Breast tomosynthesis trials show promise

April 1, 2005

Years ago, tomosynthesis was cited as one reason radiology needed digital mammography. A digital detector could be arced around the breast to capture multiple views from different angles, stripping away tissue that obscures cancers. But despite the commercial introduction of digital mammography five years ago, the use of tomosynthesis remains limited to a few isolated medical centers.

Years ago, tomosynthesis was cited as one reason radiology needed digital mammography. A digital detector could be arced around the breast to capture multiple views from different angles, stripping away tissue that obscures cancers. But despite the commercial introduction of digital mammography five years ago, the use of tomosynthesis remains limited to a few isolated medical centers.

Hologic showed such a device at the last two RSNA meetings and plans to do so again next year, but only as a work-in-progress. Siemens exhibited the first images from its tomosynthesis prototype at the 2004 RSNA meeting. GE Healthcare, which built the first tomosynthesis device to be used in clinical trials, has been even more circumspect, keeping its technology out of public view.

Just four academic sites in the U.S. have any experience with tomosynthesis. One uses GE equipment, another uses the Siemens prototype, and two others use a Hologic design. (A fifth site, Bethesda Naval Hospital, has used the GE prototype to examine soft-copy reading and image processing issues.)

This low profile does not reflect the promising clinical results. Using a GE prototype to assess 450 patients, Dr. Daniel Kopans and colleagues at Massachusetts General Hospital observed improved sensitivity: They identified seven cancers not seen with 2D imaging. Specificity also increased, as false positives plummeted.

"We found in our practice that about 25% of women who are recalled because of an abnormal mammogram just have overlapping normal breast tissue. Tomosynthesis completely eliminates that," said Kopans, director of breast imaging at MGH.

Researchers at Duke University received a prototype last October that is based on Siemens' Novation DR full-field digital mammography system. Much of their work has focused on the technology itself. But cases demonstrating the clinical value of tomosynthesis have already cropped up, according to Dr. Jay Baker, chief of breast imaging at Duke. The day before his interview with Diagnostic Imaging, Baker had reviewed a case in which a tumor was evident on tomosynthesis but not on screening mammography.

"Once we learn how to read these things routinely, this is the sort of lesion that will not be passed by," he said.

Hologic prototypes have been used to examine more than 200 women for breast cancer. Dr. Laurie Fajardo and her research team at the University of Iowa have studied about 100 of them. All the patients had been referred for a diagnostic study. Fajardo found tomosynthesis helpful, particularly in examining dense breasts.

Not surprisingly, the equipment makers are the most enthusiastic advocates of tomosynthesis. Andy Smith, principal scientist at Hologic specializing in tomosynthesis, predicts that the technology will revolutionize mammography, eliminating needless recalls while detecting cancers that would otherwise be missed.

"With this comes the potential for greatly reduced healthcare costs, because women are not coming back for workups they don't need," he said.

Such savings may be years away, however. Proving the clinical value of tomosynthesis will be challenging. True positives of the type reported by the clinical sites now operating are relatively easy to verify, as lesions seen on tomosynthesis can be confirmed by biopsy. The difficulty lies in proving that breast tomosynthesis is not missing lesions.

Researchers must establish the "ground truth" for false negatives, which means patients must remain free of lesions for at least a year after a negative report in order to receive a clean bill of health. This delays getting the product to market.

An additional issue is whether breast tomosynthesis should be used for diagnostic or screening purposes. Proving the diagnostic value of a mammography device is easier than proving its efficacy as a screening tool. Thousands of women may have to be screened to detect a single cancer.

Digital mammography struggled for years to surmount FDA red tape, but the challenge for tomosynthesis may not be as daunting. Kopans discussed its potential with FDA officials about a year ago.

"They were very supportive," he said. "I don't think getting FDA approval is going to be that difficult once the advantages of this technology are clear."

CLINICAL USAGE

The MGH team got off to a fast start, landing the first tomosynthesis clinical grant seven years ago. But the GE prototype on which they based their early work was damaged in a move in February 2003 and could not be repaired. The group has been awaiting its replacement so they can begin using tomosynthesis to screen 3000 women for breast cancer under a National Institutes of Health grant awarded last year.

"With this pilot we will have no problem proving scientifically that the recall rate is reduced," Kopans said.

The prototype, based on GE's Senographe 2000DS, is expected to arrive this year. The wait is frustrating, considering that eight years that have elapsed since this work began.

Kopans believes GE could have introduced a commercial product two years ago. GE's top executive, Joe Hogan, agrees, but claims that doing so would have been ill-advised.

"You can't launch a product that doesn't have a workflow piece to it, that's not efficient, and that has no proven clinical efficacy," Hogan said. "We are working on this now."

GE has invested heavily in the technology. It has increased the size of the digital detector in the latest version of its full-field Senographe 2000DS to support volumetric imaging, modified the detector's crystal structure to increase detective quantum efficiency, and changed the electronics to take signals off the back of the panel for increased efficiency.

The company hoped to install the first of its Senographe 2000DS systems capable of tomosynthesis at MGH by late February. Nine additional sites could be installed before the end of the year.

"But before we install these others, we want to get some feedback from MGH," said Jean Hooks, general manager for GE global mammography.

The company is pressing ahead on another front as well, developing a number of advanced tomosynthesis prototypes for use by specific research sites. The University of Michigan in Ann Arbor, for example, plans to use such a system to examine the fusion of tomosynthesis with ultrasound.

All the prototypes currently in clinical testing are built around a flat-panel detector and an x-ray tube that arcs across the breast. But they do have differences. The GE and Siemens systems use fixed detectors, while the Hologic detector moves with the tube. The Hologic and Siemens detectors are based on amorphous selenium, which directly records x-ray strikes. GE's detector is based on amorphous silicon, which uses a scintillator to produce flashes of light when struck by x-rays.

None of these devices is likely to eliminate the breast compression that most women find uncomfortable, although theoretically, it could be avoided. The main purpose of breast compression in conventional mammography is to minimize the layering of tissue that can obscure lesions. The inherent ability of tomosynthesis to electronically de-layer the breast should eliminate the need to compress tissue. It might be necessary, however, to apply some compression to pull the breast tissue from the chest wall into the field-of-view.

Tomosynthesis does promise to cause less discomfort. Some sites are reducing the pressure they apply to the breast. And because tomosynthesis electronically slices through the breast, only one view-and one compression-may be required, rather than the two needed in conventional mammography.

BREAST CT

An emerging competitor of tomosynthesis may not require any compression, however. Researchers at the University of California, Davis have developed a dedicated CT scanner that orbits just the breasts. Patients lie face down with a breast positioned in an opening in the table, as in an x-ray-guided stereotactic biopsy procedure. The table surface surrounding the cutout is engineered with a swale to allow a portion of the chest wall to extend downward into the field-of-view. Gentle pressure is applied to immobilize the breast and pull the tissue away from the chest wall.

A single flat detector mounted on a gantry below the table rotates around the breast, acquiring 30 frames per second. A single scan generates 100 to 160 coronal CT images of the breast. Conventional mammography produces just two orthogonal images.

"We are excited about the image quality, and things look very promising, but we haven't done the necessary science yet to make any legitimate claims," said John M. Boone, a professor of radiology and biomedical engineering at UC Davis.

The necessary research is under way, however; nine of a planned total of 10 healthy volunteers have been scanned with the device at UC Davis Medical Center as part of a phase I trial. The first of a total of 190 women with breast cancer has been scanned in a phase II study.

Several U.S. research teams are developing similar CT-based prototypes. A group at the University of Rochester Medical Center has constructed a cone-beam CT system capable of breast scanning using the same detector as the one at UC-Davis. The device can acquire up to 288 2D projections for 3D reconstruction in a single scan within 9.6 seconds.

Both tomosynthesis and dedicated breast CT make use of flat panels. They differ, however, in the number and orientation of the images.

The Siemens prototype can take up to 49 low-dose images, although the actual number is likely to fall between 20 and 30, said Thomas Mertelmeier, Ph.D., senior research scientist for Siemens women's health technology.

Early tests at MGH acquired 11 images. Clinical sites using the Hologic technology are shooting the same number of images per arc. Hologic equipment acquires the images in 18 seconds, but scan time will drop to a few seconds on the commercial product, according to Smith.

Regardless of the number of images or the time period, the total dose of a tomosynthesis exam equals or is slightly lower than the dose delivered during a 2D mammography exam. This is achieved by dividing the dose by the number of views.

"That is the magic of it," Smith said. "You get the 3D image sort of for nothing."

Reading those data raises a different concern, however. Rather than reconstructing them as a volume, Fajardo reads tomo images in stack mode, running through them in sequence, one after the other.

"You can have them in a video loop or you can punch a button and scroll through them," she said.

Mertelmeier is convinced that computer-aided detection algorithms could play a role in managing the data. But developing CAD algorithms will not be easy. They may have to consider the frames immediately preceding and following any one frame, if they are to pick up the subtle nuances of cancer, Baker said.

"We will have to train the computers to look three-dimensionally, to look at certain locations in the same space of several frames that have been acquired," he said.

IMPLEMENTATION HURDLES

The technologies for acquiring, processing, and displaying data all must meet the approval of the FDA. Even then, tomosynthesis is not assured of widespread adoption. Its use must first be reimbursed.

GE has begun working with several major third-party payers to lay the groundwork. The company has agreed to share early clinical results with Blue Cross Blue Shield and other third-party payers as a foundation for later reimbursement decisions.

To ease the transition to tomosynthesis, manufacturers are considering the feasibility of equipment capable of both tomosynthesis and full-field digital mammography. Don Alvarez, Siemens' director of women's health, believes this issue will be addressed as the technology gets closer to market. But the design of the product could affect the company's approach to the FDA, according to Mertelmeier.

"Workflow issues and the FDA will be the keys to whether tomosynthesis gets broad application," he said.

The data to support regulatory submissions could start flowing in earnest this year. In addition to GE's plans to install 10 Senographe 2000DS systems by year-end, Hologic executives expect to extend their network from the current two sites to six.

Kopans is hopeful that breast tomosynthesis will reach the marketplace in the next few years. The potential health benefits demand it.

"Our prototype, with all of its shortcomings, was far superior to any other mammography system in the world," he said. "I wish it wasn't taking so long to get this technology out to women, where I believe it will have a major impact on cancer detection and diagnosis."

Mr. Freiherr is business editor of Diagnostic Imaging.