TRIP evolves to keep up with image overload

April 30, 2006

A status update and news of the latest initiative launched by the Transforming the Radiological Interpretation Process drew a near capacity crowd on Saturday, at the final session of the Society for Imaging Informatics in Medicine meeting. TRIP's mission is to identify more efficient ways for imaging informatics to deliver high-quality healthcare.

A status update and news of the latest initiative launched by the Transforming the Radiological Interpretation Process drew a near capacity crowd on Saturday, at the final session of the Society for Imaging Informatics in Medicine meeting. TRIP's mission is to identify more efficient ways for imaging informatics to deliver high-quality healthcare.

TRIP was launched in 2002 as an attempt to manage image overload spawned by new digital technologies. At the time, the TRIP committee predicted that by 2006, daily workloads would increase to 80,000 images requiring 22.2 hours to review and interpret.

The committee's first mandate was to learn from others. A literature review spanning multiple industries was conducted to learn how individuals navigate and interpret large data sets. To the committee's surprise, very little had been published about the psychophysics of interpreting large data sets, said Richard Morin, Ph.D., chair of SIIM and a medical physicist at the Mayo Clinic in Jacksonville.

Of even greater surprise was the irrelevance to medicine of the experiences by organizations such as NASA and the U.S. Geological Survey, among others.

"These groups manage large data sets very efficiently," Morin said. "But they work on a single set of images over a period of days, whereas radiologists need to work with a day's worth of images in a span of minutes."

Of even greater surprise to TRIP members was the interest of NASA scientists in their endeavors. NASA noted that some aspects of diagnostic imaging data management were so innovative that the agency itself would benefit from adopting them.

Unique factors about working with a large data set of diagnostic images include:

  • the urgency of results;

  • safety limitations and restrictions;

  • the cost of error and omission may be fatal to a patient;

  • tremendous variability of human data within and among individuals, one of the primary reasons that humans are utilized to interpret this data;

  • the difficulty of validating performance; and

  • poor understanding of human perception and its relationship to the art of medicine.

The challenge is an interdisciplinary one, requiring individual and coordinated contributions by scientists, engineers, researchers in academia and industry, vendors, and radiologists. Since its implementation, TRIP has acted as a broker and ombudsman to bring these groups together.

"The problem hasn't gone away. In fact, it has worsened, just as predicted in 2002," Morin said.

He noted that if an ACR protocol dictated radiologists spend one second to review each image in a 1500-image exam and five minutes to formulate a diagnosis, each exam would take 30 minutes.

"Given a typical caseload, a radiologist chained to a chair working without food, water, or experiencing any distractions would complete only 16 procedures in a day," he said.

But the reality is that radiologists are interrupted, and take breaks throughout the day, yet still manage to complete more than 16 procedures a day, Morin said.

"They have figured out a workaround, primarily using stack images and 3D visualization tools. But they also are getting tired trying to keep up with the deluge of large volume studies," he said.

At a meeting earlier this year, the TRIP committee refined its priorities to focus on human perception and image navigation. At SIIM, the group announced that its industry partners will install software at designated customer sites to record precise measurements of how radiologists work with large data sets.

Information collected will be analyzed to determine:

  • time spent interpreting chest CT and other high volume exams;

  • the number of mouse clicks radiologists make during an average exam review;

  • which software tools are used; and,

  • overall time spent interpreting exams.