This report is the final year report of Mr. Shree Shrestha. He is a graduate of university of Wolverhampton. We will provide you the link to his project very soon. You can download his report by clicking download here button at the end of this page.
Introduction
Historical Prospective of AI in health and medical:
Application of computing and Artificial Intelligence in the field of health and medicine is one of the greatest achievements of human civilization. History of use of computers in the field of health and medicine has been started from 1950s (Ramesh, et al., 2004). But only from 1970s with the help of expert systems such as INTERNIST-I, MYCIN, ONCOSIN, CASNET, we can see the first application of Artificial Intelligence in health and medicine (Ishak & Siraj, 2008).
The application of artificial intelligence in health and medicine was limited to US till 1980. On 13-14 September 1985, with the motive to establish an active research community, an international conference was held in Italy. During those time, major problem to sustain artificial intelligence in the sector of health and medicine was lack of availability of data. To eliminate this problem the concept of Electronic Health Records, also known as EHR, was introduced in 1960s by Lockheed. Later, US government started using EMR after 10 years, in 70s with the Department of Veteran Affairs (Atherton, 2011).
Classification approach in Machine Learning:
Classification and Regression is the two major approaches of machine learning in supervised learning. In classification, model learns with the data which is fit into it and base on its learning from the training data it becomes able to classify new data. There are two types of classification approaches. First one is binary classification and second is multi-class classification. In binary classification, classification is made with just two output classes. For example, whether the person is suffered from diabetes or not, either a person is male or female or either the mail is spam or not. Where as in multi-class classification, the model classifies the output with more than two classes (minimum three). For example, classifying cat, dog and mouse from a group of animals, multiple disease classification, documents classification, etc. (Soofi & Awan, 2017). Classification approach used in this project is binary classification. Model will be trained with multiple algorithms like Logistic regression, Random forest, Neural Network, etc. Out of all classifiers, one with best performance will be used in the system. Model will classify whether the patients have diabetes or not and will give the output on the basis of 1 and 0 based on the medical report given to the model as input.
Figure 1: Supervised Learning classification techniques
Classification system and disease prediction.
In every country, many hospitals and clinics has been established either by the government or by private companies with the mind set of providing health related facilities such as counseling related to health and medicine, normal check-up to big surgeries. The major challenge for all those health service providers is providing high quality of service with minimum cost with zero error. To obtain desired results and server whole nation, implementation of artificial intelligence-based computer systems can be done, which can help in decision making process that also in minimum time period. Development of system which can detect or classify any particular disease inside a human body can be one of the greatest achievements of human civilization. Any system that makes a prediction or detection of a disease using classification algorithms of machine learning can be called as disease detection or classification system. To make a prediction or detection of various diseases, many researches has been carried out with the help of artificial intelligence and machine learning. This project, “Diabetes Detection, Know Your Diabetes”, is also one of the examples of classification system which classifies disease (C.Dharuman, et al., 2017).
Research methods used:
- Journals and articles
- Reports
- Conferences papers
- Newspapers and online news
- Documents from various web
- E-Books
Overview of Diabetes:
Diabetes is one of the dangerous and commonly found disease in many people round the world. Main cause of diabetes is high blood sugar level. This disease is one of the major causes for other severe illness and disease inside a person’s body and can even take a person towards death. If diabetes is not cared and control, it can also cause damage to other parts of the body like kidney, heart, eye, nerves, etc.
Diabetes is one of the kinds of metabolic disease categorized by hyperglycemia. Common symptom of diabetes can be seen from following cases like weight loss, urine problem, abnormally increase in thrust, excessive appetite and problem in vision. Longstanding problems of diabetes includes Retinopathy, Nephropathy, Peripheral neuropathy, Autonomic neuropathy.
- Retinopathy:
Retinopathy is the disease related to retina which cases serious problem to the vision of eyes.
- Nephropathy:
Nephropathy is the disease related to kidney. Those people who has diabetes since longer period has the higher chance to suffer from this disease.
- Peripheral neuropathy:
Peripheral neuropathy is a disease which affects in the peripheral nerves of human body. This disease can affect only single nerve to different nerves of the body at the same time.
- Autonomic neuropathy:
When there is damage to the nerves which various day to day information to the spinal cord and brain, we can say, person autonomic neuropathy (American Diabetes Association, 2014).
There are various types of diabetes. Some of the most common types of diabetes are as follows:
- Type 1 diabetes:
In type 1 diabetes, body does not create any insulin. In the case of type 1 diabetes, human body’s immune system attacks the pancreas gland which is responsible for making insulin. Because of the absence of the insulin in the body, the level of blood rises up than normal and causes diabetes. If the person suffers from type 1 diabetes, he/she must take insulin in daily basis in order to maintain the sugar level and stay healthy. Most of the victims of this diabetes are children and younger people, and less chances for the old age people.
- Type 2 diabetes:
Type 2 diabetes is found in most of the diabetes patient. This disease is caused when the blood sugar level is very high than normal. When the human body does not create enough insulin or lose the ability to utilize the created insulin then the person suffers from type 2 diabetes. Unlike type 1 diabetes, type 2 diabetes can be cared and stop the development of the disease, if proper care is given in the early stages. Mostly old age people aged from 45 or above suffers from type 2 diabetes. Major causes of type 2 diabetes are overweight, heredity and unhealthy life style.
- Gestational Diabetes:
This kind of diabetes is only found in female. If the person is pregnant, she is most likely to have gestational diabetes. Symptoms of gestational diabetes is only seen after 24th week of the pregnancy. This diabetes id very dangerous even to infant that is why a mother should carefully look after her health and stay updated of you blood sugar level (Rodgers, 2017).
Project Academic Questions:
- Can diabetes be accurately predicted using Machine Learning and Deep Learning model?
- How the doctors will get benefit by this model?
Project Background:
Many people have died till date because of misclassification of the disease. Considering this situation, if a system became able to help the doctors to detect or predict a disease in accurate way, on the basis of the given symptoms, then this can be the ultimate solution of this problem. With the innovation of high-performance computing, availability of data and advancement in the algorithms, artificial intelligence are in good practice now a days in the field of disease diagnosis. “Diabetes Detection: Know about your diabetes”, will be the web-based application, which will have diabetes detection model, that helps people to check and know if they are suffering from diabetes or not. Along with the detection model, it will have basic but authentic information regarding diabetes, its symptom as well as things to eat and guidelines to follow for healthy lifestyle
Details of artifact produced:
On daily basis, we hear the news of various cases of fraud happening in many hospitals, death of the patients because of the misguidance or improper treatment from the doctors and charge of big amount of fee even for a normal follow-up. To solve these problems in case of diabetes, “Diabetes Detection: Know about your diabetes”, has been built and will help people who wants to check their diabetes’s status staying in home and know the result in advance before going to the doctors for further checkup and treatment. To build the final model for the web-application, different machine learning algorithms such as random forests, decision tree and neural network were built and tested. As neural network gave the best accuracy among all algorithms, it has been finalized to fit into the web application. With the help of Artificial Neural Network (ANN), “Diabetes Detection: Know about your diabetes”, will detect whether the patient have a diabetes or not based on their medical report. To check the diabetes, user must have the report of their insulin level, glucose level, blood pressure, skin thickness, body mass index (BMI) and age. For female, total number of times she has become pregnant. After providing the data, ANN will detect and show the result on the basic of 0 and 1 which is “Yes” and “No”. An Artificial Neural Network (ANN) is a data processing paradigm which is inspired with the biological nervous systems i.e. human brain working mechanism. ANN is the combination of large numbers of highly connected neurons, working in unison to solve a specific problem. In general, Artificial Neural Network are in massive practice round the globe in many sectors because of their ability to build nonlinear relationship between input variable and output variables directly from training data.
Problem Statement:
Because of unhealthy life style the number of diabetes patients are increasing rapidly in every part of the world. Even though people brag of being modernize and literate, most of people are lacking behind in term of proper guidance to prevent from diabetes. Frauding from every sector like hospitals are also increasing and takes massive fees. Because of carelessness of medical personnel human casualties are also increasing.
Aims and Objectives:
Goal of this project is to detect the diabetes, provide useful data about diabetes and study about different machine learning algorithms such as Linear Regression, Random Forest and Artificial Neural Network to find out which one is the best algorithm that can be used accurately to detect diabetes.
Aims:
- To build a web-application and prediction system which will detect whether a person is suffering from diabetes or not.
- To provide information about causes and consequences of diabetes.
- To explain the precautions to prevent from diabetes and advices and suggestions to control diabetes / sugar level if in case a person is already suffering from diabetes.
People
- Aware general to stay fit and healthy.
Objectives:
- To prevent people from getting scammed by the doctors or medical staffs
- To help people to check about their diabetes/ sugar level easily with the help of simple web application.
- To provide the tips regarding physical exercises or activities to control blood sugar level.
- To provide information about hygienic food.
Scope and limitations:
Scope
Diabetes Detection System can be deployed for various purposes. Such as:
- It will be very useful to general public. Everyone can test their diabetes status at very 1st level staying at home.
- It will help medical staffs, doctors and contributes a lot in medical field.
- Doctors will also get help and one level of alert if patients have first tested with diabetes detection system.
- Time and money will be saved.
Limitations
However, systems like these also have their own major drawbacks. Such as:
- Not only this all artificial intelligence integrated system will just predict the result but will not point how and which major factor is responsible for the particular.
- It cannot be used as a final and only measure to check diabetes.
Structure of report
- Introduction
Introduction part explains all the details from the surface level such as how the system will be built from front-end as well as backend, which algorithm will be used. This section will further explain about the problem domain and will clearly explain, how after the development of the system, the problem will get minimized, and finally, how this system will work as the best alternative as well as helping hand for medical department.
- Literature Review
Literature Review is the core part of the report. This section focuses on history of AI in the field of medical field. This section will major focus on providing theoretical basis of the project by explaining about different algorithms which can be used in this system and analyzing similar systems.
- Implementation of Prediction System
Execution of Detection System section will focus on the AI part of the project. This chapter will analyze the dataset that will be used to train the model and how the system will make prediction on the basic of the dataset. This section will focus on every single detail that has been implemented to build a system. It will also explain all the developmental stages of detection system in increments.
- Implementation of Web-Application
Execution of Web-Application will shop the step by step execution of webapplication. Explanation about the framework used and all the pages which has been integrated in the web-application.
- Testing
Testing of the whole system will be done. White-box testing will be carries out. Lower level testing like unit testing and integration testing will also be done in other to check the internal implementation of the system from unit by unit.
- Conclusion
Concluding all the project and listing the further improvements that can be done to increase the benefit from the project.
- Critical Evaluation
This section will be the critical evaluation of the final project, evaluation of the final system, findings and future escalation.