Dengue Diagnosis Expert System

Author : Nuha Nor Haminudin
Date added : 2003-09-21

Brief Project Background

Hear this common myth : There’s nothing we can do to detect dengue at the very early stage. There are no possible ways to differentiate between dengue fever and normal fever. And it’s all depend on one’s body strength whether they’ll be able to stay alive or have to willingly accept the fact that they’re going to die because of dengue fever.

My story is just about using ICT knowledge to have the possibility of saving he life of dengue infected patients. The story will trace the journey that I’ve gone through to dispel the myth quoted earlier, up to the stage that now my effort has started to see some light, after winning the APICTA (Asia Pacific ICT Award) for Best Student Software Application category. But my journey has many more routes to follow before I can truly say that I’ve achieved my mission. One of the routes led me to this essay competition forum.

Dengue Diagnosis Expert System was actually my final year project, submitted in partial fulfilment of the requirement, for the Degree of Bachelor of Information Technology (Hons.).

My first idea was to build a system that will help the blind, using voiceXML technology. However the system has been widely built and already successful in the market. I have to be more creative & innovative if I were to challenge them.

I started doing my project during 1st semester of my final year 2003. I became interested in the application of logical thinking & concept of applying human’s intelligence in machines (computer), taught in Artificial Intelligence (AI) class. Expert System (ES) was one of the branches in AI, & diagnosis was one of the areas that have been widely applied in the development of ES.

After discussion with project Supervisor, Prof. Jamilin, I paid a visit to Dr. Nasir (child specialist) to enquire as to whether it’s appropriate for me to develop an ES on Asthma, Anaemia or HIV. He advised me to build an ES that can diagnose acute diseases, that requires fast detection & early treatment action to be taken instead.

2003, dengue was a serious problem in Malaysia. I read about lots of sad cases of dengue fatal patients in newspapers & internet. An idea came to my mind. I should try to build an ES that can diagnose dengue & specifies the appropriate action that should be taken at an early a stage as possible. Dr. Nasir stated that it’s a good idea since there’s never been a system like this being built yet.

The key to the success of this project, & the key to survival, is the early detection from observed symptoms. I realized the difficulty in gathering reliable data without help from any vector research institutes. I need to find reliable statistics of the dengue symptoms in a very limited time. I am so blessed with the emerging internet technology as I managed to find lots of reliable data from various resources. I did various readings on dengue, AI, ES & other related topics. I believe this system will be able to save human’s life. I am not doing this with just a mere assumption as every happening can certainly be mathematically analyzed.

Results

I came to know about the APICTA competition. My father was the one who introduced me to the MSC - APICTA 2003 competition. I sent my final year project to compete in Best Student Software Application Award category. On July 2003, I was selected as one of the finalist for the category. And fate determines all. I won the award, & knew that I’ve been given commendation & trust to produce a more reliable system.

My system was built to help diagnose one of the serious diseases that occur mostly in a developing country. As dengue is not very commonly found in already developed countries, it’s our own duty to try to tackle our own problems. I am going to elaborate more about the power & strength of my project. It’s now up to you to judge my findings and ideas.


Allow me to begin with first voicing out the problem statement; “The traditional approach in diagnosing dengue fever at the early stage, which is the manual analysis by general physicians, proves to be weak & inefficient. It has lots of disadvantages such as unreliability, slow, expensive & risky”. After thorough discussion with physicians, supervisor, patients, colleagues, parents & the others, I manage to come out with the ‘Hypothesis’ and ‘Goal’. The Hypothesis is that; “Once dengue is confirmed through blood test analysis, it can already be too late! And thus our Goal is; “To determine whether Patient is PROBABLY or NOT PROBABLY infected by dengue fever at as an early a stage as possible, based on a Set of Reliable Questions (On Dengue Possible Symptoms) that are referred to a set of Inferences.

We are going to do the diagnosis, “by checking the Signs and Symptoms of dengue infection against a databank of information gathered from Dengue Experts (the physicians)”. It’s important to take note that, once detected there’s possibility of dengue infection; specified action must be taken accordingly! The objectives of this system are 1st to detect dengue infection at an as early a stage as possible; 2nd, to specify the treatment action that should be taken by physicians; 3rd, is to capture the “Window of Mercy”.

‘Window of Mercy’ is a term that I have created. We are trying to capture this period of ‘Window of Mercy’ in order to save human’s life. The time frame for dengue virus to attack human’s body was divided into 3 different segments. The 1st segment is whereby the dengue virus start to attack human’s body. Signs & symptoms such as very high fever and lost of appetite, can already be observed within this segment. This is the early stage & this is the target that we are trying to achieve! The 2nd segment is the ‘Window of Mercy’, the time frame that we are trying to capture; as this is the time frame where human’s bodies are at their strength to fight dengue viruses. Since there’s no vaccine for dengue, the only treatment is by assisting the body. And thus, within the time frame of Window of Mercy, if given appropriate treatment, such as being given lots of fluids, I believe that the chances of survival are very high. We must capture this Window of Mercy! And we are not going to wait until it reaches the hemorrhages level, which is in the 3rd segment of dengue time frame. The traditional dengue diagnosis process mostly will wait until this critical level. And we should never wait until it reaches hemorrhages, which will only be proven through a number of blood tests analysis. Most dengue infected patients met their fatal end due to hemorrhages & shock syndromes which cause bleeding in the internal organs.

I’ve also developed & established the ‘Probability values’. This is the concept that utilizes knowledge of statistics & the ability of providing a solution based on mathematical inferences. I’ve developed a tree structure that cumulates all possible dengue symptoms, and percentages of being infected by dengue were calculated based on weightage concept. Using five parameters of dengue symptoms, thirty two possible outcomes were proven, as the result of using permutation formula (the nCr). For example, if parameter 1 (fever >= 41°C & occurs between 5 - 7 days), parameter 2 (headache/ myalgias/ arthralgias/ abdominal pain/ joint pain), parameter 3 and parameter 5 are satisfied, but parameter 4 is not satisfied, then the Possibility of patient being infected by dengue is 93 %. And according to the agreed and specified algorithm, if the percentage of having dengue is less than 49%, patients should be treated as outpatients, & if more than 49%, patients must be treated as inpatients.

Lessons

I will now answer this question; “What ICT training was necessary for implementation?”. My first answer is of course skills of finding related and reliable information on the Internet ! As stated earlier it was quite hard to gain data from any personnel or authority, Internet is the only reliable medium for any kinds of research areas. Then comes the need to master Prolog Programming Language, understanding the concept of Flex Engine, understanding WebFlex technology, and the capabilities to integrate between Medical Knowledge and computer technologies. I also need to come out with problem solving technique, using knowledge gained throughout the years, such as mathematical inferences technique (Statistics etc.) in deducing a reliable conclusion.

Skills of knowledge representation technique in Artificial Intelligence are also needed so that an idea or solution will be able to be visualized. An example used in this system was the Inference Tree knowledge representation technique. Another important skill that should not be left out here is the skills in finding reliable resources to learn and master the WinProlog programming software. I was having quite a difficult time regarding this matter as this software is not commonly used, and thus I need to find help from various sources. Other than understanding the software technique through observation, try and error from sample programs, I seek help from the original WinProlog developer in United Kingdom. I discuss with them through emails. Another method used in mastering this software was by discussing with a lecturer from another university, as this software was not taught in my own university. I need to find ways on how to make the users understand the real concept of my program. There is a need to explain in an understandable and user friendly ways, considering all levels and groups of users. I used the software engineering methodology so that the software developed will be in cost effective and timely manner.

What opportunity or problem occurred that called for the use of ICTs in my program ? The opportunities are, program developed will help physician diagnose dengue at early stage as possible. It will assist physicians in making reliable decision. Knowledge gained from experts on dengue is stored in databanks, and Inference Engine will infer the symptoms to come out with reliable answers. Questions asked are derived from the experts (having years of experience in diagnosing dengue). Then the answer will be represented to users through Graphical User Interface. Appropriate action will be taken, and it will assist doctors to make fast and quick decision. Another opportunity was the ability to incorporate medical knowledge into ICT capabilities.

In describing the problems that came out of the project, I rather use the term ‘Challenges faced’ instead. I see every problem as challenges that I must face, and I must do my best to solve them. The first challenge relates to programming issues. Expertise in Win-Prolog is very difficult to get as it is not a commonly used software. I had to seek help from the original developer of Win-Prolog in United Kingdom and Universiti Kebangsaan Malaysia. Combination Theory Programming was also not available to me. I have to do it manually using only 5 parameters, whereby in fact, if using more parameters, the accuracy of the system will be more enhanced. The second challenge relates to data acquisition problems. As stated before, it was very hard to gain cooperation and trust from related authorities. Data acquisition process was conducted through interviews with hospitals, Vector Research Institute, Physicians and doctors; observation of real life situations; reading materials; and also through internet research. The third challenge was the cross reference between ‘Domain Experts’ (people who possesses the skills and knowledge to solve a specific problem in a manner superior to others, such as doctors) and ‘Knowledge Engineer’ (person who designs, builds and tests an Expert System, in this case it was me).

‘What have I learned in the project so far that I think others - starting a similar project - would need to know ?’. My answer is, the platform has been developed, and the algorithm has also been established. They just need to be benchmarked by trusted bodies such as WHO. We just need to gain more knowledge on various dengue virus strains behaviour so that more accurate inference can be developed, and be tested by experts in each medical field.

‘How did ICTs help my project ?’. ICTs does help me a lot to make my project a success until this level, by providing the platform to search the knowledge on dengue, reliable and understandable medical diagnosis process, and also by enabling me to learn the necessary programming languages in a very flexible manner.

I will now say something about the technologies used. With the use of Internet, and its wide storage of knowledge, made it possible to find information on any topics people can think of. Artificial Intelligence provides useful integration between human natural knowledge processing capabilities with Information Technology. And it is also possible to combine Information Technology with Medical Research, and thus enhancing both fields to provide benefits for the development of humankind.

‘Did the project in any way contribute to poverty alleviation ?’. Yes, as the program developed will enable economical dengue detection diagnosis to be undertaken. Developing and least developing countries can use cheaply my diagnosis program developed. As it contains databank of dengue experts analysis and experience, which can be expensive and difficult to obtained in normal circumstances, this project will thus enable lives especially from these countries to have the possibility of being saved at nominal cost and thus saved valuable financial funds to be deployed elsewhere.

Development Impacts

As stated, I have used five parameters to come out with 32 sets of combination in determining the percentage of dengue infection. The parameters should be added, to increase the reliability of this Dengue Diagnosis Expert System.

The Knowledge Base has to be challenged and commented by “Domain Experts” such as dengue experts, general practitioner experts, programmer experts, and also by statisticians. Overall, the platform has been built, it just need to be benchmarked by trusted and reliable bodies such as WHO (World Health Organization) and IMG (International Medical Group).

This Expert System also can be modified to diagnose other acute diseases such as SARS (Severe Acute Respiratory Syndrome) by modifying the knowledge base of dengue fever.

In conclusion, I truly hope that my idea will be heard and recognized. I admit that my programming skills are not that much, but I just do my best. I hope this Dengue Diagnosis Expert System will be able to save lives and will be able to serve as a platform for other medical diagnosis process. It is not impossible to combine between different fields such as medicine and information technology, as their main target are the same!

Project Information

Organisation : Universiti Tenaga Nasional
Country of activity: Malaysia [MY]

Are there any partners involved : None

Contact Information

Nuha Nor Haminudin
serenity_nuha@yahoo.com
No.2, Jalan USJ 4 / 9E, UEP Subang Jaya,, 47600, Selangor Darul Ehsan, + 60 - 03 - 8024 3858

Disclaimer: No stories on this website shall be reproduced or stored in any other retrieval system without the written permission of the infoDev/IICD. Although every precaution will be taken in the preperation and maintenance of this collection of stories, neither infoDev, IICD or the submitting parties assume any responsibilities for errors or omissions. In addition, no liability is assumed fordamages resulting from the use of the information supplied in the stories.