NEW ORLEANS - Fast image access is one of the strongest selling points of an enterprise-wide PACS. But sending too much image data can slow a network down, and force clinicians to sort through information they don't need.
In order to move data as efficiently and cost-effectively as possible, it’s helpful to track clinician image-access patterns, according to researchers from the University of Alabama Birmingham School of Medicine. Dr. Barton Guthrie, associate professor of neurosurgery at the university, discussed the topic at this week's Healthcare Information and Management Systems Society meeting in New Orleans.
According to Guthrie, PACS network architecture can be optimized based on clinician usage profiles. Appropriate routing strategies, intelligent pre-fetching, and customized displays can all be implemented based on usage patterns, he said.
Thus, usage profiles enable image resources to be matched to the clinical needs of users. The ability to precisely measure image resource utilization makes it easier to create effective contracts based on diagnosis-related groups (DRGs) as well, he said. And tailoring the system to users' needs maximizes provider time, since clinicians no longer have to search through unnecessary image data to find what they're looking for.
"What I really want to do is have a system that understands 80% or 90% of what I need," Guthrie said. "Speed of access is going to be solved (with an enterprise-wide PACS), but the real problem is shuffling through the data that’s available, and what you don’t want to do is assault a clinician with data that’s not necessary."
To define the elements of clinical use of medical images, including image content and patterns of use and workflow, the researchers retrospectively reviewed the image utilization patterns of the university's neurosurgical oncology department between April 1 and June 30, 2000.
The University of Alabama has an internally developed enterprise-wide clinical image management system (CIMS), which employs an active patient image repository (APIR). The APIR features a 300-GB RAID 5 system for caching that holds a study for eight weeks after its last query.
The archive is powered by a DICOM Archive Manager (Emageon, Birmingham, AL), an Oracle database, and StorageTek 9710 tape media (Storage Technology, Louisville, CO), storing a total of 5 TB of information. (The archive has since migrated to StorageTek’s 9840 tape media.)
The researchers used CORBA services logs and archive logs to track clinician image queries. In all there were 400 individual queries to the CIMS, including calls for 233 unique studies and 297 clinical sessions. Of the 233 studies, 75% were MRI, 21% CT, and 4% were CR studies, Guthrie said.
Of the 400 queries, 202 were for current studies (seven days old or less), while 198 were for historical studies (eight days old or more), according to the researchers. Sixty-eight percent of the CIMS sessions were for queries of current studies, while 50% were associated with a historical study. In 32% of the cases, only the historical study was viewed.
In CR, 10 studies were queried and nine were viewed, with an average of 1.7 series acquired per viewed study. The average number of series viewed was 1.7, and 100% of the acquired image data was viewed by clinicians, Guthrie said.
For CT studies, 49 were queried and 47 viewed, with an average of 2.5 series acquired per viewed study. An average of 1.2 views, and 80% of acquired image data per study, were viewed by the clinician.
MRI image data acquisition, however, was not as efficiently matched to the needs of this service, Guthrie said. The clinicians queried 174 studies and viewed 134. An average of 1.5 series were viewed per study, and only 23% of acquired image data per study was viewed by the clinician, Guthrie said.
"Unneeded patient data impacts the bottom line, and insufficient (reviewed) data impacts patient care," he said. "I’ve had some discussions with my radiology colleagues about this; of course their comment, and they may be right, was that well, you’re just not taking adequate care of your patient. We’ll see how that turns out, but it certainly points up the need for us to get together and get more effective (data acquisition) protocols."
Overall, 70% of requested studies were available in the RAID disk cache in the CIMS' current configuration. If desired, this figure could be improved by efficient pre-fetching or by enlarging the cache, he said.
Images required retrieval from the archive in 30% of cases, which took three to five minutes. However, access time has been reduced since the StorageTek 9840 media was adopted. Since clinicians' need to access the long-term archive is unpredictable and would therefore represent a complex pre-fetching pattern, the archive needs to be very responsive to queries, he said.
By Erik L. RidleyAuntMinnie.com staff writer
February 9, 2001
Click here to post your comments about this story in our PACS Digital Community. Please include the headline of the article in your message.
Copyright © 2001 AuntMinnie.com
