3D screen - digital breast tomosynthesis

Mammography has come under scrutiny in recent years. In late 2011, it was announced that the UK's NHS Breast Screening Programme (NHSBSP) would undergo an independent review to determine whether the benefits outweighed the drawbacks.

The document published as a result of the study concluded that such initiatives conferred "significant benefits and should continue", estimating that 1,300 deaths a year could be prevented.

Nevertheless, there are many limitations associated with the 'gold standard' mammogram used in these circumstances. Is the technology due an update?

"Standard mammography is based on 2D imaging, whereby all the tissue information is projected onto a 2D plane," explains Dr Michael Michell from the UK's King's College London University Hospital.

Michell is a consultant radiologist at the university hospital's NHS trust, and director of the National Breast Screening Training Centre. An expert in all aspects of screening and diagnosis, he takes us through the challenges that are encountred with conventional mammography techniques.

Tissues and issues

The big drawback is the image produced in standard screening examinations can be difficult to read because of tissue overlap. This can cause two main problems in practice, he says. One is that significant pathologies - such as small cancers - may be hidden. Secondly, the resulting image might erroneously suggest that an abnormality is present.

"In clinical, particularly screening, practice, the result is that we don't see the cancer in some patients and they don't get diagnosed following screening. And in some instances, we recall patients for further tests when there isn't actually cancer present," Michell reveals.

Both are obviously worrying scenarios. False negatives result in 15-30% of cancers going undetected and can mean treatment starts too late. On the other side of the coin, unnecessary recalls cause much anxiety to the women involved and are a cost burden on health services. For every four women told to return for more tests, only one will be diagnosed with the disease.

Technical advances in digital mammography, however, have allowed the development of a relatively new technique: digital breast tomosynthesis (DBT), which has now become a significant addition to standard diagnostic imaging. A three-dimensional image is produced, which might make tumours easier to spot, particularly in patients with dense breasts whose tissue can make mammograms trickier to read.

DBT is performed with the breast compressed just like conventional mammography, but the X-ray tube moves across it in a limited arc, during which several low-dose exposures are made. Multiple images of the breast at different angles are produced.

The digital data that is obtained is then reconstructed in 1mm-thick slices, which can then be viewed individually in sequence, or put on a cine loop for doctors to analyse.

"Because we're just looking at thin reconstructive slices, it gets over much of the problems caused by overlapping normal tissues, so we can see some much more easily," Michell explains.

There's abundant evidence in support of the technology, much of it conducted at King's, but plenty of papers from the rest of the world, too. A recent trial conducted at Washington Radiology Associates, purportedly the largest to date, suggested the method was superior to conventional mammography.

Similarly, a study from the University of Pennsylvania found DBT lowered recall rates from false positives and increased cancer detection rate by 22%. The technique has shown promise at reducing recall rates in all groups of patients, including younger women, who are more likely to have dense breast tissue.

Clear winner

At the start of 2009, Michell and colleagues at King's commenced a trial comparing the reading accuracy of adding DBT with both standard digital 2D and now old-fashioned analogue film-screen mammography. Michell and his team decided to study a group of women who were recalled from screening due to some abnormalities seen on a conventional scan.

A total of 759 lesions analysed by the different techniques, and the results were clear-cut. Readers were found to be 79% accurate when interpreting film screen, which increased to 90% with 2D mammography and an impressive 97% when DBT exams were combined with each method.

"Unsurprisingly, tomosynthesis, when added onto the 2D imaging, was more accurate in terms of identifying and categorising malignant and benign conditions, reveals Michell.

"From that, we would predict using tomosynthesis in screening practice might well reduce the amount of unnecessary recalls and improve our ability to predict small cancers."

On the basis of Michell and colleague's results, King's brought DBT into its routine clinical practice for the work up of abnormalities found with 2D mammography. It's been a useful addition to the hospital's diagnostic repertoire, but what Michell is really interested in is if tomosynthesis could be used in routine screening practice.

He reveals that King's is thinking about setting up a large-scale UK trial, potentially commencing in late 2015, to elucidate this, but cautions that it will take a while to recruit sufficient numbers of participants and the researchers will have to follow those cases for several years to get final results. It sounds like a lot of hard work, but it's what the NHS will require if it is to recommend DBT as part of routine screening practice.

Right now, the jury's still out. The service has so far approved the use of tomosynthesis for assessments of lesions detected after screening, and many units have recently acquired, or are looking to acquire, DBT equipment for their assessment centres.

"In terms of implementing tomosynthesis in routine screening practice, I think the position of the national screening programme is that they require more data than is available to make decisions as to whether to implement it or not," says Michell. "Such data will probably need to come from a large, randomised study."

Storing up problems

Lack of data is not the only reason for prudence on the NHS's part, however. In a report explaining its position, published in December last year, several other potential drawbacks were pointed out. For one, the files associated with DBT are much larger than those acquired using 2D imaging. It's bound to have implications for local and archival storage. Similarly, the format in which tomosynthesis images are available might not be compatible with most PACS workstations.

Another worry noted is the radiation dose associated with DBT, which is typically about 2.2mGy for 50 to 60mm-thick breasts. It's not much greater than exposure in conventional mammography, but as 3D scans have to be performed in addition to 2D ones, a patient will garner twice the usual dose of radiation. This might be justified if DBT is used to diagnose patients with signs of the disease, but it's a tougher argument to make when screening the general public, the majority of whom will be healthy.

But with so much evidence to suggest that DBT is more accurate, why is an additional 2D scan really necessary at all? Michell explains: "I think there is a generation of radiologists at the moment who are used to looking at a 2D mammographic image, and it's on those images that the results of trials showing the effectiveness of mammography and screening are based."

As there's no data indicating tomosynthesis can replace 2D imaging completely, radiologists won't be able to move fully into another dimension. Most professionals agree that a combination of the two readings will ultimately produce the best results.

Double dose of reality

So how can the problem with double dosing be resolved in the future? It turns out technology has been developed that can present the best of both worlds, minus the radiation. It's called synthetic 2D and it uses an algorithm to produce a two-dimensional image from data acquired during DBT.

"In terms of radiation, if you combine tomosynthesis with a conventional 2D mammogram, of course it increases radiation dose," says Michell, "and the tomosynthesis examination by itself is slightly greater in dose compared with conventional 2D.

"Data from some trials done abroad shows that synthetic 2D is probably going to be of sufficient quality to be used instead of the 2D X-ray image, though."

If that's the case, it clearly resolves any question over radiation exposure, because it will be approximately the same for both techniques, and radiologists end up with a 2D and 3D information for the same dose.

Which is all very well, but of course, unless there's a clear argument for DBT to save lives, the NHS is unlikely to change its guidelines. After all, increased detection doesn't necessarily correlate to a decrease in mortality rate.

It's estimated that up to a tenth of breast cancers are currently treated even though they wouldn't have gone on to develop into anything life-threatening, so screening that detects more cancers only helps women if the test reduces their risk of dying.

"We can't say what the effectiveness of screening would be in terms of changing the mortality from breast cancer because we don't know whether tomosynthesis is picking up biologically significant growths, or whether it's identifying very low grade ones that might not have made any difference, or reveal themselves, in one's normal lifespan," says Michell. "The only way to answer that question is to do a randomised trial."