A Machine Learning Approach for Identifying Disease-Treatment Relations in Short Texts

Abstract

The Machine Learning (ML) field has gained its momentum in almost any domain of research and just recently has become a reliable tool in the medical domain. The empirical domain of automatic learning is used in tasks such as medical decision support, medical imaging, protein-protein interaction, extraction of medical knowledge, and for overall patient management care. ML is envisioned as a tool by which computer-based systems can be integrated in the healthcare field in order to get a better, more efficient medical care. This paper describes a ML-based methodology for building an application that is capable of identifying and disseminating healthcare information. It extracts sentences from published medical papers that mention diseases and treatments, and identifies semantic relations that exist between diseases and treatments. Our evaluation results for these tasks show that the proposed methodology obtains reliable outcomes that could be integrated in an application to be used in the medical care domain. The potential value of this paper stands in the ML settings that we propose and in the fact that we outperform previous results on the same data set

GOAL OF PROJECT

The work that we present in this paper is focused on two tasks: automatically identifying sentences published in medical abstracts (Medline) as containing or not information about diseases and treatments, and automatically identifying semantic relations that exist between diseases and treatments, as expressed in these texts. The second task is focused on three semantic relations: Cure, Prevent, and Side Effect

ANALYSIS ON EXISTING SYSTEM

In order to embrace the views that the EHR system has, we need better, faster, and more reliable access to information. In the medical domain, the richest and most used source of information is Medline,4 a database of extensive

life science published articles. All research discoveries come and enter the repository at high rate (Hunter and Cohen [12]), making the process of identifying and disseminating reliable information a very difficult task The tasks that are addressed here are the foundation of an information technology framework that identifies and disseminates healthcare information. People want fast access to reliable information and in a manner that is suitable to their habits and workflow. Medical care related information (e.g., published articles, clinical trials, news, etc.) is a source of power for both healthcare providers and laypeople. Studies reveal that people are searching the web and read medical related information in order to be informed about their health. Ginsberg et al. [10] show how a new outbreak of the influenza virus can be detected from search engine query data

PROBLEM DEFINITION

The problems addressed in this paper form the building blocks of a framework that can be used by healthcare providers (e.g., private clinics, hospitals, medical doctors, etc.), companies that build systematic reviews8 (hereafter,

SR), or laypeople who want to be in charge of their health by reading the latest life science published articles related to their interests. The final product can be envisioned as a browser plug-in or a desktop application that will automatically

find and extract the latest medical discoveries related to disease-treatment relations and present them to the user. The product can be developed and sold by companies that do research in Healthcare Informatics, Natural Language

Processing, and Machine Learning, and companies that develop tools like Microsoft Health Vault. The value of the product from an e-commerce point of view stands in the fact that it can be used in marketing strategies to show that

the information that is presented is trustful (Medline articles ) and that the results are the latest discoveries. For any type of business, the trust and interest of customers are the key success factors

Disadvantage

IDEA ON PROPOSED SYSTEM

Our objective for this work is to show what Natural Language Processing (NLP) and Machine Learning (ML)

techniques—what representation of information and what classification algorithms—are suitable to use for identifying and classifying relevant medical information in short texts. We acknowledge the fact that tools capable of identifying reliable information in the medical domain stand as building blocks for a healthcare system that is up-to-date with the latest discoveries. In this research, we focus on diseases and treatment information, and the relation that

exists between these two entities. Our interests are inline with the tendency of having a personalized medicine, one in which each patient has its medical care tailored to its needs. It is not enough to read and know only about one study that

states that a treatment is beneficial for a certain disease. Healthcare providers need to be up-to-date with all new discoveries about a certain treatment, in order to identify if it might have side effects for certain types of patients. We envision the potential and value of the findings of our work as guidelines for the performance of a framework that is capable to find relevant information about diseases and treatments in a medical domain repository. The results that we obtained show that it is a realistic scenario to use NLP and ML techniques to build a tool, similar to an RSS feed, capable to identify and disseminate textual information related to diseases and treatments. Therefore, this study is aimed at designing and examining various representation techniques in combination with various learning methods to identify and extract biomedical relations from literature.