Physics and Development of Breast Tomosynthesis

Screening mammography has undergone many improvements since widespread screening began in the 1980s. These improvements have led to improved image quality, reduced radiation dose, and more accurate examinations. In the early 2000s there was a major shift in imaging technology, with the introduction of digital mammography. Although the transition to digital mammography systems took nearly a decade, there are now very few analog systems still in use. The transition to digital mammography also enabled the use of advanced imaging methods that were not practical with film mammography. Tomosynthesis is one of these advanced imaging methods that is rapidly being adopted for breast cancer screening in the United States. The adoption rate is much faster than that seen for digital mammography because the initial clinical results are far more compelling than those seen with the introduction of digital mammography.

Although digital mammography involves a different method for detecting mammography images, it is a two-dimensional (2D) image and suffers from superimposed tissue that may mask or mimic a breast cancer. Mammography sensitivity and specificity are inversely related to breast density. The structures that appear dense or white on a mammogram are composed of fibrous or glandular tissue. This fibroglandular tissue has nearly identical x-ray attenuation to that of breast cancer, and as a result it may hide breast cancer or superimposed tissue may appear suspicious for cancer. For this reason, as breast density increases, the ability to detect cancer decreases and the number of women recalled for additional testing increases. The limitations of conventional mammography for women with dense breasts have led many states to require women be notified if they have dense breasts, allowing them to discuss with their physicians supplemental screening methods.

Tomosynthesis is a three-dimensional (3D) imaging method that allows visualization of the tissue in a series of images spaced at 1-mm intervals through the breast. Instead of a single-projection image that is used for conventional digital mammography images, a series of 9 to 25 images are taken as the x-ray source moves in an arc above the breast ( Fig. 2.1 ). These images are referred to as projection images and are obtained at 1- to 3-degree increments as the x-ray source moves in an arc above the breast. These images are reconstructed using methods similar to computed tomography (CT) reconstruction into a series of images with only structures in a small range of thickness in focus. In mammography the breast is typically compressed to a thickness of 40 to 70 mm, and this results in 40 to 70 tomosynthesis images for an average size breast.

A clinical tomosynthesis image is shown in Fig. 2.2 and demonstrates the marked reduction in complexity of the background in the reconstructed tomosynthesis images compared with the conventional mammogram. This reduction in superimposed tissue allows the detection of breast cancer that might otherwise be hidden by fibroglandular structures above or below the cancer and better differentiation of normal and abnormal breast structures; this results in a reduction of women recalled for additional imaging who do not have breast cancer.

Tomosynthesis systems are now available for imaging many parts of the body; however, only in breast imaging has widespread adoption occurred. This is likely due to some of the specific needs for breast imaging and the specific strengths of tomosynthesis imaging. Breast imaging requires extremely high spatial resolution, and the images must be obtained at very low dose. In addition, high contrast of small structures, such as calcifications, spiculations, and lesion margins, is critical and requires the use of low-energy x-rays. Tomosynthesis allows the use of low-energy x-rays, low dose, and high resolution needed for breast imaging while also reducing superimposed structures.

Development of Tomosynthesis

The initial work on breast tomosynthesis was performed at Massachusetts General Hospital (MGH), starting in 1995. Our group started with phantom imaging, then specimen imaging, and finally clinical imaging. Three medical physicists, including Bradley Christian, Laura Niklason, and myself, did the initial development work. The initial physics and specimen imaging work was supported by a grant from the Department of Defense Breast Cancer Research Program. Although our work was the first on breast tomosynthesis, there had been a lot of previous research on tomosynthesis for imaging other body parts.

The key technological breakthrough needed for the implementation of tomosynthesis was the development of a flat panel digital detector. A group at General Electric Corporate Research and Development provided the first detector that made breast tomosynthesis possible. This detector had the high resolution needed for mammography and the rapid image readout needed for tomosynthesis. The team at General Electric included Henri Rougeot, Beale Opshal-Ong, Cynthia Landberg, Donald Castleberry, Jeffrey Eberhard, and many others. The detector development was supported by a grant from Defense Advanced Research Projects Agency.

The clinical evaluation of tomosynthesis was led by Daniel Kopans at MGH. The first clinical unit was constructed by General Electric to support this research. The research was again funded by the Department of Defense Breast Cancer Research Program, and their support provided the funding to transition breast tomosynthesis into a clinical reality.

After the initial clinical work, there was a lag in development of tomosynthesis. General Electric, which had supported the early research, decided against commercialization at that time. In 2005 Hologic, Inc., decided to develop a clinical tomosynthesis unit, and I joined Hologic at that time. Hologic presented data for the US Food and Drug Administration (FDA) approval to a panel in 2010 and received approval in early 2011. Elizabeth Rafferty, MD, was the principal investigator for the FDA studies and presented the data at the FDA panel meeting.

After just a few short years, tomosynthesis is now rapidly replacing conventional mammography for breast cancer screening. Looking back, there are several people who deserve special recognition for proving the clinical superiority of tomosynthesis and driving the commercial development. First, Elizabeth Rafferty, MD, from MGH led all of the Hologic reader studies investigating the best methods for using tomosynthesis and provided the clinical leadership for each study. In addition to publishing many of these results, she led the training effort and has trained many of the radiologists currently using tomosynthesis. Her training of the early adopters led to the superior clinical results from a wide range of practices, many of which have resulted in publications. Next, Per Skaane from Oslo, Norway, took on the challenge of performing the first large-scale, prospective screening trial of tomosynthesis. His team imaged 25,000 women over 2 years and doubled their workload by having each case read by four different radiologists. This research demonstrated the clinical efficacy of tomosynthesis for both the single reader method used in North America and the double reader method used in Europe. He also reported on the clinical performance of synthetic mammograms combined with tomosynthesis. David Gur from Magee Women’s Hospital and the University of Pittsburgh Medical Center was a leader in evaluating tomosynthesis from the beginning and, along with Margarita Zuley and Jules Sumkin, provided much of the basic research demonstrating the clinical utility of tomosynthesis. David Gur was instrumental in working with Dr. Skaane and myself on study design for the Oslo study and brought together the group of 13 institutions in the United States that published the landmark study in the Journal of the American Medical Association evaluating tomosynthesis screening performance in 450,000 women.

Finally, on the commercial side, Jay Stein was the key person behind several decisions that made tomosynthesis a commercial success. First, the decision to add tomosynthesis to conventional mammography made clinical adoption much faster. Next, Jay Stein worked with Chris Ruth to develop the first synthesized mammograms. This development has been critical for widespread commercial adoption.

System Design

The design of tomosynthesis systems varies greatly among different manufacturers. Differences in design may have significant impact on clinical performance. In this section the key design parameters and their impact on image quality will be discussed.

Acquisition Angle

Tomosynthesis projection images are acquired over an angular range of 15 to 50 degrees. This angular range affects image quality in several ways. Wider angle acquisition reduces the amount of superimposed tissue that may be present in the reconstructed tomosynthesis images. If the tomosynthesis acquisition angle was increased to slightly greater than 180 degrees, a true CT image could be obtained, resulting in complete removal of superimposed tissue. Thus tomosynthesis images acquired at 50 degrees may have reduced superimposed tissue compared with those obtained at 15 degrees. However, there are benefits associated with narrow acquisition angles, including better depiction of calcifications.

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here