Dan Xia BS, Presenter: Nothing to Disclose

In current breast tomosynthesis, image representation with non-isotropic spatial resolution is used for reducing computational time. This can, however, lead to artifacts in iterative reconstruction of breast tomosynthesis images. In the work, we investigate the effect of non-isotropic image representation on the reconstruction accuracy. Based upon the investigation, we devise schemes for reducing artifacts in iterative reconstruction.

In the work, we focus on investigating the effect of non-isotropic image representation on reconstruction accuracy of iterative algorithms. The iterative algorithms under study include the total-variation (TV) based, expectation maximization (EM), and algebraic reconstruction technique (ART) algorithms. Tomosynthesis data are generated at 12 and 20 views over 50 degrees from phantoms, including a breast phantom. We have reconstructed images by using image representations with different degrees of non-isotropic spatial resolution. Specifically, in each image representation, the ratio between the in-plane and longitudinal resolution for an image voxel is selected to be a value less than 1.

We have reconstructed images by use of TV-based, EM, and ART algorithms for image representations with different ratios of in-plane and longitudinal resolution. Our results demonstrate that non-isotropic image representation can lead to significant artifacts in reconstructed images. The appearance and severity of the artifacts depend not only upon the ratio between the in-plane and longitudinal resolution but also upon the iterative algorithms. The TV-based algorithm seems to be less susceptible to the effect than the EM and ART algorithms. Through the selection of algorithm parameters, the artifacts can be reduced.

The non-isotropic image representation can significantly affect reconstruction accuracy obtained with iterative algorithms in breast tomosynthesis.

Breast tomosynthesis has received renewed interest because it can provide 3D information about the breast. This work concerns iterative reconstruction of accurate breast tomosynthesis images.

Xia, D, Image Representation with Non-isotropic Spatial Resolution on Iterative reconstruction Accuracy in Breast Tomosynthesis.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5009568.html