|
|
||||||||
Computers in Radiology |
1 Department of Radiology, Mayo Clinic, 200 First St. S.W., Rochester, MN 55905.
Received June 25, 2001; accepted after revision July 30, 2001.
Address correspondence to J.W. Gurney.
OBJECTIVE. The purpose of this study was to index the contents of the pre-MEDLINE American Journal of Roentgenology into a database that could then be accessed over the World Wide Web.
CONCLUSION. The database is composed of more than 8,000 citations from the years 1914 (volume 1) to 1965 (volume 95). Using a Web browser, a user can search the database by year, volume, title word, page, or author. Results are returned sorted by year and page number in a bibliographic format. The database is accessed approximately 200 times per month.
The journal article, peer-reviewed by colleagues, has been the forum for the communication of ideas and advancements in medicine for more than 200 years. Even before Roentgen's discovery, physicians thought they were overwhelmed with information. Finding information on a specific topic was especially difficult [1].
A physician seeking information had to make a pilgrimage to the library and manually review all the journals, scanning the table of contents of each for the topic of interest. It was just such an experience that stimulated John Shaw Billings, Surgeon General in charge of the Army Medical Library, to embark on a long-range project of hand indexing the world's medical literature. After 5 years of effort, the first Index Medicus was published in 1879 [1]. This subject catalog greatly eased the researcher's burden. Rather than adding new information, the index assimilated existing citations so that they could be easily found. John Shaw Billings' idea would eventually develop into MEDLINE, and the Army Medical Library, composed of 485 books, would eventually be transformed into the National Library of Medicine.
In the early 1960s, the National Library of Medicine obtained funds to begin a process of mechanization using the then new computer technology to help collate and produce Index Medicus. Developers quickly realized that these computer tapes could be searched. Computerized searching began in 1964 with MEDLARS (Medical Literature Analysis and Retrieval System) [1]. In 1965, trained librarians performed 1,623 searches. In 1971, MEDLINE (MEDLARS online) was launched. Today, anyone with Internet access can search the MEDLINE database free of charge. Hundreds of millions of searches are performed yearly [2]. The database contains more than 11 million citations from more than 4,000 journals from 1965 to the present. Unfortunately, the literature before 1965 is not indexed online and can be accessed only by browsing through old volumes of Index Medicus or by reviewing the tables of contents of selected journals. The purpose of this study was to produce a pre MEDLINE database of articles published in the American Journal of Roentgenology (AJR) for volumes 1-95 (1914-1965) and publish this database in an easy-to-use search format on the World Wide Web.
Materials and Methods
My two college-age nieces were employed to transfer the citations into a computer database. Both used portable computers to index each volume of the AJR. The citations were entered into EndNote software (ISI Research Soft, Berkeley, CA), a cross-platform bibliographic database program. Articles in each volume of the AJR were entered into a separate EndNote document. The volume, year, and journal name were added through a find-and-replace command for each document. Authors' names were double-checked for accuracy. The article titles were examined with a spelling checker (Spellswell Plus; Working Software, Santa Cruz, CA). Once the index was complete, the records were again proofread. Each EndNote document was exported as a tab-de-limited text file and then imported into a database (FileMaker Pro 5 Unlimited; FileMaker, Santa Clara, CA).
The FileMaker database contained the following fields: a primary key composed of an autoentered unique serial number and text fields corresponding to fields in the EndNote databases for authors, title, journal, volume, and inclusive pages (Fig. 1). A second page field containing only the first page number was calculated by parsing the inclusive page field. This field was used to sort the results of a search by page number.
|
Web Companion, the FileMaker common gateway interface, was used to connect the database to the Internet. Hypertext Markup Language (HTML) templates were designed using GoLive 5.0 software (Adobe, San Jose, CA). HTML templates are used by the common gateway interface to input searches into the database and to format the output from the results of a database search back to the client who requested the search. The search template contains a form and text input for searching by title, word, author, year, or page number (Fig. 2). Three templates were constructed for output, depending on whether the search is successful in retrieving records.
|
For successful searches, the citation fields are organized in a standard bibliographic format (Fig. 3). A maximum of 20 records are displayed with each search, with the option to proceed through lists of 20 citations each for the entire found set. The database sorts citations by year and the first page number of the article.
|
The database and the associated templates were placed in a file folder on a dedicated server (Macintosh G3; Apple, Cupertino, CA) with continuous Internet connection. The uniform resource locator for the database is http://www.chestx-ray.com/Research/Research.html.
To obtain perspective about how often pre-MEDLINE citations are currently used, the online AJR (www.ajronline.org) was searched for the numbers 1950 through 1959 in those issues for which the full text is presented online (January 2000May 2001) as of the first writing of this article. The bibliographies of each article were then manually searched for the occurrence of such a number. (The search was not specific for the year because the number could be part of an address or page number in an article rather than the year of publication.)
Results
The articles were indexed over a 6-month period. The design of the FileMaker database and HTML templates was constructed during the same time. The database contains 8,627 AJR records, each record corresponding to a single citation, and occupies 5.4 MB of disk space. An example of a client's search screen and results is shown in Figure 4A,4B. Depending on the client's Internet connection, the time between executing a search and receiving results is only a few seconds. The database was first made available in December 2000. Currently, approximately 200 page views per month are requested from the database.
|
|
Of the 1,019 articles published in the AJR from January 2000 through May 2001, 27 citations were found from the published literature during the 1950s.
Discussion
More than 90% of physicians have access to the Internet, and their most common Web activity is searching the literature for references, especially from the National Library of Medicine and other database vendors that use the MEDLINE database [3]. To my knowledge, this AJR database is the first database of pre-MEDLINE medical literature to be published on the Internet.
For MEDLINE, trained indexers use the 19,000 word MeSH (medical subject headings) to categorize each article. Each citation receives an average of eight MeSH terms [1]. This system adds value to the bibliographic database by categorizing each article with a standard vocabulary. Text title word searches are limited and may not retrieve all relevant citations. For example, a search for "lung cancer" will retrieve articles that contain that phrase but not citations that use the terms "bronchogenic cancer" or "nonsmall cell carcinoma." However, the appropriate MeSH term ("lung neoplasms") would retrieve articles with all those terms in the title. No attempt was made to categorize the AJR database with the MeSH terminology. Fields for MeSH terminology could be easily added to the database; however, to do so would require trained indexers to read each article and apply the appropriate terms.
AJR has undergone several name changes [4]. In 1913, the Quarterly Transactions of the American Roentgen Ray Society was expanded into a monthly publication and renamed the American Journal of Roentgenology. In 1923, it was renamed the American Journal of Roentgenology and Radium Therapy, and in 1952, the American Journal of Roentgenology, Radium Therapy, and Nuclear Medicine. In 1976, the American Journal of Roentgenology assumed its current title [4]. The style for abbreviating the title is AJR, but the official MEDLINE abbreviation is AJR Am J Roentgenol. In the AJR database, all volumes use AJR as the journal abbreviation. Changing the name in the database, if desired, would be a trivial matter.
The AJR database, in conjunction with MEDLINE, is a complete bibliographic history of the publication of a single periodical. Because the field of radiology is relatively young, the combined AJR and MEDLINE databases serve as a chronologic record of the development of a specialty. A researcher can now easily retrieve references almost 90 years old from any Internet-connected computer. Researchers can find original observations, technical developments, and other information that would otherwise not be readily available. To enhance scholarship, the database could be easily expanded to other pre-MEDLINE radiology journals.
Is the older literature worthwhile? The most often cited paper in the AJR during the period 1955-1986, "Observations on Growth Rates of Human Tumors," was published in 1956 and is not accessible in the MEDLINE database [5, 6]. Current scholarly activity, as evidenced by citation bibliographies, is generally limited to work published in MEDLINE. Most researchers assume that the 30-odd years covered by MEDLINE is all that needs to be researched in order to publish [7]. Indeed, in this recent 17-month survey of issues from the AJR, only one article of every 38 contains a citation from articles published in the 1950s. Previous works may disappear from memory, and valuable information will be lost. However, in this era of limited resources, money and time should not be wasted rediscovering what is already known. For example, if researchers were aware of publications in the 1920s dealing with gallbladder interventions, these researchers would be aware of some problems seen in their own research [8]. By not accessing the older literature, we researchers may miss numerous opportunities to learn from the work of our predecessors. The World Wide Web excels as a method of accessing and delivering information. This method, using commercially available software, is convenient for searching previously hard-to-find information.
References
commerce/hearings/0301lin.pdf.
Accessed October 15, 2001
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |