The main difference between digital mammography and film screen mammography is about logistics. With a direct digital image, you instantaneously see whether it’s good or not and don’t have to fiddle with cassettes, film holders, or film processors so it’s much faster for the patient and for the technologist.
Digital is better logistically, but in addition, because it is an electronic image, it can be manipulated using computers, and this has opened the door for things like tomosynthesis to improve on conventional two-dimensional mammography.
When I initially read about tomosynthesis in the late 1970s and thought about applying it to breasts, we couldn’t do it because digital detectors had not yet been devised that could collect the images to be read on the computer. The development of digital mammography has made tomosynthesis possible.
Basically, when we look at a standard mammogram (digital or film), we’re looking through the entire breast on the image so you see everything from one side of the breast to the other all stacked up. I liken this to a book with clear pages. If you hold the book up to the light, you can see all the words, but they are superimposed on top of each other. Each page is superimposed upon the other so you can’t tell what’s on an individual page and it’s hard to read. Similarly, a cancer on a standard two-dimensional mammogram can be hidden by normal breast tissue.
Tomosynthesis allows us to page through the book, looking at each page individually. Just as the words become much clearer by reading each page separately, seeing cancers becomes much easier on tomosynthesis.
With tomosynthesis, we take a series of low-dose mammograms from different angles so the breast is held in the same way that a standard mammogram holds the breast in the mammography machine. By taking mammograms from different angles we’re looking through the breast from different directions. This allows us to ‘synthesize’ slices (pages) parallel to the detector.
As we move the X‑ray tube through an arc and take pictures, the structures close to the digital detector will appear to move a shorter distance than structures that are further away from the detector. This is called parallax. We then take all the pictures that we have taken – right now we take 15 pictures – and have the computer line up the pictures so that everything that is in a given plane in the breast in the breast lines up.
We start one millimeter above the detector and have the computer line up all the images so that everything in a plane one millimeter lines up perfectly in all 15 images. Everything in that plane is reinforced 15 times and the structures in the plane are clearly seen. Everything that is in a different plane will misregister and will not be reinforced so you’re effectively seeing just the plane that’s one millimeter above the detector.
The computer can then shift the images and add them again so that every thing in the plane two millimeters above the detector lines up on all 15 images reinforcing everything in that plane while everything out of that plane is misregistered and not seen. By shifting and adding the images we can synthesize all planes through the breast from the original 15 low dose images. Actually, we have now developed more sophisticated computer algorithms that work better than ‘shifting and adding’.
In fact, the technique permits that synthesis of an unlimited number of images and spacing. For various physics reasons we find that one millimeter spacing works well, so we make pages through the breast with one-millimeter separation. However, when I look at a tomosynthesis slice, the resolution within the image is the same resolution as the digital detector.
One tenth of a millimeter is the size of the resolution element in our detector so that’s the size of the detail that we see on each slice. Even though the slices are a millimeter apart, they’re a tenth of a millimeter in resolution so the ability to see sharply and clearly is pretty much the same as the digital detector, but we get the rest of the breast out of the way.
Advantages of tomosynthesis
Tomosynthesis eliminates the forest, if you will, when we are looking for specific trees. For example, a birch tree in a pine forest is pretty hard to see when we are standing outside looking in, but if we could go row by row through the forest, suddenly there’s the row with the birch tree in it and we have no trouble seeing it. That’s what tomosynthesis does for the mammogram.
A real advantage is that radiologists understand what we’re looking at since these are the same structures that we see with conventional X‑ray mammography, but we see them with greater clarity because we’ve gotten rid of what’s in front and what’s in back. So it allows us to see cancers more clearly because they’re not hidden by the forest, or in other words, the whole breast on a conventional mammogram.
The second advantage is we can see the margins of an abnormality more clearly because the rest of the breast doesn’t get in the way. We use the margins to decide whether something may or may not be cancer. If something has ill-defined margins or has lines coming out from it that look like a star, which we call spicules, we can see the spicules more clearly and differentiate benign from malignant with greater accuracy.
The other fundamental benefit comes when we look at screening mammograms of women who come in who have no reason to think they have cancer. In our practice we call back about eight women out of 100 to get additional pictures because we see something on the mammogram that we’re concerned about but we’re not sure whether we have to biopsy or not
Two women out of the eight (25 percent) women we call back will have nothing more than normal tissues that superimpose on the conventional mammogram causing a false alarm. With tomosynthesis, we completely eliminate the superimposition so those women will never get called back because there’s nothing superimposed on the tomosynthesis.
Because we see things with greater clarity, we reduce the callback rate. In December, my colleague Richard Moore, who heads our research program, presented our results on callback rates from a nearly 3000 women study. He found that using tomosynthesis reduced the callback rate by 37 percent. So the false alarm rate goes down with tomosynthesis as well as improving our ability to see cancers more clearly.
Future usage of tomosynthesis
We thought about using tomosynthesis in the late 1970s, but had to wait for digital detectors to be developed in the 1990s. We identified the General Electric digital detector as being best suited for a prototype device and we worked out the physics in the mid 1990s. The Massachusetts General Hospital now has a patent for digital breast tomosynthesis. We received two grants from the Army – one that helped us pay General Electric to build the prototype machine, and one to do various pilot studies that allowed us to show that the technology works the way that we had expected.
We’re just finishing up a 3000-woman study funded by the National Cancer Institute. In our preliminary studies we weren’t blinded. We compared the standard mammograms directly to tomosynthesis images of the same volunteers. In other words, when we first collected and looked at the images, we knew what the mammogram showed.
Our recent 3000-woman study is a blinded study where one radiologist reads the conventional screening mammogram and the other radiologist reads the tomosynthesis without knowing the other’s results – the two-dimensional or the standard reader doesn’t know what the tomosynthesis showed and the tomosynthesis reader doesn’t know what the standard mammogram shows.
Then we’re comparing the results. That’s where we have the data on reduced callback rates. With conventional mammography (we used digital mammograms) we called back more women than we would have had they been screened with tomosynthesis.
We’ve screened almost 3500 women in various studies at the Massachusetts General Hospital confirming that the technology works, and we are now looking at the advantages. So far, we’ve been able to prove a decreased callback rate. In reader studies our data shows that cancer detection and differentiating benign from malignant is better with tomosynthesis. Proving an increased cancer detection rate is a much bigger study because, fortunately there aren’t that many cancers in 3,000 women. Consequently, we cannot prove which technique is better with our present study, but all our data suggest that tomosynthesis will improve cancer detection.
I believe that companies developing the devices are very close to getting clinical approval. Several companies are collecting data where they will compare tomosynthesis to conventional, two-dimensional film or digital mammography to provide the FDA with the information needed to get the FDA to approve the use of tomosynthesis in the general population. Those studies should be complete over the next year and I would hope by 2009 tomosynthesis will be available to the general public.
Tomosynthesis is not going to find every single cancer because there are some cancers that hide even in the slices. But from a radiologist’s point of view, it reduces one of the big problems that we have which is the superimposition of normal tissues. So I call it a mammogram only better.
Any radiologist who looks at conventional mammograms can interpret tomosynthesis images. It’s the same information only you see it with greater clarity. Consequently, the ‘learning curve’ is very short and people should be able to interpret these studies pretty quickly without a whole lot of teaching.
We expect from our preliminary data that we’ll find more cancers with tomosynthesis or find them earlier, and we’ll reduce the false positive rate. Usually when you reduce the false-positive rate, you also reduce the ability to find cancers, but with tomosynthesis it looks like we can go in both directions so that we find more cancers and reduce the false positive rate, which is great for women.
We are excited about this new technology, but it is very important that women realize that conventional mammography, whether it’s film screen or digital, has already saved tens of thousands of lives. Mammography is not perfect but studies have clearly shown that the death rate from breast cancer is down in the United States by 25 percent since 1990 and it’s down primarily due to early detection by conventional mammography.
Prior to 1990, the death rate hadn’t changed for 50 years so it proves that early detection through conventional mammography saves lives. We think that tomosynthesis will build on that and that we’ll be able to find additional cancers earlier and save additional lives, but women should not wait for tomosynthesis, but should take advantage of conventional mammography screening now.
Tomosynthesis will be the next big step forward in the battle against breast cancer. It’s not perfection, but I think we’ll reduce the false positive rates, we’ll make fewer women anxious with callbacks and it should increase the detection rate as well.
I strongly support laboratory research to ‘find a cure’, but when you look at all the money that’s gone into therapy, what’s really been saving lives is early detection with mammography. Mammography has improved over the years: digital mammography is modern and tomosynthesis is cutting-edge. Those are saving lives; granted, we should still be investing in better therapy and finding ways to cure breast cancer but we haven’t figured that out yet.
What we have figured out is how to find many cancers earlier. I would strongly urge that women take advantage of that now while we’re working hard to find better ways to find more cancers earlier and while we’re hoping that someone in the laboratory is going to come up with a cure. So I would urge women to get regular mammograms now and look forward to tomosynthesis in the near future.
Dr. Daniel B. Kopans is considered a world expert in breast cancer detection and diagnosis. He is often noted for his focus and extensive knowledge on mammography, breast imaging systems, and related technologies. Kopans is currently the senior radiologist and founder of the Breast Imaging Division at Massachusetts General Hospital (MGH), and a professor of radiology at Harvard Medical School.