Algorithm addresses match of CT and MR data sets

Later this year, Carestream Health will release for testing a PACS algorithm that automatically registers several data sets, synchronizing slices to allow comparison of present and prior CT and MR exams. By early next year, this capability is expected to be atthe fingertips of Carestream PACS owners.

The algorithm registers related data sets in a fraction of a second. It can do so using data from the same modality, such as CT, or from different ones, as in the case of CT and MR imaging.

It performed flawlessly in demonstrations done in May for Diagnostic Imaging at both the Stanford International Symposium on Multidetector-Row CT and the Society for Imaging Informatics in Medicine meeting. At the MDCT conference in Las Vegas, Haddas Padan, director of product management at Algotec, the Carestream company that developed the algorithm, picked a point from a specific slice in one data set. In a moment, the algorithm took over, automatically finding the same point in a different data set, then presented the two slices in a split-screen format.

Reformatting the data can be helpful when looking at lesions or vasculature, said Eddie Moore, senior sales engineer at Carestream, who demonstrated the algorithm at the SIIM meeting.

"We are able to look at vessels and see the actual diameter," Moore said.

A major goal is to improve the efficiency by which radiologists compare studies by allowing these comparisons on a standard PACS rather than sending them to dedicated workstations or placing them on thin-client servers. But certain side benefits will be at least as important.

"Time savings and flexibility will be appreciated, but the most important contribution is that my confidence in the diagnosis will improve," said Dr. Matthew Bassignani, an associate professor of radiology at the University of Virginia, who watched as Padan put the Carestream algorithm through its paces.

Bassignani, who serves as medical director of UVA Imaging Centers, currently links present and prior exams manually for an average 50 cases per day. This requires the synchronization of between 100 and 150 data sets. With this kind of workload, the two or three minutes needed to link data sets for each case adds up to hours by the end of the day.

The kind of autoregistration promised by Carestream's new algorithm not only finds the exact same points in the data sets but aligns the surrounding data to the same plane as well. This may be needed simply because the patient was scanned in a slightly different position than during previous scans. Or the data may be fundamentally different, as occurs when comparing CT and MR results, an increasingly common practice after cancer treatment as oncologists try to reduce patient exposure to ionizing radiation.

"Our clinicians are starting to alternate between CT and MR in hopes of decreasing the amount of radiation that their patients receive," said Bassignani, who provided clinical commentary during the demo at the MDCT conference.

"Whether we go to only MR for all oncological imaging in the future or continue to alternate CT and MR, the situation will get increasingly complex. This software will match the CT and MR data sets, which is something you can't do now."

Whereas CT scans are typically formatted in the axial plane, in which the data were acquired, MR data may be reconstructed in virtually any plane, including the oblique. Because the Carestream algorithm registers data points in 3D space, the data can be reconstructed in any plane.

"Without this, you have to do the comparison in your mind," Padan said. "Here, the software actually shows it to you."

Underlying the enhanced visualization possible by interlinking points in the volumetric data sets is the reproducibility among the various studies and different radiologists. Manual interlinking is skill- and time-dependent. Padan noted that the algorithm uses the exact same process to interlink data sets time after time.

And it does so in less than a second.

"Autoregistration allows radiologists to get the software to do the work for them," she said. "The software looks at the volume of each data set, finds the registration matrix between those different data sets, and brings the results to the radiologist. From the radiologist's point of view, it is just a click away."

The interlinked data sets can be scrolled on screen, one slice giving way to the next in lockstep. Or, if radiologists have marked key frames of interest in prior studies, they can jump directly from one key frame to the next with a single button push while current and prior studies in split windows match exactly. This allows specific structures or lesions to be compared visually or quantitatively, determining whether a tumor or aneurysm, for example, has increased or decreased in size.

And because this capability will be part of a PACS, the user has immediate access to features, such as bone subtraction and vessel segmentation, that allow further analysis, Padan said. In the demonstration at the MDCT conference, she displayed two data sets in split screen on the left-hand monitor, while showing other postprocessed views, including a 3D reconstruction, on the right-hand monitor.

"With this we can easily take measurements that are later inserted automatically into reports," she said.The report, generated using the Carestream PACS, can be further automated to include key images, as well as DICOM information that describes how the study was done.

As the demo ended, Bassignani put the increased efficiency in the context of radiology's big picture.

"When a clinician wants to know if his treatment is working, I have to be able to say, yes the patient is getting better or no he's not getting better. I can't give him 'fuzzies,'" he said. The way I can increase my confidence will be through technology like what we are talking about here."