While data mining has been a hot button issue of late, with some hospitals criticized for using information gathered from patient records to market to well-to-do patients, researchers in Germany believe a tool they've created puts such efforts to much better use.
Software developed by two German universities and a semantic knowledge management solutions company can help medical professionals analyze the free-text data linked with medical images in radiology reports, making for easier data mining efforts, AuntMinnie.com reports.
According to the researchers, the tool--which was the basis of a presentation at last week's European Congress of Radiology--improves access to images, while at the same time differentiating between similar terms used in conjunction with those images.
That improved access, AuntMinnie.com reports, ultimately could help to confirm disease diagnoses, as well.
The tool can pull and sort information based on size, time and date, image references and abbreviations, among other factors, according to Philipp Daumke, a co-founder of German-based Averbis GmbH, which helped to create the software. It provides "semantic links" between images and descriptions in reports.
It also was able to identify more than 32,500 image references--and retrieve more than 32,100 medical images--from more than 133,300 radiology reports, according to AuntMinnie.com.
"There is a huge variation [in terminology], and we are able to find all of these variations," Daumke said in his presentation.
Health IT solutions vendor Medstreaming announced at the recent Healthcare Information and Management Systems Society's annual conference in Las Vegas the launch of a web-based tool that enables customers using its electronic medical record to mine any EMR information online, which includes imaging data.
As we reported last summer, a data mining tool, when used in combination with a clinical decision support system at Massachusetts General Hospital, helped to justify pricey imaging tests. The tool looked at radiology data in conjunction with an algorithm and reported similar patient characteristics.