James T. Dobbins III, PhD

Digital tomosynthesis is a method of producing section images using a digital detector and conventional x-ray system with a moving x-ray tube. These section images have superb spatial resolution in the plane of the image, but less resolution in the depth direction. Despite this limitation, tomosynthesis holds great promise for improving detection of certain types of lesions over conventional chest radiography, and at a lower cost and lower x-ray dose than CT.

Tomosynthesis is a contemporary version of traditional geometric tomography, a technique that has been known since the 1930s. Conventional geometric tomography, although widely popular several decades ago, is seldom practiced today outside of excretory urography. The reason conventional tomography waned in popularity is because it could acquire only a single section image at a time, thus making it time-consuming and dose-intensive if more than one section were desired. Furthermore, conventional geometric tomography was plagued by the superposition of blurry artifacts from structures lying outside the plane of interest.

Digital tomosynthesis has eliminated virtually all of the problems associated with conventional tomography. Tomosynthesis allows the retrospective production of an arbitrary number of section images from a single pass of the x-ray tube, allowing the full three-dimensional volume of the patient to be reconstructed, albeit with somewhat less depth resolution than CT. A number of deblurring algorithms have been developed to eliminate residual blur artifacts from structures outside each plane of interest, yielding tomosynthesis section images that render a well-defined section of anatomy. The strong current interest in tomosynthesis has been made possible by flat-panel imaging receptors that have come to market in the last 5 to 10 years; they have proved to be an ideal platform for implementing tomosynthesis.


Figure 1. Tomosynthesis image through the mid lung, depicting excellent rendition of the vasculature, absence of overlying ribs, and a clearly visualized pulmonary nodule. Courtesy of Duke University Medical Center. (Click the image for a larger version.)

A series of projection image “snapshots” is acquired as the x-ray tube is moved along a path by a motorized tube assembly. In chest imaging, this path of motion is typically linear and vertical, and controlled by an overhead tube crane. The projection images show the anatomy from different orientations due to parallax, and can be shifted and added together to render structures in one plane in sharp focus while objects in other planes are blurred out in the summation image. Deblurring algorithms are then applied to eliminate the out-of-plane blur; these methods use a variety of approaches including linear algebra, unsharp masking, maximum likelihood expectation maximization, or iterative restoration. Details on these various approaches are outlined in a recent review article. 1 A typical set of acquisition parameters for chest imaging is 71 projection images, 20-degree total tube movement, and 5 mm spacing of reconstructed sections. The resulting section images can be viewed dynamically as a stack of reconstructed images.

There have been many applications of tomosynthesis reported in the literature, including dental, orthopedic, chest, breast, and angiographic imaging. Currently, the two areas receiving the most research and commercial interest are applications in breast and chest imaging. In the chest, tomosynthesis is being investigated in an NIH-funded trial at Duke University Medical Center for its application to improving detection of pulmonary nodules. Pulmonary nodules are often difficult to see, not because of their size, but because of their poor conspicuity relative to a confusing background of overlying anatomy. Tomosynthesis can improve the visibility of pulmonary nodules by producing section images that have removed the ribs and overlying vasculature. Very early results at Duke have indicated that substantial improvement in visibility of pulmonary nodules in tomosynthesis images is possible compared with conventional PA radiographs. These results must be confirmed in the larger ongoing NIH trial.


Because tomosynthesis can be implemented using a standard digital x-ray system, it has only modest incremental expense over a dedicated chest imaging room. It requires a high-quality digital detector with rapid readout, such as can be provided with current flat-panel detectors. These flat-panel rooms are considerably more expensive than traditional screen-film chest rooms, but are becoming more commonplace as older screen-film rooms are replaced; these flat-panel detectors are rapidly becoming the gold standard in conventional chest radiography. The only notable requirement of a room to be tomosynthesis-ready is the addition of a motorized tube mover. Additional software is also needed to do the tomosynthesis reconstructions.

There is virtually no additional technologist time required for tomosynthesis, because the projection images required for tomosynthesis reconstruction can be acquired without repositioning of the patient immediately after acquisition of the standard digital PA chest radiograph. The entire acquisition sequence is completed in about 10 seconds, well within a single breath hold for most patients. Thus, the impact on patient throughput is expected to be minimal.

Chest tomosynthesis should become commercially available within the next year. Although the final assessment of its role in chest imaging awaits completion of an NIH trial, early indications suggest that it will become a useful adjunct to conventional chest radiography for the detection of lesions such as pulmonary nodules.

James T. Dobbins III, PhD, is associate professor of radiology and biomedical engineering, and director, Medical Physics Graduate Program, Duke University, Durham, NC.


  1. Dobbins JT, Godfrey DJ. Digital x-ray tomosynthesis: current state of the art and clinical potential. Phys Med Biol. 2003;48:R65-R106.