physical activity is
a key determinant of energy expenditure, and thus is fundamental to energy
balance and weight control, Physical activity reduces risk for cardiovascular
diseases and diabetes and has substantial benefits for many conditions, not only
16
those associated with obesity. The beneficial effects of physical activity on the
metabolic syndrome are mediated by mechanisms beyond controlling excess body
weight. For example, physical activity reduces blood pressure, improves the level of
high-density lipoprotein cholesterol, improves control of blood glucose in overweight
people, even without significant weight loss, and reduces the risk for colon cancer and
breast cancer among women (WHO, 2004).
2.3.5
Urban-Rural Differences
17
2.3.6
Gender
In the first half of the last century, the prevalence of type 2 diabetes was higher among
women than among men, but this trend has shifted, so more men than women are now
diagnosed with type 2 diabetes. This change in the gender distribution of type 2
diabetes is mainly caused by a more sedentary lifestyle particularly among men,
resulting in increased obesity. However, recent data have also shown that men develop
type 2 diabetes at a lower degree of obesity than women a finding that adds support
to the view that the pathogenesis of type 2 diabetes differs between men and women.
Observations of sex differences in body fat distribution, insulin resistance, sex
hormones, and blood glucose levels further support this notion (Frch, K., 2014). The
body fat distribution, especially the abdominal visceral fat is associated with increased
type 2 diabetes risk. Body fat distribution differs by sex (Logue J. et al., 2011), and in
general men have more abdominal fat, whereas women have more peripheral fat
also denoted as apple versus pear shape. Looking into the abdominal fat, men
also tend to have more visceral and hepatic fat than women do, whereas women have
more subcutaneous fat than men do. In contrast to visceral fat, subcutaneous fat is
associated with improved insulin sensitivity and is therefore protective against type 2
diabetes. Thus, the phenomenon that men develop diabetes at a lower body mass
index than women can be explained by the fact that men have more visceral fat for a
given body mass index than women and thereby a higher relative risk for developing
type 2 diabetes (Logue J. et al., 2011).
2.3.7
There is also ample evidence that type 2 diabetes has a strong genetic basis. The
concordance of type 2 diabetes in monozygotic twins is ~70% compared with 20
30% in dizygotic twins (Valeriya L. et al., 2013). The lifetime risk of developing the
18
disease is ~40% in offspring of one parent with type 2 diabetes, greater if the mother
is affected and approaching 70% if both parents have type 2 diabetes. In prospective
studies, we have demonstrated that first-degree family history is associated with
twofold increased risk of future type 2 diabetes (Valeriya L. et al., 2013). The
challenge has been to find genetic markers that explain the excess risk associated with
family history of type 2 diabetes. A significant proportion of the offspring of
Cameroonians with type 2 diabetes have either type 2 diabetes (4 percent) or IGT (8
percent) (Mhanya et al. 2000). A positive family history seems to be an independent
risk factor for type 2 diabetes, but this was not the case in the Cape Town study
(Levitt et al, 1993), in which family history has not an independent risk factor.
2.3.8
Prediabetes
In 1997 and 2003, the Expert Committee on Diagnosis and Classification of Diabetes
Mellitus (Expert Committee on the Diagnosis and Classification of Diabetes Mellitus,
1997, Genuth S, et al., 2003) recognized an intermediate group of individuals whose
glucose levels do not meet criteria for diabetes, yet are higher than those considered
normal. These people were defined as having impaired fasting glucose (IFG) [fasting
plasma glucose (FPG) levels 100 mg/dl (5.6 mmol/l) to 125 mg/dl (6.9 mmol/l)], or
impaired glucose tolerance (IGT) [2-h values in the oral glucose tolerance test
(OGTT) of 140 mg/dl (7.8 mmol/l) to 199 mg/dl (11.0 mmol/l)]. Individuals with IFG
and/or IGT have been referred to as having prediabetes, indicating the relatively high
risk for the future development of type 2 diabetes. IFG and IGT should not be viewed
as clinical entities in their own right but rather risk factors for type 2 diabetes as well
as cardiovascular disease (ADA, 2014).
19
20
2.4.1
Testing enables health care providers to find and treat diabetes before complications
occur and to find and treat prediabetes, which can delay or prevent type 2 diabetes
from developing. Although not all tests are recommended for diagnosing all types of
diabetes, but the any one of the following tests can be used for diagnosis:
A1C Test, also called the haemoglobin A1c, HbA1c, or glycol haemoglobin
test
Fasting Plasma Glucose (FPG) Test
Oral Glucose Tolerance Test (OGTT)
Random Plasma Glucose (RPG) Test
2.4.1.1 A1C Test
The A1C test is used to detect type 2 diabetes and prediabetes but is not recommended
for diagnosis of type 1 diabetes or gestational diabetes. The A1C test is a blood test
that reflects the average of a persons blood glucose levels over the past 3 months and
does not show daily fluctuations. The A1C test is more convenient for patients than
the traditional glucose tests because it does not require fasting and can be performed
at any time of the day. The A1C test result is reported as a percentage. The higher the
percentage, the higher a persons blood glucose levels have been. A normal A1C
level is below 5.7%, and A1C of 5.7 to 6.4 %, indicates prediabetes.
People
diagnosed with prediabetes may be retested in 1 year. People with an A1C below 5.7
percent may still be at risk for diabetes, depending on the presence of other
characteristics that put them at risk, also known as risk factors. People with an A1C
above 6.0%, should be considered at very high risk of developing diabetes. A level of
6.5 percent or above means a person has diabetes (NDIC, 2014).
21
The Fasting Plasma Glucose (FPG) test is used to detect type 2 diabetes and
prediabetes. The FPG test has been the most common test used for diagnosing
diabetes because it is more convenient than the OGTT and less expensive (NDIC,
2014). The FPG test measures blood glucose in a person who has fasted for at least 8
hours and is most reliable when given in the morning. People with a fasting glucose
level of 100 to 125 mg/dl have impaired fasting glucose (IFG), or prediabetes. A level
of 126 mg/dl or above, confirmed by repeating the test on another day, means a
person has diabetes.
2.4.1.3 Oral Glucose Tolerance Test
According to National Diabetes Information Clearinghouse (NDIC, 2014), OGTT can
be used to diagnose type 2 diabetes, prediabetes, and gestational diabetes. Research
has shown that the OGTT is more sensitive than the FPG test, but it is less convenient
to administer. When used to test for type 2 diabetes or prediabetes, the OGTT
measures blood glucose after a person fasts for at least 8 hours and 2 hours after the
person drinks a liquid containing 75 grams of glucose dissolved in water. If the 2-hour
blood glucose level is between 140 and 199 mg/dl, the person has a type of
prediabetes called impaired glucose tolerance (IGT). If confirmed by a second test, a
2-hour glucose level of 200 mg/dl or above means a person has diabetes.
2.4.1.4 Random Plasma Glucose (RPG) Test
The random plasma glucose (RPG) test is sometimes used to diagnose type 2 diabetes
during a regular health check-up. If the RPG measures 200 micrograms per decilitre
or above and the individual shows symptoms of diabetes, then a health care provider
may diagnose diabetes (NDIC, 2014).
2.4.2
22
Frequent urination
Disproportionate thirst
Intense hunger
Weight gain
Increased fatigue
Irritability
Blurred vision
Itchy skin
23
24
al., 2006). The three-layered MLP with 40 categorical input variables and modified
learning method achieved a diagnosis accuracy of over 90%.
Support vector machines are a new and promising classification and regression
technique proposed by Vapnik and his co-workers (Cortes & Vapnik, 1995; Vapnik,
1995). SVMs, developed in statistical learning theory, are recently of increasing
interest to biomedical researchers. They are not only theoretically well-founded, but
are also superior in practical applications. For medical, clinical decision support and
biological domains, SVMs have been successfully applied to a wide variety of
application domains, including MDSS for the diagnosis of tuberculosis infection
(Veropoulos, et al, 1999), tumour classification (Schubert, et al, 2003), myocardial
infarction detection (Conforti & Guido, 2005), biomarker discovery (Prados et al,
2004), and cancer diagnosis (Majumder, et al, 2005).
To overcome the limited generalization performance of single models and simple
model combination approaches, more precise model combination methods, called
ensemble methods, have been suggested. This multiple classifier combination is a
technique that combines the decisions of different classifiers that are trained to solve
the same problem but make different errors. Ensembles can reduce the variance of
estimation errors and improve the overall classification accuracy. Many ensemblebased approaches have been proposed in recent research, including an ANN ensemble
for decision support system (Ohlsson, 2004), an ensemble of ANNs for breast cancer
and liver disorders prediction (Yang & Browne, 2004), MDSS with an ensemble of
several different classifiers for breast diagnosis (West, et al, 2005), and multiple
classifier combinations with an evolutionary approach (Kim, et al, 2006).
2.5.1
25
Diabetes is known as one of the most common diseases that has significant
burden on patients and healthcare systems. Nowadays, there are many researches in
the field of diabetes monitoring. These researches are coming as a sequence of
evolutions. Mashael S. B. (2013) emphasized that the first evolution in diabetes
monitoring was the use the computers to manage patient data and save their
records which include personal information, treatment progress and historical
information. Then, these monitoring systems were developed and become as dual
sides systems. In this type of systems, diabetes can be controlled remotely by
which called tele-monitoring and tele-medication. In such systems, the health care
providers offer tele-support and monitoring services to patients through a desktop
computer at home. Patient also can enter data about his daily intake food, activities,
and medication and get a right advice about his condition. Dorsey and Mayer (1994
and 1995) stated also that genetic algorithm is a useful search procedure that searches
from one population of points to another; thus directing the search to the best solution
so far rendering it as a global solution to non-linear functions.
2.5.2
Predictive models
Predictive modelling is concerned with analysing patterns and trends in historical and
operational data in order to transform data into actionable decisions. This is
accomplished by analysing and modelling the dynamics of the application-specific
data. In its raw form, this data is of limited value and is mainly used for reporting
what has happened. However, when the data is compiled into a compact model, it is a
powerful tool for proactively predicting what will happen.
In the abstract, a predictive model is a computational structure that can accurately
forecast an outcome of interest (i.e. output or dependent variable) when provided with
26
input data (i.e. independent variables) that have a measurable causal or coincident
relationship to the output. In order for predictive modelling to be useful in a given
application, two fundamental principles must hold:
i. Outcomes must have some level of predictability from known data. That is, similar
patterns represented across model inputs should be indicative of similar outputs;
ii.
There exist some measurable relationship between the set of known data values
that will be used as model inputs and the resulting output value(s) that the
model is tasked to approximate; and
iii.
Relationships that existed in the past will continue to hold in the future such that
it is reasonable to use past observations to infer future behaviour.
When these principles are adhered to, predictive modelling can approximate the
relationship between the known input data measures and the resulting output.
2.5.3
There are generally two classes of predictive modelling applications that differ by the
type of output the model produces:
i.
Forecasting: Forecasting model generate outputs that are continuous-valued. That is,
the output should be a value ranging from the minimum to the maximum
allowed. These models are used in applications such as forecasting/estimating:
sales, volumes, costs, yields, rates, temperatures, scores, etc. and
ii.
27
2.6
Related Works
A few number of prediction systems exists concerning Diabetes Mellitus and other
related diseases such as Cardiovascular diseases prediction with varying factors and
data mining methodology applied.
2.6.1
regions
and
therefore
patients
responding
to
these
Associative Classifiers
Jyoti et al (2011) designed the IHDPS system as a GUI based Interface to enter the
patient record and predict whether the patient is having Heart diseases or not using
Weighted Association rule based Classifier. The prediction is performed from mining
the patients historical data or data repository. In Weighted Associative Classifier
(WAC), different weights are assigned to different attributes according to their
predicting capability. The system has been implemented in java Platform and trained
using benchmark data from UCI machine learning repository. The system is
expandable for the new dataset.
28
Nave Bayes
The DSHDPS was developed by Subbalakshmi et al (2011) using Naive Bayesian
Classification technique. The system extracts hidden knowledge from a historical
heart diseases database. It is one of the most effective models to predict patients with
heart diseases. This model could answer complex queries, each with its own strength
with respect to ease of model interpretation, access to detailed information and
accuracy. DSHDPS can be further enhanced and expanded. For, example it can
incorporate other medical attributes besides the one used. It can also incorporate other
data mining techniques. Continuous data can be used instead of just categorical data.
29
Table 2.1
S/N
1.
Author(s)
Kevin
P,
Razvan
B,
Cindy M, Jay
S, and Frank S.
A. (2014)
Research Title
A
Machine
Learning
Approach
to
Predicting Blood
Glucose Levels
for
Diabetes
Management
Scope
Blood
Glucose
Levels
2.
Jyoti
Soni,
Uzma Ansari,
Dipesh
Sharma,
Sunita
Soni
(2011)
Heart
Disease
Predictio
n
3.
G.Subbalaksh
mi,
MTech,
K.
Ramesh,
MTech,
M.
Chinna Rao,
PhD (2011)
Intelligent and
Effective Heart
Disease
Prediction
System
using
Weighted
Associative
Classifiers
Decision
Support in Heart
Disease
Prediction
System
using
Naive Bayes
Heart
Disease
Predictio
n
Strengths
The system incorporate
Support Vector
Regression (SVR) model,
informed by a
physiological model and
trained on patient specific
data, has outperformed
diabetes experts at
predicting blood glucose
levels and can predict
23% of hypoglycaemic
events 30 minutes in
advance
The system incorporates
patient health record with
a
detailed
genetic
analysis. There is a need
to combine these factors
to provide a better overall
determinant of risk.
The
system
extracts
hidden knowledge from a
historical heart disease
database. It is one of the
most effective models to
predict patients with heart
Limitations
The SVR system was
able to predict 23% of
the
hypoglycaemic
events with a false
positive rate under 1%.
Remarks
The system performs
prediction using blood
glucose datasets collected
from Type 1
DM
patients
and
SVR
model with Physiological
features
The
prediction
is
performed from mining
the patients historical
data, which is from
UCI machine learning
dataset,
which
is
mainly
used
for
research purpose.
The system uses only
categorical
data
without incorporating
continuous data.
30
disease. It is implemented
as
web
based
questionnaire application
and can serve as a training
tool to train nurses and
medical
students
to
diagnose patients.
4.
S. Pruna, N. D. One-Side
Harris, and R. Desktop
Dixon,(2000).
Diabetes
Monitoring
System
Diabetes
The system improves the
Monitorin quality
of
diabetes
g System services by providing the
clinicians
with
a
computerized
diabetes
registry. The clinicians
have many options for the
management
of
the
creation, correction, and
visualization of patients'
records.
The
system
was
developed
using
a
modular design and
object oriented method
approach
and
its
architecture was based
on the Good European
Health Care Record
(GEHR)
31
CHAPTER THREE
RESEARCH METHODOLOGY
3.1 Introduction
The research methodology focused on the identification of the different variables required
for predicting the risk of T2DM in patients from Specialist in the College of Medicine,
Obafemi Awolowo University, Ile Ife via the use of structured interview followed by the
formulation of the fuzzy logic based model for predicting the risk of T2DM in such
patients through the use of MATLAB fuzzy logic toolbox.
3.2 Variables Description
In this study, the work is limited to six paramount risk factors of the T2DM only since the
work is intended to provide a system, which aids preventive medicine via the earlier
detection of the disease risk. The causatives variables of T2DM were classified according to
the groups that they belong to and may only be used to identify the status of the individual
risk to these groups (see Table 3.1).
The risk factors of those set of variables that help in the identification of the risk of T2DM
include:
i.
Body Mass Index (BMI): this is a measure of the ratio of the height (in meters) to
the square of the weight (in Kg) used in identifying the likelihood of obesity. The
risk of diabetes and cardiovascular disease increases and the body mass index
ii.
increases.
Age: this is another major determinant of the Type 2 Diabetes Mellitus disease
because the higher the age (from 30 years old) the higher the likelihood of the
iii.
T2DM disease.
Family History of Diabetes: This is another identification of the existence of
family members who have had T2DM or are still living with the disease. The risk
32
of T2DM increases with the existence of family members especially the first
iv.
generation members.
Blood Pressure: this is the measure of systolic and diastolic blood pressure of the
individual and has a benchmark. The risk of T2DM increases with the increase in
v.
blood pressure.
History of Gestational Diabetes: Gestational diabetes is the type of diabetes that
usually affects the women during pregnancy. The risk of T2DM increases with
vi.
33
Table 3.1
S/N
1.
Labels
None/ 3rd Generation, 2nd
2.
Age
3.
4.
History of Gestational
5.
Diabetes
Blood Pressure
6.
Gender
Male, Female
34
35
0,x a
xa
, a x b
ba
f ( x ; a , b , c) =
(3.1)
cx
,bx c
cb
0, c x
or
( (
f ( x ; a , b , c ) =max min
xa cx
,
, 0 (3.2)
ba cb
) )
Where a and c are the base of the triangle and b is the apex point of the triangle, and x is the
label value within the interval a and c
In this study, all variables were divided into three labels each represented by its own
triangular membership functions defining their respective base points and apex points. Thus,
in the process of fuzzification all variables were mapped using the following membership
functions as defined below, for the three (3) labels of each variable and that of two (2) labels
of each variable as follow in equations 3.3, 3.4, 3.5, 3.6, and 3.7 respectively:
For the variables with two labels, the membership functions will be:
label 1=f ( x ; 0.00, 0.25,0.5 ) .(3.3)
While for the variables with three labels, the membership functions will be:
label 1=f ( x ; 0.00, 0.16,0.33 ) ..(3.5)
36
3.4 Fuzzy Logic Model for Predicting Type 2 Diabetes Mellitus Disease Risk
The fuzzy logic model for predicting the risk of T2DM involves the process of fuzzification
defining the input and output variables in the Fuzzy Inference System (FIS), construction
of rule-based for the inference engine, the aggregation of the rules and then the
defuzzification of the results of the aggregated membership function.
The first process in modelling a fuzzy logic system is Fuzzification, and this is used to
convert each of input data to a degree of membership function in the MATLAB fuzzy logic
toolbox. Thus, the triangular membership function is chosen for fuzzification of both inputs
and output variables. In the process of fuzzification, each input data was mapped with the set
of rules to establish the degree of fitness on how each rule matches the particular input. It is
to be noted that the triangular membership function was used to map the degree of
membership of the labels of each variables used for input and output variables.
The schematic representation of the fuzzy logic system for T2DM disease risk predicting
system in figure 3.1 below shows the set of variables used as inputs ofr the model and the
risk as the output variable for the system.
37
Age
Rule 1
Rule 2
Rule 3
Rule N
Rule N
AGGREGATION
DEFUZZIFICATION
(All
Rules
are aggregated
output variable
T2DM DISEASE
RISK
INFERENCE
ENGINE into a single fuzzified
Type 2 Diabetes Mellitus Risk Predic
Using
ZIFICATION (Use Triangular Membership Function to map the variables to their respective label
(Low, Moderate, High)
= Value) AND (Body-Mass-Index = value) AND (Age = value) AND (History-of-Gestational-Diabetes = value) AND (Blood-Pressure = Value) AND (Gender = value) THEN (
History of Gestational Diabetes
Blood Pressure
Gender
Figure 3.1: Schematic Representation of the T2DM Disease Risk Inference System
38
39
Table 3.2
S/N
1.
Risk Factors
Labels
0.16
nd
4.
Age
7.
2 Generation,
0.49
1st Generation
0.83
30
30 45
> 45
Normal ( 24.9)
0.16
0.49
0.83
0.16
0.49
Obesity (30.0)
Positive
0.83
0.75
11.
Blood Pressure
Normal,
Pre hypertension
High
0.16
0.49
0.83
14.
History of Gestational
Negative
0.25
Male
Female
0.25
0.75
Diabetes
15.
Gender
40
Figure 3.2
41
Identify the factors affecting diabetes mellitus disease and their corresponding
influence. This is to highlight the factors that are considered to be associated with
diabetes mellitus disease and how significance their influence is so that an accurate
ii.
iii.
information.
Generate a model using fuzzy logic approaches. The model was developed by
iv.
identifying the variables that are required in type 2 diabetes mellitus disease.
Develop a prototypical type 2 diabetes mellitus disease risk predicting system with
using fuzzy set approaches. Efforts were made to ensure that the system is able to
predict the likelihood of occurrence of diabetes mellitus disease.
The system aim to assist doctor in predicting the patient type 2 diabetes mellitus disease risk
status thereby reduces the number of people coming to the hospital and easing the doctors
task. It will also allow people to know how prone they are to developing type 2 diabetes
mellitus disease without visiting the hospital based on their body mass index, blood
pressure, sedentary lifestyle, health history and their current health status, though some
information will still be needed from the doctor for accurate prediction.
42
Fuzzy Inference System Editor: The MATLAB fuzzy logic toolbox contains fuzzy
inference system (FIS) editor that was used to define both the input and output
variables. The FIS Editor handles the high-level issues for the system by determining
the number of input and output variables alongside their names. The Fuzzy Logic
Toolbox does not limit the number of inputs. However, the number of inputs may be
limited by the available memory of the machine. If the number of inputs is too large,
or the number of membership functions is too big, then it may also be difficult to
ii.
iii.
iv.
system.
The Rule Viewer and the Surface Viewer are used for looking at, as opposed to
editing, the FIS. They are strictly read-only tools. The Rule Viewer is a MATLAB
based display of the fuzzy inference diagram shown at the end of the last section.
Used as a diagnostic, it can show (for example) which rules are active, or how
v.
43
Figure 3.3
44
Hardware Requirements
For the proper functioning of the diabetes mellitus disease risk prediction system, the
following items will be needed for the hardware:
a. A Computer with internet access and at least a Pentium III processor;
b. An input and pointing device;
c. A hard Disk of at least 1GB of size is required in order for the repository to run well
without congesting other programs; and
d. Random Access Memory of at least 512MB is required.
3.8.2
Software Requirements
The following software will be needed for the proper functioning of the diabetes mellitus
disease risk prediction system:
a) Windows Operating System (Wins. 7 and above).
a) MATLAB Fuzzy Logic Toolbox
45
CHAPTER FOUR
TYPE 2 DIABETES MELLITUS DISEASE RISK MODEL DEVELOPMENT
4.1 Simulation of the Fuzzy Logic Model for Predicting Type 2 Diabetes Mellitus
Disease Risk
The simulation of the fuzzy logic model for the prediction of Type 2 Diabetes Mellitus
disease risk was simulated using the fuzzy logic toolbox available in the MATLAB 2013
Development Environment. Using the formulated triangular membership functions defined
for each input and output variable, the membership functions and the respective fuzzy
inference model for the risk of T2DM using six risk factors as shown in figure 3.2 above
were used as the inputs. The triangular membership functions in figures
46
Figure 4.0:
47
48
49
50
51
52
53
REFERENCES
ADA 2014: American Diabetes Association: Diagnosis and Classification of Diabetes
Mellitus Diabetes Care Volume 37, Supplement 1, January 2014. <See
http://creativecommons.org/licenses/bync-nd/3.0/> DOI: 10.2337/dc14-S081
Akinkugbe OO. The non-communicable diseases in Nigeria Report of a national survey.
Ibadan, Nigeria: Intec printers limited, 1997.
A. Kollmann et al. , Feasibility of a mobile phone-based data service for functional
insulin treatment of type 1 diabetes mellitus patients, Journal of Medical Internet
Research, vol. 9, 2007, p. e36
Balla J. I, Iansek R., & Elstein A. (1985). Diagnosis in presence of pre-existing disease
Lancet, 325(8424), 326329
Baxt W. G. (1990). Use of an artificial neural network for data analysis in clinical decision
making: The diagnosis of acute coronary occlusion. Neural Computing, 2(4), 480
489.
Baxt W. G. (1995). Application of artificial neural networks to clinical medicine. Lancet,
346(8983), 11351138.
Centers for Disease Control and Prevention. National diabetes fact sheet: national estimates
and general information on diabetes and prediabetes in the United States, 2011.
Atlanta: U.S. Department of Health and Human Services, Centers for Disease Control
and Prevention.
Chae Y. M, Park K. E, Park K. S, & Bae M. Y. (1998). Development of medical decision
support system for Leukemia management. Expert Systems with Applications, 15,
309315.
54
Chinenye S, Young E (2011) Diabetes Care In Nigeria. The Nigerian Health Journal, Vol.
1 1, No 4, October - December, 2011
Conforti, D., & Guido, R. (2005). Kernel-based support vector machine classifiers for early
detection of myocardial infarction. Optimization Methods and Software, 20(23),
401413.
Cortes, C., & Vapnik, V. (1995). Support-vector networks machine learning (pp. 237297).
Boston, MA: Kluwer Academic Publisher.
Ellenius, J., & Groth, T. (2000). Transferability of neural network-based decision support
algorithms for early assessment of chest-pain patients. International Journal of
Medical Informatics, 60(1), 120.
Dorsey R. E., & Mayer W. J. (1995). Genetic algorithms for estimation problems with
multiple optimal, non-differentiability, and other irregular features. Journal of
Business and Economic Statistics, 13(1), 53-66.
Dorsey R. E., & Mayer W. J. (1994). Advances in Artificial Intelligence in Economics,
Finance, and Management (J. D. Johnson and A. B. Whinston, eds.). Optimization
using genetic algorithms, Vol. 1. Greenwich, CT: JAI Press Inc., 69-91.
Dybowski, R., & Gant, V. (1995). Artificial neural networks in pathology and medical
laboratories. Lancet, 346(8984), 12031207.
Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Report of the
Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Diabetes
Care 1997; 20:11831197
Famuyiwa OO, Nwabuebo IE, Abioye AA. Pattern of histocompatibility (HLA) antigen
distribution among Nigerian (West African black) diabetics. Diabetes1982; 31: 11191122 [PMID: 6757024 DOI: 10.2337/diacare.31.12.1119.
55
Frch, Kristine. Gender and T2DM [internet]. 2014 Aug 13; Diapedia 3104972816 rev. no.
10. Available from:http://dx.doi.org/10.14496/dia.3104972816.10
Gonzlez EL, Johansson S, Wallander MA, Rodrguez LA (2009). Trends in the
prevalence and incidence of diabetes in the UK: 1996 2005. J. Epidemiol.
Community Health. 63: 332-336.
Geurts, P., Fillet, M., de Seny, D., Meuwis, M. A., Malaise, M., Merville, M. P., et al. (2005).
Proteomic mass spectra classification using decision tree based ensemble methods.
Bioinformatics, 21(14), 31383145.
Genuth S, Alberti KG, Bennett P, et al.; Expert Committee on the Diagnosis and
Classification of Diabetes Mellitus. Follow-up report on the diagnosis of diabetes
mellitus. Diabetes Care 2003; 26:31603167
Guven, A., & Kara, S. (2006). Classification of electro-oculogram signals using artificial
neural network. Expert Systems with Applications, 31(1), 199205.
Harris, N. L. (1990). Probabilistic belief networks for genetic counseling. Computer
Methods and Programs in Biomedicine, 32(1), 3744.
Harris M I:Non-insulin-dependent diabetes mellitus in black and white Americans. Diabetes
MetabRev 6:71-90,1990
Hu EA, Pan A, Malik V, Sun Q (2012-03-15). "White rice consumption and risk of type 2
diabetes: meta-analysis and systematic review". BMJ (Clinical research ed.) 344:
e1454.doi:10.1136/bmj.e1454.PMC 3307808.PMID 22422870.
Humink M.G, Goldman L, Tosteson A.N, Mittleman M.A, Goldman P.A, Williams L.W,
Tsevat J, Weinstein M.C (1997). The recent decline in mortality from coronary heart
disease. 1980-1990. The effect of secular trends in risk factors and treatment.
Journal of the American Medical Association, 277(7):535-542. [PubMed: 9032159].
56
H.-S. Kwon et al. , Establishment of Blood Glucose Monitoring System Using the
Internet, Diabetes Care, vol. 27, Feb. 2004, pp. 478 -483
Idowu P.A, Ogunlade O., Sarumi O. A, Balogun J.A (2015). Development of a Fuzzy Logic
Based Model for Cardiovascular Disease Risk, AICTTRA 2015 Proceedings. 45
-46
Joint WHO/FAO Expert Consultation on Diet Nutrition and the Prevention of Chronic
Diseases and World Health Organization Department of Nutrition for Health and
Development. WHO technical report series: Diet, nutrition and the prevention of
chronic diseases (2003). Report of a joint WHO/FAO expert consultation; Geneva.
28 January- 1 February 2002; Geneva World Health Organization.
Jyoti Soni, Uzma Ansari, Dipesh Sharma, Sunita Soni (2011). Intelligent and Effective
Heart Disease Prediction System using Weighted Associative Classifiers
International Journal on Computer Science and Engineering (IJCSE), 3(6).
Kevin P, Razvan B, Cindy M, Jay S, and Frank S. A. (2014) Machine Learning Approach to
Predicting Blood Glucose Levels for Diabetes Management. Modern Artificial
Intelligence for Health Analytics: Papers from the AAAI-14
Kim, M.-J., Min, S.-H., & Han, I. (2006). An evolutionary approach to the combination of
multiple classifiers to predict a stock price index. Expert Systems with Applications,
31(2), 241247.
Kinnear TW. The pattern of diabetes mellitus in a Nigerian teaching hospital. East Afr Med J
1963; 40: 288-294 [PMID: 14032897].
Li, Y. C., Liu, L., Chiu, W. T., & Jian, W. S. (2000). Neural network modeling for surgical
decisions on traumatic brain injury patients. International Journal of Medical
Informatics, 57(1), 19.
57
Liu, T.-F., Sung, W.-K., & Mittal, A. (2006). Model gene network by semi-fixed Bayesian
network. Expert Systems with Applications, 30(1), 4249.
Logue J, Walker J, Colhoun H, et al (2011) Do men develop type 2 diabetes at lower body
mass indices than women? Diabetologia 54: 3003-3006
Lopez A.D, Mathers C.D, Eszati M, Jamison D.T, Murray C.J.L. (2006). Global burden of
disease and risk factors. Washington, DC: World Bank.
MacDowell, M., Somoza, E., Rothe, K., Fry, R., Brady, K., & Bocklet, A. (2001).
Understanding birthing mode decision making using artificial neural networks.
Medical Decision Making, 21(6), 433443.
Majumder, S. K., Ghosh, N., & Gupta, P. K. (2005). Support vector machine for optical
diagnosis of cancer. Journal of Biomedical Optics, 10(2), 2434.
Malik VS, Popkin BM, Bray GA, Desprs JP, Hu FB (2010-03-23). "Sugar Sweetened
Beverages,
Obesity,
Type 2
Diabetes
and
Diabetes
mellitus
Disease
International
Journal
of
Advanced
Computer
Science
and
58
Diabetes
2014;
5(6):
905-911
URL:http://www.wjgnet.com/1948-9358/full/v5/i6/905.htm
Available
from:
DOI:
http://dx.doi.org/10.4239/wjd.v5.i6.905
Ohlsson, M. (2004). WeAidU A decision support system for myocardial perfusion images
using artificial neural networks. Artificial Intelligence in Medicine, 30(1), 4960.
Osuntokun BO, Akinkugbe FM, Francis TI, Reddy S, Osuntokun O, Taylor GO. Diabetes
mellitus in Nigerians: a study of 832 patients. West Afr Med J Niger Pract1971; 20:
295-312 [PMID: 5136797].
59
Oli JM. Remittant diabetes mellitus in Nigeria. Trop Geogr Med1978; 30: 57-62 [PMID:
675828]
Pencina M.J, DAgostino R.B, Larson M.G, Massaro J.M, Vasan R.S (2009). Predicting the
30-year risk of cardiovascular disease: The Framingham Heart Study. Circulation,
119(24):3078-3084. [PMC free article: PMC2748236] [PubMed: 19506114].
PHAC 2011: Public Health Agency of Canada. CANRISK: the Canadian diabetes risk
questionnaire user guide for pharmacists.Ottawa (ON): Public Health Agency of
Canada; 2011.
Prados, J., Kalousis, A., Sanchez, J. C., Allard, L., Carrette, O., & Hilario, M. (2004).
Mining mass spectra for diagnosis and biomarker discovery of cerebral accidents.
Proteomics, 4(8), 23202332.
Qu, Y., Adam, B.-L., Yasui, Y., Ward, M. D., Cazares, L. H., Schellhammer, P. F., et al.
(2002).
Boosted
decision
tree
analysis
of
surface-enhanced
laser
60
S. Pruna, N. D. Harris, and R. Dixon, Black Sea Tele Diab: building an information system
for management of diabetes, Proceedings 2000 IEEE EMBS International Conference
on Information Technology Applications in Biomedicine, Arlington, VA , USA: 2000,
pp. 284 - 289
Stern M P, Rosenthal M, Haffner S M, Hazuda H P, Franco L J: Sex difference in the effects
of socio cultural status on diabetes and cardiovascular risk factors in Mexican
Americans: the San Antonio Heart Study. AmJEpidemiol120:834-851,1984.
Stockwell, D. R. B. (1993). LBS: Bayesian learning system for rapid expert system
development. Expert Systems with Applications, 6(2), 137147.
Subbalakshmi G, Ramesh, K, Chinna Rao M. (2011). Decision Support in Heart Disease
Prediction System using Nave Bayes International Journal on Computer Science
and Engineering (IJCSE), 2(2).
Turkoglu, I., Arslan, A., & Ilkay, E. (2002). An expert system for diagnosis of the heart
valve diseases. Expert Systems with Applications, 23(3), 229236.
Van Dam R.M, Li T, Spiegelman D, Franco O.H, Hu F.B (2008). Combined impact of
lifestyle factors on mortality: Prospective cohort study in US women. British
Medical Journal, 337: a1440. [PMC free article: PMC2658866]. [PubMed:
18796495].
Vanhecke T.E, Miller W.M, Franklin B.A, Weber J.E, McCullough P.A (Oct 2006).
Awareness, knowledge, and perfection of heart disease among adolescents. Eur J
Cardiovasc
Prev
Rehabil.
13(5):718-723.
Doi:10.1097/01.hjr.0000214611.91490.5e.PMID 17001210.
Valeriya L. and markku L., (2013) doi: 10.2337/dcs13-2009 diabetes care august 2013 vol.
36 no. Supplement 2 S120-S126
61
Veropoulos, K., Cristianini, N., & Campbell, C. (1999). The application of support vector
machines to medical decision support: A Case Study. In Proceedings of the ECCAI
advanced course on artificial intelligence (ACAI 1999), 51 July 1999, Chania,
Greece.
West, D., & West, V. (2000). Model selection for a medical diagnostic decision support
system: A breast cancer detection case. Artificial Intelligence in Medicine, 20(3),
183204.
West, D., Mangiameli, P., Rampal, R., & West, V. (2005). Ensemble strategies for a medical
diagnostic decision support system: A breast cancer diagnosis application. European
Journal of Operational Research, 162(2), 532551.
WHO (1999): Malnutrition Related Diabetes Mellitus. Tech Rep Ser1985; 727: 20-24
WHO (2002): World health report 2002. Diagnosis and Classification of Diabetes Mellitus.
Geneva: World Health Organization; 2002.
WHO (2002): World health report 2002. Reducing risks, promoting healthy life. Geneva:
World Health Organization; 2002.
WHO (2004): Global strategy on diet, physical activity and health: NCD Global Strategy
Diet Physical Activity health 2004
WHO (2006): World Health Organization: Definition and Diagnosis of Diabetes Mellitus
and Intermediate Hyperglycemia: Report of a Who / IDF Consultation. Geneva,
World Health Org., 2006.
WHO Preventing Chronic Diseases: A vital investment 2005 (Accessed April 23, 2009)
36.
WHO World Health Statistics (2009). Geneva: World Health Organization; 2009e.
WHO (2009b). Global health risks: Mortality and burden of disease attributable to selected
major risks. Geneva: World Health Organization.
62