Are you ready for your 3D mammogram?

Via my friend and fellow cancer-warrior Francesco Fondi of Wired (Italy), news that Fujifilm in Japan is launching what it calls "Real 3D Mammography," a medical imaging system that enables technicians to view mammographic images in a kind of 3D. The idea is to see and interpret the detail of internal anatomical breast structures more clearly than is currently possible with a 2D image.

The new system costs about $181K, and is designed to work with "Amulet f," Fujifilm's digital X-ray equipment for breast cancer screening (sold separately). I hereby volunteer to be a test hamster for this thing some day, even though I realize the radiation payload is a little higher with this than with current mammography. But wow, I'd love to see this level of detail about what is going on inside my body right now, as I go through chemotherapy.

Takuya Otani, reporting in Nikkei Electronics & Digital Health Online:

Because the three-dimensional structures of breast tissues can be checked all at the same time, it is possible to determine if a tumor mass is in contact with a mammary gland as well as to measure the depth of microcalcification, Fujifilm said.

When a picture of a breast is taken with the Amulet f, it takes two images from different angles. Then, by displaying the two images on a special LCD monitor and using polarized glasses, it becomes possible to see a 3D image.

The special monitor is manufactured by combining two LCD monitors and a part called "half mirror." (...)When the Amulet f is used to take pictures, a patient is exposed to X-ray radiation twice. But, with Fujifilm's own methods of taking and processing images, the total amount of X-ray increases by only 30-50%, compared with a normal mammography, the company said.

Read more. At the Fujifilm website, an explainer with images including the one shown above.

Two images are displayed on the two high-definition monitors. By wearing the polarized 3D glasses, a 3D image can be viewed through the half mirror.


  1. Umm..not to down play the coolness that is going on here, but why are we displaying it as a “faux” 3D image?  (Faux is in we are using a 2D monitor to create a 3D effect.)  If there are two different angle X-ray images taken why not interpolate the two into an actual 3d mesh/image that could be rotated and viewed on a standard 2D screen?  

    1. You still need to be able to visualize each image individually.  There is usually loss of fine detail when processing the source images, and mammographers are looking for calcifications that are 0.2 mm in diameter.  That kind of detail will get lost on a 3D image.  But true, you could still fuse them and just make the original images available separately.

      In my opinion, tomosynthesis shows more promise than this technology. The limitation of overlapping tissue still exists with this method, just somewhat reduced. Tomosynthesis, but taking multiple slices through the breast, does a much better job of dealing with the problem of overlapping tissue.

      1. Here’s a nice prototype of a dedicated breast CT that creates a full 3D image from 300 X-ray projections. As Scott says, the advantage of a 3D image over 2D in diagnosis is that sometimes the tumor that you’re surying to find is hidden behind some other tissue. On a 3D image you can ‘look around’ it. And nowadays you can get very high resolutions on a CT: there are industrial and scientific sub-micron CTs! 

        1.  How well are lymph nodes (or even smaller breasts) imaged using that hanging-down-through-the-hole-in-the-table method, I wonder?

          1. Small breasts shouldn’t be a problem.  I have never seen that technology in use, but it is the same positioning as for breast MRI and small breasts are imaged fine.  The axillary (armpit) lymph nodes don’t look like they would be imaged on the CT, but there is no effort made to evaluate them unless there is clinical concern.  Often some lower nodes are imaged on mammograms, but that is not intentional.  Ultrasound is the gold standard to evaluate them currently anyway.

    2. Because that’s an ill-posed problem: there are many possible 3D structures that would give identical projection images. You’d need many more than just 2 projections to reconstruct the whole structure. You have a good chance of converging on the correct structure if you add some assumptions, but it’s not guaranteed and as mentioned earlier it is very computationally intensive.

  2. The X-ray tech might be neat, but the monitor setup is a completely ridiculous way to view what is essentially the same as the old 3-d postcards.  In fact, I suspect just flipping back and forth between the two images would give most of the effect.

  3. A huge industry thrives on mammography screening. This obscures rational decisionmaking, not only for the individual woman, but also for health professionals.
    Although screening may spot the cancer, as it did for Xeni, it may also reveal harmless tumors that might never have developed into cancer. It is a double-edged sword.

    [obligatory reference to Woody Allen’s “Everything You Ever Wanted To Know About Sex (But Were Afraid To Ask)”]

    To balance that lowbrow comment: I’m sure the way in which this image works is that the top screen is polarized with one type of light, which is then superimposed over the oppositely polarized light of the bottom screen, so the lenses of the glasses filter out one each to create a 3D movie-style fusion. Cory will HATE it.

  5. Are there non-breast applications? Seems like this would be extremely useful in identifying Multiple Sclerosis plaques. I just recently spent too much time with my doctor trying to correlate spots in my top-view MRI slices to side-view MRI slices.

    Maybe breast cancer is just where the money is to get started, right now.

    1. Were the images a paper copy?  Because when MRIs are read on a workstation, it is VERY easy to correlate spots on 2 views (most software will let you click on the spot on 1 view and show it to you in the other 2 views).

      1.  No, I think the software was trying to help navigate the images, but we still had to try to visually correlate the spots to get a feel for their volume. They were pretty tiny/fuzzy, so I think that complicated things. I kind of wondered how much was hiding between the slices, too.

        1. Ah, yeah, the CD software is terrible.  The reading on the original workstation should give that info.  The sequences that are good for picking up plaques have relatively low resolution for various reasons (the slices are about 5 mm).  So if it is smaller than 5-10 mm, it may be missed.

  6. I would be interested in a blind study to determine whether all this extra money in the form of fancy 3D screens will actually allow technicians to better spot the tumors.

    It’s not actually giving you any more data than if you just had a rotatable “3D” model on your 2D screen. Both methods would make it equally easy or hard to detect a tumor behind other tissue. And in neither case are you getting an actual 3D image where you could move your head to get a different angle on the image.

    But if it turns out it’s good enough that it actually detects more tumors, well then bravo, and let’s have more of it.

  7. I’ve worked with sterotactic -xray tables for mammography.  The summary is fairly accurate.  This is for figuring out “What the hell is that?”  Soft tissue doesn’t have much definition and breasts are very complex (fat/tissue/glands/ligaments).  A 2D image looks like a big splatter.  The trick is to look for tell-tail marks (such as calcification) that look marginally different.  Even then, there are many many false positives.  I’d say they are more of a problem than false positives.

    90% false positive rate is common.  Of the 10% that are a mass, 90% are easily treated.  And the vast majority of breast cancers are slow growing (taking 5 to 10 years to reach the size of a pea).  So many women are unnecessarily frightened by results that turn out to be nothing.  And when missed, can be caught later.  So the detection rate (with good scans done at timely intervals) is pretty good, while the harm and cost from false positives is something that has lots of room for improvement.

    3D gives you a bit more information.  Maybe you can tell if something is two layers stacked in the image giving the impression of a mass (i.e. ligaments, not a mass).  Maybe reduce false positives.  Maybe save a few nights worrying if you have cancer and avoiding a biopsy.

    Note: MRI, CT, and ultrasound all pick up different things.  Each has a high false positive rate, but not on the same structures.  Each has major drawbacks.  All told, 2-D x-ray is still a good screening method because it is cheap enough to be widely deployed.

    Note 2: successive 3D scans never look the same.  Breasts move too much (normally something I approve of, but professionally lots of trouble).  And still look like a big splatter.  You’ll still end up with some doctor using a pen to go “It’s about here.”

    1. My doctor spends five minutes pre-apologizing to me before he sticks his finger in for two seconds. Besides the obvious panoply of anal pleasure references available here, I would like to point out that it’s ickier having his finger in my mouth than my ass.

    2.  If you are prostrate and being buggered, it’s time to find a new doctor.  Unless his office is open on weekends.

Comments are closed.