Information Retrieval in Healthcare

on

From past few decades, the amount of medical information is increasing at an explosive rate. As the research in Healthcare progressed, the medical literature has seen enormous growth. The amount of patient data has significantly increased over past 30 years not only in the form of text but also other media such as images, video, protein sequences, chemical structures etc. It is estimated that 30% of world’s data is dedicate to medical imaging.

Searching and retreival of information from data require complex queries, So searching with standard does not suffice the need of professionals.About 4.5% of queries in web searches are related to HealthCare. The bar graph shows percentage of different consumer searches in Healthcare related topics.

Health Related Searches


There are basically two categories of information Retreival :
(1) Knowledge-Based Information Retrieval
(2) Patient Specific Information Retrieval

Knowledge Based Information Retrieval

Information related to medical literature i.e information need from scientific papers , medical journals, reports etc. An important example of Knowledge based retrieval system is PubMed developed by National Library of Medicine, which contains 15 million bibliographic citations from MEDLINE.

The method of retreival involves two types : - Exact Match retreival and Partial Match retreival.

Exact match Retrieval

One of the example of exact match search system is PubMed. PubMed uses boolean retrieval method to extract medical literature related information using standard boolean variables such as OR, AND etc. For this search engine, user has to give some important keywords as shown below.



But, this is very painstaking and task for beginers as query formulation has to be done intelligently. Many derivatives of PubMed exists which search using specific property or search need.

Partial Match Retrieval


This is much simpler for user point of view as user have to enter the query directly and view the ranked results. It is implemented using TF-IDF. Many PubMed based search engines use this to facilitate ranked retrieval.

Patient Specific search


Information about individual patients and to search about health conditions from patient data. These of type of searches depend on patient’s background such as :

Information need : A Patient who is recently diagonised may need basic information about disease while a patient from longer disease may need advanced information regarding that disease.
Knowledge level and language : Knowledge level and language vary from person to person.Patient specific search engine are designed keeping this characteristics in mind.


Application in genomics


Most of the genomics data is available in form of text or text with HTML hyperlinks. Importance of IR in genomics search can be incorporated from the fact that user always prefers keyword -based search even when genomics data is not available in textual format.
One such tool is GeneCard which was built at Weizmann institute. It combines information of individual gene from various sources and provides partial match search to the users. It is specific to gene symbol as there is high similarity between symbols, only difference occurs at the numerical prefixes.


Other popular healthcare related search engines


Healia It was developed with consumers need in mind from National Cancer Institute . It provides high quality search results, its personalized search option allows to find relevant details and uses Health Thesaurus.

MEDgle MEDical GLobal Electronic generates medical conditions from symptoms , diagonises , physicians and medical techniques.
Medstory It is a microsoft health search engine for both professionals and consumers. It contains two tabs: one of which focusses on consumers’ needs (Health Tab) and other focus on health professionals. It uses other users’ queries to refine it’s search results. Search results can be further categorized such as Drug and Substances, Conditions and Procedures.


References





Comments

Post a Comment