Expectations for tomosynthesis debut rise amid nagging doubts

December 28, 2007

Not since the height of the CT slice wars have vendors said so much about products that were so far from market. They have been talking up breast tomosynthesis for years, adding details with each passing RSNA meeting, whetting the appetite of a marketplace enamored of digital imaging. Looming larger, year after year, until it can no longer be ignored is the question-spoken or unspoken-that begins with the word "when."

Not since the height of the CT slice wars have vendors said so much about products that were so far from market. They have been talking up breast tomosynthesis for years, adding details with each passing RSNA meeting, whetting the appetite of a marketplace enamored of digital imaging. Looming larger, year after year, until it can no longer be ignored is the question-spoken or unspoken-that begins with the word "when."

Hologic, which owns more than half the U.S. market for full-field digital mammography sales, has an answer. Its RSNA booth focused visitors' attention on a tomo system that Hologic executives described as being in its final form, predicting that the company will have a tomo-capable unit for sale in the U.S. this year, possibly by summer.

GE Healthcare and Siemens Medical Solutions, the number two and number three players, respectively, in the U.S. market for FFDM, showed tomo systems on the RSNA exhibit floor as well, along with plenty of clinical data demonstrating their potential. But the companies made no predictions about when the units would be ready for release.

The final arbiter, they say, will be the FDA, the bureaucratic equivalent of a huge question mark hanging over tomosynthesis, which Hologic recognizes but does not see as a problem. In the end, what the FDA will say and when the agency will say it may be the least of all the uncertainties facing tomosynthesis.

While U.S. regulators hold absolute sway over which systems ultimately go up for sale and when, those questions barely hint at the complex issues that underlie commercialization of this technology.

Unknown factors relating to workflow punctuate basic uncertainties surrounding tomosynthesis, uncertainties as to how products will operate and how they will handle the volumes of data these systems will deliver. That so much is unknown is brought into sharp focus by the steep regulatory path tomosynthesis will have to climb, one that separates this technology from the majority of medical imaging devices.

Unlike other products, systems offering tomosynthesis must get FDA approval, not just clearance. The relatively quick and easy 510(k) process paves the way to the U.S. market for most radiological systems, but this process will not apply to tomosynthesis. This technology, characterized by the 3D acquisition and reconstruction of x-ray images of the breast, must go through a premarket approval (PMA) process, which requires clinical tests to assure safety and efficacy. Instead of the 90-day review period for 510(k)s, FDA staff will have twice that long to review tomo products. They then can follow up with as many questions over as long a period as they like.

None of the companies believes the PMA process will be easy. But Hologic executives say the process will not be overly challenging.

"We believe the technology, the science and the medicine are good, and we believe the FDA agrees," said Andy Smith, director of imaging science at Hologic. "We don't see enormous impediments."

The fact is, however, no one knows for sure. Digital mammography has a limited track record. The FDA has approved just four FFDM products. Three of these-Senographe from GE, Selenia from Hologic, and Novation from Siemens-are capable of evolving to support tomosynthesis. The last, Siemens' Novation system, was approved in 2004; the first, GE's Senographe 2000D, in 2000.

DMIST RESULTS

It wasn't until late 2005, when results from the Digital Mammographic Imaging Screening Trial were published, that FFDM became widely accepted as an unqualified improvement over film-based mammography. Tomosynthesis promises to be better than FFDM, but the data to prove those promises are not yet in hand, forcing proponents to speak in hypotheticals.

All three companies cite the same basic argument in favor of tomosynthesis, that acquiring projections at several angles on an arc across the breast has the potential to reveal lesions that would otherwise be obscured. The multiple views can provide the parallax needed to separate lesions from dense fibroglandular tissue. Patients may also experience less discomfort, as less compression may be needed. Neither of these advantages comes as a surprise to the imaging community.

Vendors have been singing the praises of tomosynthesis for years. At the last RSNA meeting, their voices rang out louder than ever before.

Siemens displayed a prototype of its tomosynthesis device along with a viewing station. GE dedicated a special area of its booth to tomosynthesis but kept most visitors away, ushering in only selected customers who could see a version of the system currently in use in clinical trials.

"We did this to get feedback, possibly to find future collaborators," said David Caumartin, general manager for global marketing at GE.

Hologic dedicated a section of its RSNA booth to demonstrating how 3D mammography might be implemented. The company showed a system capable of tomosynthesis and a specially designed workstation that created an interactive exhibit allowing visitors to test their skills in reading 2D and 3D images.

Tomo goes back to the mid-1990s, when a research group at Massachusetts General Hospital began looking into the idea years before the first FFDM system was commercialized. These two variations on the digital theme took root so close together because they share the same basic technology.

Now, a decade later, the solid-state detectors being used in FFDM systems are nearly the same as the ones their makers expect to install in units capable of tomosynthesis. The biggest difference will be a faster readout of data. Vendors expect to take a dozen or more exposures of the breast in rapid order. This requires the detector to capture and transmit the data, then very quickly wipe clean the "digital slate" so the next exposure can be made.

Hologic has already developed the electronics for doing this. The company could begin building them into its Selenia FFDM system even before tomosynthesis is ready for the general market. If Hologic can do this, so can Siemens, which buys the flat-panel detectors for its Mammomat NovationDR from Hologic.

Siemens might switch to a different supplier of flat panels, Analogic subsidiary Anrad. But if it does, Anrad would have to meet the required specs, according to Jonny Eser, Siemens product manager for women's health, whether it is for FFDM or tomosynthesis.

X-RAY FINE-TUNING

Just as the detector may need only some fine-tuning, so might the x-ray tubes. And these changes may already be in place before tomo is widely available. Hologic, for example, is transitioning to a tungsten-based tube as a means to improve performance and limit dose. The first of these tubes are expected to appear on Selenia systems designed for FFDM.

When delivering x-rays with a tomo system, GE plans to build on the approach it currently uses with its Senographe portfolio. These products use the first few milliseconds of exposure to calculate the density of the breast, then adjust the power level and align either the molybdenum or rhodium target in the tube to deliver the optimal dose of x-rays.

The most apparent difference between current FFDM systems and units capable of tomo will be the motor that drives the detector and x-ray tube across the breast. This, however, may present the least difficult challenge of any facing the developers of tomosynthesis. Current FFDM units already allow the arm holding the imaging chain to move in an arc. Only a simple drive mechanism, using software to allow precisely timed exposures, need be added.

Lurking beneath the surface, though, are a lot of thorny issues that, despite years of work, still remain.

One is how to handle the incoming data. The multiple exposures in a tomo scan could generate up to a gigabyte of information. And these are just the raw data. Three-D data sets will have to be processed into slices, which together could create a data set 10 times as large as the raw data.

Storage is getting cheaper, so the data generated by tomo may not be a big issue. But the time needed to process these data and the computing muscle to do it will still be an issue.

"You might be able to make fantastic images, but what good are they if they take hours to reconstruct?" Caumartin said.

Radiation Dose

Adding a dimension of its own to the problem is the question of how many projections should be obtained in an exam. In its clinical trials, Hologic is using a 15 degrees arc and 15 exposures taken over five seconds. Other manufacturers are leaning toward a number between 11 and 15, but some sites are experimenting with 25 or even 50 per arc. It is not yet apparent, they say, which is the optimal number.

Then there's the issue of radiation. Would 15 images rather than one make the risk all that much greater? All three vendors answer with a resounding no.

"The total radiation for a tomo scan will be the same as for a conventional digital mammogram," Smith said.

The machines will split the dose delivered in a conventional mammogram by the number of projections taken in tomosynthesis, he said, a strategy echoed by executives from Siemens and GE.

Any one of the projections would translate into a very noisy image, but these images will never be made. The underlying data will be processed by algorithms into a 3D model that can be processed again into slices. These algorithms will work much like those that process 3D data sets in CT and MR.

Siemens and GE have been developing such algorithms for decades, and they are migrating some of that knowledge to tomosynthesis.

The time needed to interpret an exam cannot increase. If anything, tomosynthesis should be more efficient, so that productivity can increase.

"We will not have more radiologists or fewer patients," Caumartin said. "So the challenge is to develop a workflow that will make tomo effective and easy to use."

This could be tough. Tomosynthesis cannot produce images in familiar formats, which means early adopters of this technology will have to learn to read breast images all over again, which could affect productivity.

There is the added consideration that digital mammography itself is relatively new. Only about 20% of the 1300 sites performing mammography in the U.S. have invested in FFDM systems. Many of these signed on only in the last couple of years. For the rest, jumping into tomosynthesis from film-based mammography could be overwhelming.

Vendors plan to ease the medical community into tomosynthesis by offering a first generation of machines that can do both 2D and 3D imaging.

"If someone wanted to buy the 2D portion only, it would be exactly the same as our 3D tomo system," Smith said. "Then they could purchase a software upgrade in the future to convert to 3D. Such a system would use exactly the same image receptor."

This strategy has the advantage of protecting future buyers of digital mammography systems from obsolescence. Already in the near future, FFDM units will roll off assembly lines equipped to allow a field upgrade to tomosynthesis. And prospective customers not owning such systems can rest easy. They will find vendors eager to take aging systems as trade-ins on tomo-capable systems . . . when the time is right.

Greg Freiherr is business editor of Diagnostic Imaging.