Another Dimension: Improving Breast Screening

Overlapping tissue within the breast can reduce the visibility of lesions or mimic the appearance of an abnormality. Dr Maureen Gillan of the University of Aberdeen tells Elly Earls how a new 3D technology – digital breast tomosynthesis – has the potential to significantly improve the accuracy of breast screening.

In standard two-dimensional film or full-field digital mammography (FDDM), overlapping tissue within the breast can reduce the visibility of potentially harmful lesions or simulate the appearance of something suspicious. This can lead to unnecessary recalls and biopsies, causing severe psychological stress for the women involved. Current mammography techniques are also less sensitive in women with dense breasts who are at a higher risk of developing cancer.

As the NHS breast screening programme (NHSBSP) is now inviting women aged 47 and upwards for screening (rather than 50 and upwards as it had been previously) as well as younger women who are judged to be at risk of developing breast cancer due to their family history, and because dense breasts are much more common in younger women and pre-menopausal women, the shortcomings of the current technologies are becoming a concern.

"The problem with 2D mammography is that you're trying to take a picture of a 3D structure," says Dr Maureen Gillan, research fellow at the University of Aberdeen. "Even when you're taking images from two views – from above and below and from side to side – the concentration of tissue in the middle of the breast can make things difficult. Overlapping tissue can make something look suspicious and, at the same time, cancers can get hidden because of overlapping tissue."

Digital breast tomosynthesis (DBT), a new method of obtaining high-quality images that uses digital X-rays and a computer to generate 3D images of the breast, is currently under investigation at the University of Aberdeen in a three-year study and could provide a solution. "It could be that tomosynthesis is more effective in dense tissues, and that's one of the things we'll be looking at in the study," says Gillan.

Reducing tissue overlap

"Overlapping tissue can make something look suspicious and, at the same time, cancers can get hidden because of overlapping tissue."

So how does DBT work? A newly developed form of 3D imaging, it has the potential to improve the accuracy of mammography by reducing tissue overlap. In DBT, multiple projection images of the breast are acquired from different angulations of the X-ray tube. The images are then processed using algebraic reconstruction algorithms to produce tomography sections through the breast. These sections can be viewed on a soft copy workstation either as slices or sequentially in a dynamic cine mode.

"You scroll through the slices, a bit like focusing in and out with a microscope," Gillan explains. "Things will come into focus as you go through the breast and then they will start to disappear. That's why it could make it easier to see small cancers."

This overcomes the limitation of overlapping breast tissue in standard 2D projection mammograms as well as improving the conspicuity of lesions.

As it is still early days, the basic physics of the technology have yet to be optimised and the range of angles, as well as the number of projections, differs depending on which company's machine is being used.

The range of angles over which projection images are acquired is called the tomography angle and the approach adopted by the different manufacturers can be classed as a narrow or wide angle. The Hologic and Sectra systems use narrow tomography angles of 15° and 11° respectively, the GE and XCounter systems operate at slightly wider angles of 25° and 26° respectively, while Siemens has a very wide tomography angle of 50°. The advantages and drawbacks of using narrow or wide tomography angles remain a matter for research and debate, but it is currently claimed that a wide angle provides better depth resolution while a narrow one enhances in-plane resolution.

The number of DBT projections also ranges from 15 with the Hologic system to 25 with the Siemens system.

DBT systems have only become commercially available in the past two years; consequently most of the published data has been acquired through small-scale pilot studies. And while it is still too early in the technology's evolution to use DBT in a randomised trial in a screening setting, the study at the University of Aberdeen, in which Gillan is involved, is a first attempt to ascertain whether the new technique could be better than 2D imaging.

"Reporting times are likely to be longer for DBT than for 2D examinations because the quantity of information to be processed."

The £1.6m trial funded for three years will be led by Professor Fiona Gilbert, University of Aberdeen. The study will invite around 7,000 women in Aberdeen, Glasgow, Manchester, London and Guildford to take part. The targeted population will be made up of women who have been recalled after an abnormal screening mammogram and younger women whose family history gives them a high or moderate risk of developing breast cancer and who are attending annual screening.

"This is a population where you'd be likely to find something abnormal, so it should be easier to compare the performance of DBT to 2D," says Gillan. "If you were to run the study with routine screening cases that contain <1% cancer cases you would have a very large number of cases that were absolutely normal."

Trial participants will receive a standard mammogram as well as a DBT examination, and the 2D and 3D images will then be collected and reviewed independently by radiologists at another centre.

"We'll be looking at the cancer-detection rates and the recall rates," Gillan notes. "Some of the readers will be sent only the 2D images for a case, some will be sent only the 3D images and some will be sent both. This will tell us whether one or other or the combination of the two types of imaging is better."

On top of this, the team will be conducting sub-analyses. "We'll be measuring density to see if the tomosynthesis works better with dense breasts," Gillan explains. "We'll also analyse which types of cancer might be seen better by one imaging modality over another, or if one is better able to find very small or subtle lesions."

Screening setting

The study will by no means be a definitive evaluation of the effectiveness of DBT, as the trial's experimental setting presents many limitations. The retrospective nature of the study means there will be no time limits in terms of reading or reporting. But in a screening setting, reading time would become much more of an issue. "In screening, cases need to be read and results sent out in a short timeframe, so another trial would need to be done in a screening setting to see, in practice, how effective DBT would be," Gillan acknowledges.

"In DBT, multiple projection images of the breast are acquired from different angulations of the X-ray tube."

Currently, it is accepted that reporting times are likely to be longer for DBT than for 2D examinations because the quantity of information to be processed by the observer is much greater.

Another thing that needs to be decided upon, following further studies and research, is exactly how DBT will be used – on its own or in combination with 2D, in a diagnostic or a screening setting, or both. "Initially we thought that perhaps it could replace 2D, but if you look at the publications more recently, people tend to be looking at a combination of the two techniques," says Gillan. "There's also a question mark over the visibility of micro-calcifications. It seems that they are seen better in 2D, although not many reports have been published on this issue so it's difficult to weigh up the evidence."

If, after further studies and research, it was decided that DBT was a good option for the NHSBSP– something Gillan feels would not be realistic for at least five years – there would be further challenges as hospitals would have to integrate the technology into their current systems.

"At the moment, the UK is in the process of moving over from analogue film-based mammography to digital mammography, and tomosynthesis is just a modification of digital mammography," says Gillan. "So the images collected will be similar, although the file sizes are larger, which has implications in terms of storage on the PACS systems."

Moreover, if DBT did show signs of promise, cost-effectiveness studies would have to be carried out by the centres in the NHSBSP. "If centres are replacing analogue machines with digital mammography systems, they might need to look at buying one that can also perform tomosynthesis," Gillan says. In April 2010, approximately one-third of NHSBSP centres had at least one FFDM unit; however, most of these units are unable to undertake DBT.

Although it is far from certain that DBT would be an effective addition to both the NHSBSP and other breast-screening programmes across the world, the potential of the technology is undeniable. And as awareness of breast cancer increases among younger, pre-menopausal women, the results of the University of Aberdeen study could have profound implications for both the diagnostic and screening settings.