Date of Graduation

2007

Document Type

Thesis

Degree Type

MS

Committee Chair

V. Jagannathan

Abstract

Paper based health records has proved to be not only costly but also has many disadvantages. Such records are difficult to organize and are hard to manage. This has led to the adoption of computerized electronic health records (EHR). To facilitate interoperability and consistency in exchange of information between various computerized clinical applications standard medical terminologies plays a vital role. These standard terminologies are based on controlled medical vocabularies like SNOMED, ICD9, and LOINC etc. Coding clinical documents with controlled medical vocabularies benefits the health care system in a wide range of applications like laboratory reporting, clinical research, computerized patient entry etc. This document deals with using the SNOMED medical vocabulary subset from UMLS for coding clinical documents. Coding of clinical documents would be more effective if we can avail the available natural language processing techniques. This work utilizes the UIMA platform which has a collection of natural language processing (NLP) tools. It also provides a framework on how these open NLP tools can be plugged into the UIMA architecture and how they can be used in developing any application. This report discusses the use of a statistical n-gram natural language processing technique that proved effective in assigning SNOMED codes to composite medical terms. With the component based architecture of UIMA which allows sharing and using of components between projects the thesis discusses how two components developed in separate projects are combined to form a SNOMED annotator. Thus this document describes the process of developing a SNOMED annotator for coding clinical documents which is scalable, flexible, UIMA framework compatible and extendible.

Download

Share

COinS