Annotated Bibliography

Research Question: How can machine learning be implemented into a program to diagnose illnesses in areas that lack the resources to do so?

Many areas in the world lack the resources or money. Machine learning used in conjunction with medical diagnosis could help these areas have a jump start in medical knowledge and resource.

Kononenko, I. (2001). Machine learning for medical diagnosis: history, state of the art and perspective. Artificial Intelligence in medicine, 23(1), 89-109.

Does machine learning have the capability to do intelligent diagnosis on patients? In this article, Kononenko discusses the current position of machine learning in the field of artificial intelligence. Kononenko discusses the history of algorithms that have allowed for the analyzing and modeling of a large set of data. The same principle is used in machine learning while also being able to predict what can happen using the past values. One of the more important factors in implementation is the reliability of the algorithm. How can we be sure that we can trust the computer to have a correct medical diagnosis? Kononenko describes a method in which new cases are labeled and added with all the others into the learning set, which can test for the reliability of the diagnosis.

This is an article that reviews past and present (2001) knowledge on machine learning and its use in medical diagnosis. This article is definitely scholarly, and has been cited by other writers many times. Kononenko has a PhD in computer science, and is very objective in his writing. He is only objectively stating the current (2001) status of the field, and reporting how it can expand and be augmented.   

Although this article was written in 2001, I can use this to search for more current reviews on machine learning and medical diagnosis. Using papers that review Kononenko will allow me to have a more accurate view on the “current” situation of my topic.

Leung, M. K., Delong, A., Alipanahi, B., & Frey, B. J. (2015). Machine Learning in Genomic Medicine: A Review of Computational Problems and Data Sets. Proceedings Of The IEEE In Press, 4 December 2015, doi:10.1109/JPROC.2015.2494198

In this article by Leung, machine learning is discussed alongside the topic of genomic medicine. The large data sets created by the variations in DNA are a perfect task for a machine learning algorithm to sort through and analyze. However, Leung also states that this approach of analysis is not the most ideal method. The fact that the relationship between genotypes and phenotypes are not linear and simple makes it very difficult for accurate predictions. Leung also argues the difficulty in predicting diseases due to hidden variables that the program has not yet encountered. There would need to be some way to validate the models other than by human interaction. However, Leung still believes that the role of machine learning in genomic medicine will grow rapidly. With advances to deep learning (a classification of machine learning), more techniques are being formed to analyze more complex and large data sets.

This is an article on a conference proceeding done by the IEEE. The article is very recent, and the conference was held by a reputable organization: IEEE. In the article, Leung is very objective and gives both sides of the argument to certain methods of machine learning. The quality of his argument is also very good, as his authority on the subject of genomic medicine is at a level of expertise.

This writing went a little more in depth than I initially thought. The main topic of the writing was genomic medicine, which I was not initially looking for. However, reading this helped me realize that being more specific in my topic can help me more. It makes a lot of sense for genomes to be dealt with a computer as it creates a large data set. I may narrow my topic to diseases that can be detected at the genomic level.

Kampouraki, A., Vassis, D., Belsis, P., & Skourlas, C. (2013). e-Doctor: A Web based Support Vector Machine for Automatic Medical Diagnosis. Procedia - Social And Behavioral Sciences, 73(Proceedings of the 2nd International Conference on Integrated Information (IC-ININFO 2012), Budapest, Hungary, August 30 - September 3, 2012), 467-474. doi:10.1016/j.sbspro.2013.02.078

This article proposes a method of machine learning to diagnose health problems using a procedure based on Support Vector Machines (SVM). SVM are learning models, which are supervised, that have learning algorithms that recognize and analyze data sets. It uses this information to create a classification of the data. Kampouraki proposes using SVM to create an e-doctor; a web-based application that can make automatic medical diagnosis. The application can take information from two parties: a patient and an administrator (doctor). The patient will be able to input information about his or herself to receive a diagnosis, while an administrator can update the learning algorithm with new information on diseases and train the system. The diagnosis will diagnose for a specific problem that the user will select. Kampouraki also proposes future work where the SVM application will be a center platform where many other administrators can call from. This will help create an easier way to spread the program.

This article written on the scholarly journal Elsevier is very authoritative on the subject. Kampouraki proposes his way to develop a web-based application. Kampouraki discusses how it can be implemented, as well as ways that the program can be improved. Kampouraki also included screenshots of working prototypes of the application.

This article was exactly what I was looking for. It proposes a method that could be used with low cost to diagnose illnesses. The option to select certain illnesses to diagnose for is great in that it can help narrow my research topic, while also using the SVM application. The results shown by Kampouraki is positive, and shows one way in which machine learning can be used to diagnose with low resources.