Managing PET-CT Images

The Integrating the Healthcare Enterprise joint activity by HIMSS and RSNA, which defines profiles based on standards such as DICOM and HL7, just released the new PDI (Portable Data for Imaging) integration profile as one of several supplements in their draft for trial implementation.

This integration profile specifies how images and related information can be exchanged among imaging devices using CDs. In addition, there is an invitation for vendors to participate in an RSNA demonstration. The interesting part of this demo is that attendees of RSNA 2004 will receive a sample IHE-conformant CD with sample medical images and patient information. Vendors participating as so-called IHE Integration profile Media Creator actors also will be able to generate sample CDs on site in their exhibit booths to show their capabilities to create IHE-conformant media. Vendors participating as Media Reader actors will be able to show their capabilities to import IHE-conformant media. RSNA will truly show IHE in action.

The creation of CDs for the exchange of images in a standard, DICOM manner, has been defined for many years. Initially, this was mostly used for niche applications, such as to store cardiology runs, or for cine loops from portable ultrasound. However, this application has become suddenly very popular. The reason is simple: cost reduction. If you have the images available in digital format, burning a CD costs very little, and you can easily train your file room personnel to do it. For example, a major hospital in the Dallas-Fort Worth area will gladly provide a copy of an exam for a

Among the considerations for effectively managing PET-CT images are: how do you deal with raw data? (and how long do you save it?); how will you handle reconstructed images?; and how will physicians access prior studies? For some solutions, read on.

The PET-CT business is booming. Experts predict that sales of PET-CT scanners will overtake sales of PET scanners this year and could completely replace PET scanners in the next several years. As sites clamor to deploy these high-power hybrids, it is critical to develop a plan for handling PET-CT images.

These plans can be logistically and technically challenging. Nuclear medicine departments tend to be distinct entities that aren't necessarily integrated with radiology departments - or enterprise PACS. Unlike other scans which can be reviewed as they load, all data from PET-CT studies must be loaded before fusion images can be generated, which can stress the PACS and network infrastructure. Furthermore, raw PET-CT data is memory intensive (400 to 600 megabytes per study) which can stress even the most powerful PACS. Key questions to consider as the PET-CT image management plan is developed include:

  • Will the site save raw data? If so, how and for how long?
  • How will the hospital provide access to reconstructed data?
  • How will physicians access prior exams for comparison?

Currently, hospitals rely on a variety of solutions for PET-CT image management. The 'right' solution for any given site partially depends on the technology in place. Another key factor, of course, is the budget. Common image management options include CD or optical disk, and many sites do rely on good-old fashioned film for image distribution. PACS, single-modality PACS and network solutions are utilized less often, but do hold promise for streamlining the process, cutting down on film and providing more complete 3D views than film.


The first image management hurdle to cross with PET-CT image management is the vast amount of raw data this technology brings. With PET-CT scanners yielding unwieldy raw data sets of 400 to 600 megabytes per study, some sites archive raw data with the rationale that physicians may want to reconstruct or fuse images after the initial processing has taken place. Others save raw data for a short time after verifying that reconstructed data meet clinical needs.

Raw data can be a real drag on most PACS, so hospitals rely on other media to save raw data, in the short or long term. The Nuclear Medicine Department at Walter Reed Army Medical Center (Washington, D.C.) saves all raw data from its Siemens biograph PET-CT scanner on CDs. Jaime Montilla, MD, chief of nuclear medicine, says this solution is less than ideal. His primary concerns with the CD approach are long-term survival of data and turnaround time. CDs may degrade over time, and they can be lost or scratched.

University of North Carolina Hospital (Chapel Hill) has a different take on CDs. Raw data files are saved on CDs, in part because physicians cannot reprocess raw data or fuse PET and CT images on the hospital PACS. PET Technologist Stacy Hengsterman says the combination PACS and CD approach keeps options open. When a patient returns for a repeat PET-CT scan, techs can pull old images from a CD to fuse the old study with the new exam. Hengsterman confirms, "Retrieval time is not an issue with CDs." Raw data can be easily pulled from a Siemens biograph fusion station or e.soft workstation.

Another raw data option is optical disk, which provides large storage capacity and rapid recall, says Wayne Jacobs, MD, director of nuclear medicine at Sunrise Hospital and Medical Center (Las Vegas). Sunrise currently relies on optical disks for storage and will continue to archive raw data from its Philips Gemini PET-CT system on an optical drive after an enterprise PACS is installed this summer.

Brooke Army Medical Center (Fort Sam Houston, Texas) employs a slightly different approach to the raw data crunch. Yong Bradley, MD, chief of nuclear medicine, explains, "Memory is still a problem with raw data. It takes about three times as much memory to keep raw data as it does to save reconstructed exams." Archiving raw data generated by its PET-CT scanner would tax the memory of both the radiology PACS and the mini-PACS in the nuclear medicine department, so raw data files are stored locally on the PET-CT system for about two weeks after the scan. Bradley explains, "If something is wrong or we need to re-process images, they are available on the computer that does the PET-CT scan."

Thompson Cancer Survival Center (Knoxville, Tenn.) employs a similar short-term approach to the raw data generated by its CTI Reveal PET-CT scanner. Raw data files are saved on the hard drive on the day of the exam. After it is processed, archived and reviewed with the report, raw data files are deleted from the hard drive. The entire cycle usually takes less than three days, says Karen Woten, manager of Thompson's PET Imaging Center.


The next phase of the PET-CT image management plan entails reconstructed data. The good news is that processed images are significantly easier on memory than raw data. The hitch? There can be DICOM and IS issues between nuclear medicine and the radiology PACS. Redundant solutions are common.

Walter Reed Army Medical Center stores reconstructed PET-CT data on CDs in the nuclear medicine department. This system, however, is not necessarily state-of-the-art. To facilitate image distribution, the nuclear medicine department prints one set of images, usually about 20 coronal images, for the patient file. Within the last two to three months, the department added to this solution and began placing a few PET-CT image sequences on PACS, but Montilla is researching other approaches.

Montilla explains, "DICOM here [in nuclear medicine] is not DICOM there [in radiology]. PET's real value is in therapy planning, and we are underutilizing PET because our systems can't talk to each other." Montilla believes a dedicated PET-CT image management server (mini-PACS) is the solution. The server makes it easier to archive images and export images to the radiation therapy department. In addition, images would be readily available, and CDs would not be lost, degraded or scratched. Finally, Montilla concludes, "I would be more confident that we were meeting HIPAA requirements if images were stored on a secure server than in a storage room."

University of North Carolina Hospital has a relatively streamlined plan for reconstructed data. Reconstructed data, approximately 256 frames of CT data, 255 frames of PET data, 16 frames of coronal images and 32 MIPs, are stored on an Agfa IMPAX PACS. The hospital prints films for referring physicians from the PACS or burns images to a CD.

Sunrise Hospital and Medical Center relies on a similar approach with the nuclear medicine department printing representative images, typically one sheet of coronal slices, to store in the patient's film jacket. The remaining processed data files are not saved after they are read. Jacobs explains, "We can always reprocess the raw data if necessary."

After the PACS is installed, the department plans to archive reconstructed images on the PACS for quick retrieval. Jacobs says, "The primary benefit with PACS will be quick and easy accessibility to PET-CT images at the workstation. The hardcopy storage problem will be eliminated, and we no longer have to worry about lost film."

With the new PACS, referring physicians will be able to access images on the PACS, but the nuclear medicine department also plans to print images or burn CDs on request. Jacobs continues, "Because we're adding PACS after PET-CT, we know exactly what we're dealing with. This makes the transition much smoother."

Other nuclear medicine departments dealing with legacy PACS claim the connection between PACS and PET-CT can be problematic. For example, the PACS at University of North Carolina Hospital viewed PET-CT images as CT exams not PET studies. The nuclear medicine department found a way around the problem-manually typing in patient data. Thompson Cancer Survival Center plans to phase PET-CT into its McKesson Horizon PACS after its IS department develops an interface and patient identifier to accurately record patient information from two separate registration systems. Woten says another question that won't be answered until the system is up and running is whether or not the PACS has adequate image fusion capabilities for PET-CT exams.

Brooke Army Medical Center relies on two PACS to archive reconstructed PET-CT images. The nuclear medicine department has a DICOM 3.0-compatible Siemens Sienet PACS, and the radiology department has a GE Centricity PACS. The single vendor solution in the nuclear medicine department ensures integrated service eliminates finger pointing and simplifies compatibility and communication, says Bradley. All processed PET-CT images are saved on the mini-PACS and routed to the radiology department PACS. The redundant plan provides enterprise access to PET-CT images. Clinicians throughout the hospital use the web to access PET-CT images.

Thompson Cancer Survival Center relies on multiple means to share images with referring physicians. The center not only prints images and burns CDs for clinicians but also shares PET-CT images on its server via its Reveal Network Solution. A tech at the center creates and saves a dataset of fused images and sends an email with a link to the server to the requesting physician, who then accesses the images on his workstation. Woten notes, "This is another option for getting reconstructed data to physicians, particularly doctors in outlying areas."


Managing PET-CT images can be a complex, multi-tiered process. It's important to assess what image management technology is place both in the nuclear medicine department and in the hospital and consider what might be added at a sooner or later date.

"I would implement a server [for departmental PACS] from the get-go," Montilla says. "It's easier to design the set-up and troubleshoot before deploying the PET-CT scanner. It's harder to do this after you are up and running. You'll have to cancel patients and decide whether or not to back up prior studies on the server."

Memory is a key issue for PET-CT image management. Bradley says, "If a hospital has enough money, I recommend a RAID or SANS - memory that allows the hospital to build on it. You don't have to buy all of the memory right now. You can pay as you go."

Let's face it, the ideal PET-CT management plan may hinge on the purchase of readily integrated equipment, typically a PET-CT scanner and departmental PACS or PACS. In many cases, this may not be financially feasible, but CDs, optical disks and networks can be used to develop a workable PET-CT image management plan.

The Past, Present & Future of PET-CT Image Management

Johns Hopkins University Hospital (Baltimore) has a relatively lengthy history with PET-CT image management. The hospital installed the first commercial PET-CT unit, a GE Discovery LS, in 2001. Since then, the hospital has implemented three separate image management solutions.

The first solution was to share an enterprise PACS with the radiology department. But the system did not read PET-CT studies as a single study; it created multiple relations between the studies, which made data retrieval a time-consuming undertaking. The next step was to create a departmental PACS with Integra fusion workstations. Research Associate Jeff Leal explains, "To some degree, this was rather successful. We could transfer PET-CT data as a whole study and manage multiple series as a single unit." On the downside, workstations aren't designed to be servers and could not hold more than nine months of data, which did not meet the nuclear medicine department's requirement of 12 months or longer.

The most recent solution relies on a GE Centricity enterprise PACS as an archive device for the nuclear medicine department. Leal says, "Many PACS have yet to learn to talk to PET-CT cameras. By choosing a PACS from the same vendor, we forced the PACS to talk to PET-CT." PACS workstations are useful for data management, but not PET-CT review, so the hospital uses Integra and Xeleris workstations to review fusion images. Leal admits, "It's still not perfect. We're not [totally] pleased with the speed at which data moves around. Part of that is the network infrastructure, and part of that is the increasing amount of data."

Leal predicts image management issues will continue with second-generation PET-CT cameras, which use a 256 matrix PET camera to generate more and denser slices. On the CT side, the new higher slice count scanners create thinner slices of varying thickness. "This will further force the adoption of super-fast networks and super large storage devices. A few years ago, we purchased 10 terabytes of storage for the Discovery LS and estimated that it would last seven years. If we moved to a [next-generation] Discovery ST, I think 10 terabytes would only last a few years."