Search engines for information on medical diagnosis

The continuous training and clinical practice based on evidence are the main targets for the professional doctor development. During the 90s of last century the doctrine corpus of the Philosphy of the Evidence Based Medicine () reached maturity. The practice of EBM  requires locating facts published in the literature and make a critical evaluation of them. As the available information is far too extensive and of variable quality, convenient search strategies must be developed. These strategies must be designed by specialists so that high-quality papers on the treatment, prognosis, etiology and diagnosis are recovered from the bibliographic reference databases.
The goal of this paper is to define search and recover strategies about particular diagnosis and about some aspects of them just like, morphological features, illness evolution, epidemiology, or treatment. These strategies are embebded into search engines so that the user can em ploy them straightforwardly without worrying about defining proper search strategies.
We are currently involved in with the development of an expert system for cervix-vaginal cytology. This system will diagnose anomalies in epitelial cells. One of the main components for the system is a search engine for articles related to the considered diseases.
Roughly speaking, a search engine is a piece of software able to look for relevant information items from a document collection, given a query submitted by a user.  A typical search engine is usually composed of two main parts: an indexing module, which transforms the set of documents to the most appropriate way in order to have an efficient access; and  the retrieval engine itself, which uses as input a query submitted by a user and obtains as output the set of relevant documents ranked according to the degree of relevance. Therefore, this second module compares each document to the query, by means of a retrieval model, and assigns the “relevance status value”, measuring how near or far the document is from the query.
In our case, the document collection is composed of all the papers included in the MEDLINE collection. The basic aim of our research is to build a search engine that, given a query formulated starting from diagnostics and certain of its aspects,  is able to retrieve those papers that deal with the subject of the query, sorted by relevance. In fact, this would be a meta-search engine, because the actual search engine would be PUBMED. Our application would be a front-end to this website, being able to formulate a correct query, given the syntax of PUBMED, with the specific aspects of the diagnosis, and to retrieve the relevant papers. This implies that our search engine hides the user the details of the query language.