Enter Your Email Address to Receive a
Copy of MedicExhange Member Demograhpics

"We are in the electronic age of radiology, therefore the quality of online resources is critical to our practice," says Dr. Daniel Rubin, Research Scientist (BMIR) and Clinical Assistant Professor at Stanford University.

To determine the knowledge radiologists seek, Dr Rubin and colleagues from Stanford University utilized ontology-assisted analysis of web queries from the ARRS GoldMiner and Yottalook web search engines. As a protégé project, they also downloaded the current release of RadLex.

But what do radiologists really look for? To find the occurrence of RadLex controlled terms, researchers created a computer algorithm that examines the raw search text and attempts to find the best corresponding RadLex term(s). Few radiology resources are indexed with controlled terminology. RadLex is a controlled terminology that enables processing and categorizations of queries submitted to web based radiology resources.

Due to the fact that users type their queries as unstructured text, Dr. Rubin found the computer analysis of query logs to be quite challenging as there was no automated method to deduce the information sought from such queries. "Providing controlled terminology resolves the many different ways to say the same thing. Do you know how many ways there is to say, and search for, interstitial edema/infiltrate?"

Using the Chi square test, researchers compared the frequency distributions of categories in RadLex with those in the user queries from the web sources. Using python scripts, they then implemented the query mapping algorithm and code to calculate term metrics. Dr. Rubin et al found that RadLex contains 7,413 different terms that each belonged to one category. Of the 29,669 unique search terms, 15,445 of the queries could be mapped to their algorithm.

Findings suggest that the utility of using RadLex to propose the types of information users seek and the value of RadLex from an 'ontological' perspective may be useful for analyzing and summarizing other types of radiology text, says Rubin. However, researchers found that the distribution of RadLex terms and categories for user queries significantly differed from the distribution of the categories within RadLex itself. In addition, they found that some RadLex terms were not used at all.

Generic search engine queries are providing an abundance of medical imaging results. "In lieu of traditional textbooks, radiology professionals are utilizing the internet for a number of reasons including as a clinical decision making tool,' says Dr. Richard Sharpe from the University of Maryland Medical Center.

Presenting at the 2008 SIIM meeting in Seattle, Dr. Sharpe and colleagues from the University of Maryland Medical Center investigated the variability of subspecialty search queries made using a radiology specific search engine. Though numerous generic search engines provide clinical, educational and research information, using these generic search engines has been found to be inefficient due to the fact that users must manually sift through all non-radiological and irrelevant items, states Sharpe.

To aid our daily queries, radiology specific search engines like GoldMiner and Yottalook have already been deployed. Combining Google's indexing technology, search engines like Yottalook.com search a subset of sites that are relevant to radiology, making the searches created by the radiology professional more efficient.

From August 1, 2007 to August 24, 2007, Sharpe and the team of researchers downloaded and retrospectively analyzed every 25th consecutive yottalook query. Two authors sorted the terms into 11 radiology sections. Category allocations were determined through the consensus using the criteria defined in the Diagnostic Radiology study guide produced by the American Board of Radiology.

Data analysis was performed using Microsoft excel and accepted methods for calculating 95 per cent confidence intervals for proportions with nominal data. Statistical differences were considered significant at the 95 per cent level.

Interestingly though, "Musculoskeletal imaging searches comprised the highest frequency of searches, with the next most frequently searched category being neuroradiology which had significantly fewer searches than musculoskeletal topics but considerably more than all others," said Sharpe.

One of the limitations researchers experienced in their methodology was the inability to assess the level of expertise of persons using the radiology search engine.