DIGITALBREAST TOMOSYNTHESIS (1)

2021-09-14 10:59

INTRODUCTION

Digital Breast Tomosynthesis (DBT) is a major advance in mammography screening. The introduction of computers in the 1970’s and the development of digital imaging in the 1980’s and 1990’s opened major opportunities for improved radiological evaluation. In the U.S. the development of Full Field Digital Mammography (FFDM) was delayed because the Food and Drug Administration (FDA) required the manufacturers to undertake the time consuming and expensive Pre-Market Approval (PMA) process instead of the simplified 510K path that sufficed for all other digital imaging (bone, chest, etc.).

Despite a lack of support, we began the development of Digital Breast Tomosynthesis (DBT) at the Massachusetts General Hospital in 1992, but it wasn’t until 2011 that the FDA issued the first PMA for the technology and it became available for clinical care. Since then, because of its major advantages,DBT has progressively been replacing 2D FFDM in the U.S.

THE DEVELOPMENT OF DIGITAL BREAST TOMOSYNTHESIS (DBT)

In 1978 I became Head of the “Xeroradiography Division” at the Massachusetts General Hospital. Since, in addition to x-ray, we were starting to use ultrasound and other imaging tests to evaluate the breast I soon changed the name of my Division to “Breast Imaging” which was ultimately adopted as the name around the World for our new subspecialty.

When we ed concern, but was not palpable, we would place needles in the breast (“needle localization”), guided by the images, to direct surgeons to the area so that the lesion could be excised to determine whether or not it was benign ormalignant (I developed the hookwire localization guide soon after). The surge on would send the excised tissue from the Operating Room for us to x-ray the specimen to confirm that the targeted lesion had been removed. I was struck by how much clearer lesions were seen on the specimen radiographs than they could be seen on the mammogram(Figure 1).

Figure 1

Lesion in the Breast Specimen Radiograph

Of course, part of this was that the specimen was closer to the detector so that “geometric unsharpness” was reduced, but it was also obvious that the tissues of the breast in front and in back of the lesion on the mammogram were hiding lesions. This was well known to physicists as“structure noise”. By removing the tissues that were superimposed on the lesion, its shape and margins were seen with much greater clarity than on the 2D mammogram.

In the 1970’s we were employing linear, and polycycloidal tomography for evaluating the lungs as well as the kidneys and other organs. This involved moving the x-ray tube in one direction while moving the detector (film/screen) in the other, while the exposure was being made. By having the plane of interest at the fulcrum of the motion, the details of structures within the plane of interest were seen with greater clarity while the structures in front of and in back of the plane of interest were blurred by the motions of the tube and detector. By changing the fulcrum and repeating the exposure, “slices” or planes could be obtained “up or down” through the organ allowing the structures to be visualized with greater clarity by blurring the tissues in front and in back.

However, in the 1970’s Bailar had raised the concern that the radiation from mammography might cause more cancers that would be cured, and this was used to, not only, stimulate the development of lower dose mammograms (screen/film replaced Xerograms), but radiation risk was also used to argue against screening women in their forties. Fortunately, we now know that radiation risk for the breast drops rapidly with increasing age and that there is no direct evidence of risk for women ages 40 and over, and even the extrapolated risk is much lower than even the smallest benefit ([1],[2],[3],[4],[5],[6]), but at the time it would not have been possible to advance a higher dose x-ray technique for imaging the breast. Consequently, it was impossible to even suggest using standard tomography for breast evaluation since eachslice or plane required a full exposure and it would take multiple full exposure images to “slice” through an entire breast raising the dose sign ificantly.

Fortuitously, I came across a paper by Milleret al that described a technique that could provide “slices” through the breast with no increase in dose ([7]). “Tomosynthesis” seemed to provide the opportunity that I was seeking.

PARALLAX IS THE FUNDAMENTAL PRINCIPLE BEHIND TOMOSYNTHESIS

In its simplest form tomosynthesis relieson parallax. When an x-ray is obtained of an object that lies above the detector and then the x-ray tube is moved to image the object from a different angle, the object’s projection (shadow) on the detector will move in relationship to the movement of the x-ray source and the height of the object above the detector. The “shadows” of structures further from the detector will appear to move further on the detector than those that are closer to the detector.

We applied this to the breast by using standard mammograms taken from different angles. Figure 2a, b, c shows schematically how this works. The x-ray tube is moved through an arc over the breast and three images are taken each from a different angle. Note that the “tissue” shadow moves over a longer distance on the detector than the cancer which is closer to the detector.

Figure 2a Figure 2b Figure 2c

By taking several images from differing angles a computer can be used to align the images so that the structures in adesired plane above the detector are perfectly registered and their “shadows” will reinforce one another by the number of projections. At the same time the images of structures out of the plane of interest will be misregistered and not reinforced, and they will fade into the background.

Figure 3

The computer aligns the images so that the plane of interest is reinforced (in this case 3 times) and clearly seen, while the out-of-plane information is misregistered and fades into the background.

In the first series (Figure 3a) the images are aligned so that the plane containing “tissue” is registered on all 3 images and reinforced while the “cancer” is misregistered and fades into the background.

Figure 3a

The plane containing the triangular tissues is in sharp detail while the cancer is out of plane and fades into the background.

The computer then shifts the images and adds them again and this time (Figure 3b) the cancer is in shaper detail and the “tissue” fades into the background.

Figure 3b

Tomosynthesis differs from tomography described earlier because each of the tomosynthesis projection images only requires a fraction of the dose of a full exposure FFDM. Any number of “slices” or planes can be madethrough the breast from the handful of projection images.

The advantage of tomosynthesis is that the planes can have the full resolution of the projection images. The detail in the x and y directions is only limited by the resolution of the stationary system. Some have called DBT “3D mammography”, but this is not accurate. The “voxel” is not cubic. It is diamond shaped, depending on the angle through which the x-ray source is moved to obtain the projection images. The smaller the angle through which the projection images are obtained the more uncertainty in the “z”direction so that the narrower the angle of acquisition the “thicker” the slice. The only way to get true 3-Dimensions would be for the source to go beyond a 180 degree arc which would not work for the breast in a mammographic position and begins to parallel computed tomography.

NOT ALL DBT SYSTEMS ARE THE SAME

As with any x-ray system, DBT systems varyin their design and the quality of their images.

DETECTORS

There are two main, commercially available detectors. General Electric uses aproprietary detector on which cesium iodide crystals are “grown” in columns on the detector. The x-rays are converted to light in the crystals and the light is then converted to an electrical signal.

The other main detector configuration use samorphous selenium. This is very similar to the Xeroradiography system that was used years ago to perform mammography. On its surface the detector plate has auniform electrical charge. The x-ray exposure causes a discharge of the plate that is proportionate to the amount of radiation hitting each pixel. However,instead of blowing “toner” on the charged plate (as in the Xeroradiography system), the remaining charge is directly read electronically from the detector. The x-ray signal is converted directly to an electronic signal.

The detectors vary in the size of their pixels so that their spatial resolutions vary.

ANGLE OF COLLECTION

Each of the commercial systems uses a different angle through which the x-ray tube (source) travels. The x-ray tube does not have to make a majorarc above the breast to create the tomosynthesis images. It appears that just a 15 degree arc provides excellent images. In theory, however, the larger the arc the finer the “z” resolution. There are theoretical and experimental suggestions as to the optimal angle of collection and the number of projection images collected, but since no one has done a large clinical study directly comparing the major systems, it is unclear what is the best angle of acquisition and no clear understanding as to the optimal number of collection images.

HOW BEST TO OBTAIN THE PROJECTION IMAGES

Richard Moore, who was Head of Breast Imaging Research at the MGH while we developed DBT, first proposed using multiple x-ray tubes in an array above the breast and fired in sequence to generate the projection images. The advantage is that, since the tubes are fixed and don’t move, there is no chance for motion unsharpness and the speed of image acquisition is only related to the speed of reading out the detector between images. It is my understanding that this approach is,once again, under consideration.

CONTINUOUS MOTION

Some manufacturers have decided to move the x-ray tube continuously through the arc while intermittently firing the beam ateach point on the arc. This is probably the most efficient method. The only issue is if the motion of the tube during the exposure introduces “blur” into the projection images. If the exposure is short this does not appear to be a problem.

STEP AND SHOOT

The other method that is being used has the tube move, stop, “shoot”, move, stop, shoot, etc. Stopping the tube for each exposure eliminates motion during image acquisition, however, there are mechanical issues involved in moving a heavy x-ray tube stopping it and moving it again.

HOW MANY PROJECTION IMAGES ARE NEEDED?

It is not clear as to how many projection images are needed for optimal DBT imaging. Each of the companies has their own approach, but no one has compared the systems imaging the same breast, so the differences are only theoretical.

PROCESSING THE IMAGES

“Shift and Add” was the original method for processing DBT images. Some manufacturers have adopted the “filtered back projection” approach. Tao Wu PhD developed the Iterative Maximum Likelihood Technique ([8])which has major advantages over other techniques ([9]).

REVIEWING DBT IMAGES

The number of planes that are reconstructed for each set of projection images is related to the thickness of the breast. It certainly takes more time tor eview all of the planes for each patient than to review FFDM 2D images, but as discussed later, since DBT reduces the recall rate, the time saved can be accounted toward review of the DBT screening studies.

DBT IS A SCREENING TEST

We developed DBT as an improvement in screening to replace (actually add to) 2D mammography. DBT detects more early cancers and reduces the recall rate, but it can only do this when it is used for screening. It has limited value as a “diagnostic” test,and early analysts who used it only to evaluate women recalled for additional evaluation, did not appreciate its actual importance. As discussed in “Part 2” DBT will replace standard 2D mammography as a screening test for routine screening for all women.

REFERENCES

1.Feig SA:Hypothetical breast cancer risk from mammography:A reassuring assessment.Breast 5:2-6,1980.

2.Mettler FA, UptonAC, Kelsey CA,Rosenberg RD, Linver MN. Benefits versus Risks from Mammography: A Critical Assessment. Cancer1996;77:903-909.

3.Feig SA, Hendrick RE. Radiation risk from screening mammography of women aged 40-49 years.J Natl Cancer Inst Monogr.1997;(22):119-24. Review.

4.Hendrick RE.Radiation doses and cancer risks from breast imaging studies.Radiology.2010 Oct;257(1):246-53.

5.Yaffe MJ, Mainprize JG. Risk of radiation-induced breast cancer from mammographic screening.Radiology.2011 Jan;258(1):98-105.doi: 10.1148/radiol.10100655. Epub 2010 Nov16.Erratum in: Radiology.2012 Jul;264(1):306.

6.Miglioretti DL,Lange J, van den Broek JJ, Lee CI, van Ravesteyn NT, Ritley D, Kerlikowske K,Fenton JJ, Melnikow J, de Koning HJ, Hubbard RA. Radiation-Induced Breast Cancer Incidence and Mortality From Digital Mammography Screening: A Modeling Study.Ann Intern Med.2016 Feb 16;164(4):205-14.

7.Miller ER, McCurry EM, Hruska B. entitled “An infinite number of laminograms from a finite number of radiographs.Radiology 1971; 98:249-255

8.Wu T,Stewart A, Stanton M, McCauley T, Phillips W, Kopans DB, Moore RH, Eberhard JW,Opsahl-Ong B, Niklason L, Williams MB.Tomographic mammography using a limited number of low-dose cone-beam projection images.Med Phys.2003;30:365-80 which won The Sylvia Sorkin Greenfield Award for two of the best papers (other than Radiation Dosimetry) published in Medical Physics for 2003

9.Wu T, Moore RH, Rafferty EA, Kopans DB. A comparison of reconstructional gorithms for breast tomosynthesis.Med Phys.2004 Sep;31(9):2636-47

Previous Posts： DIGITAL BREAST TOMOSYNTHESIS (2)

：Next Chapter CREATINGTHE MYTH OF "OVERDIAGNOSIS" (2)

DIGITALBREAST TOMOSYNTHESIS (1)

DART Imaging

Beijing

Guangdong

Beijing DaYing

Zhongshan DaYing

DART Imaging
Wechat

DIGITALBREAST TOMOSYNTHESIS (1)

DART Imaging

Beijing

Guangdong

Beijing DaYing

Zhongshan DaYing

DART ImagingWechat

DART Imaging
Wechat