Open-source guides imaging business tool development
A team from Brigham and Women’s Hospital in Boston, using freely available open-source software, has developed a prototype model of business intelligence (BI) tools to take advantage of all data available throughout a department. Their work was presented last week in a scientific poster presentation at the 2008 Society for Imaging Informatics in Medicine (SIIM) conference in Seattle.
“The advantages of virtually merging relevant data from various sources into one single format and location are obvious and one can easily envision how it can contribute to the quality and efficiency of knowledge discovery,” the authors wrote. “This concept is not new and has been used for many years in the business world through the adoption of data warehouses.”
The team at Brigham and Women’s, led by Luciano M.S. Prevedello, MD, utilized BI concepts to create a three-dimensional database from data streams available from their imaging department’s relational databases. Multidimensional database structures permit data cubes to be built that allow easy generation of graphical representations of the three variables that constitute the building blocks of the cube, according to the authors.
“For example, the data cube could be built with the dimensions time, modality and the number of exams, such that one could request what the number of exams is, sorted by modality on a particular date and time,” they wrote. “The dimension time could be organized in a hierarchical fashion making it possible to organize it by year, month or day.”
To demonstrate their BI model, researchers constructed a micro-environment prototype with open-source software in their medical imaging department. The team deployed a desktop PC running Microsoft Windows XP as an operating system. Onto this platform the loaded MySQL Server 5.0, Pentaho 1.5.5 as a data mining application, and Kettle 2.5 for integration in order to extract, transform and load data; input files were in XML, Microsoft Excel, and HL7 formats. With the tools, they were able to generate composite graphs and reports for MRI examination throughput in the department.
The team noted that the knowledge acquired in the process could be used to adjust or change current behavior and help the organization to move toward optimal processes or goals. They also believe that their multidimensional database structure for data mining should result in “very significant speed improvements when used in large databases.”
“Radiology departmental MRI examination throughput was the example test case; however, it is envisioned that this micro-system may be useful in understanding many other activities and processes in a digital medical imaging entity,” they observed.