Almost a year after the official launch of the Yottalook™ radiology-centered search engine, developers say they’re pleased that it is not only rapidly growing but also evolving.

Yottalook™ co-developer Khan Siddiqui, M.D., demonstrated the radiology-centered search engine during RSNA 2007. Developers are monitoring use of the site in order to refine its capabilities.

Yottalook was originally designed to use specific search algorithms to sort out radiologic content on the Web, thereby making it much easier for radiologists and other imaging specialists to find useful information. The radiology engine finds pages using the Google™ index, but its algorithms use radiology ontologies such as RadLex® to filter the results. Yottalook also drills down into teaching files, research papers and journals, opening up an “invisible web” that mainstream search engines either do not or cannot access.

Since launching the engine on January 6, 2007, Yottalook’s developers have focused on improving its functionality by monitoring exactly how and why radiologists search for information.

“They are using it exactly the way we imagined they would—to answer clinical questions at the time of interpretation,” said creator Khan Siddiqui, M.D., during an RSNA 2007 presentation.

The Yottalook™ radiology-centered search engine is also available on the RSNA.org homepage.

Dr. Siddiqui, chief of imaging informatics and Cardiac CT/MR imaging at the Veterans Affairs Maryland Health Care System in Baltimore, developed Yottalook with Woojin Kim, M.D., William Boonn, M.D., and Nabile Safdar, M.D.

“The queries that we see coming in are not research-related queries, not general queries, but very specific to in-progress imaging interpretation tasks,” said Dr. Siddiqui. “One-fourth of all of our searches happen after midnight, meaning people are using it while on call. They’re reading cases and looking for immediate information.”

Dr. Siddiqui noted that Yottalook received around 100,000 page views during August 2007. This number jumped to 213,000 in November, when it was added to the RSNA.org home page and demonstrated at RSNA 2007.

One of the major recent additions to Yottalook is Yottalook Images, which, like Google Images, allows users to preview thumbnails of search results. Unlike its Google counterpart, Yottalook Images can search through online peer-reviewed medical journals and display copyrighted images. According to Dr. Siddiqui, Yottalook Images contained 460,000 images as of November 2007 and continues to grow daily as Yottalook crawls journal sites and identifies more images.

The site also now features Yottalook Book, which uses the Google Book search function but filters results to include only imaging-related books. Radiologists seeking general anatomical information can use Yottalook Anatomy, which searches anatomy sources ranging beyond its usual index of radiology sites.

With almost a year’s worth of search query data at their disposal, the Yottalook developers are better able to fine-tune the engine, Dr. Siddiqui said. “It’s the psychology of search and information retrieval,” he said. “Users may not know what they are looking for, so the first thing they do is look for something that they already know or that is somehow related to their search target. In the process, they find the exact terminology they’re looking for and then change their search behavior accordingly.

“We track and analyze these search patterns, and the results allow us to optimize our search capabilities,” Dr. Siddiqui continued. “We can then develop an algorithm that predicts what radiologists are actually looking for and helps them get that information.”

Dr. Siddiqui used the term “AVN” as an example. The acronym can represent both “avascular necrosis” and “atrioventricular node.” However, the developers have already noted that 25 percent of all searches coming to Yottalook are related to bone disease. “For ‘AVN’ we automatically launch the avascular necrosis part of the search, because it’s related to bone disease. But at the same time, we also ask ‘Do you mean atrioventricular node?’ We’re providing guidance based on past search strategies, trying to understand what users are looking for, and then modifying our own strategies for getting the right information up as quickly as possible.”

Moving forward into the second year, Dr. Siddiqui said the Yottalook creators would like to expand the site to cover all medical imaging searches, not just those specific to radiology. “We need to understand what kind of images or information non-radiologists need and then figure out how to optimize the engine for them,” he said.

Note: This article was adapted from a story that appeared in the RSNA 2007 Daily Bulletin. The daily   newspapers from the annual meeting are available online at RSNA.org/bulletin.