Optimizing digital breast tomosynthesis

Mammography is currently the most widely used method for detecting breast cancer. The main limitation of mammography is that a mammogram is a 2D image of a 3D breast. As a result, the normal anatomy of the breast that is above or below a cancer can obscure the cancer in the image. There are two competing approaches being developed for producing 3-D images of the breast: computed tomography (CT) and digital tomosynthesis (tomo). Both techniques produce a stack of 2D slices of the 3D breast. CT is capable of producing true 3D images of the breast, with high spatial resolution in all three directions. Tomo, on the other hand, has some potential advantages over CT in terms of cost and imaging time. However, tomo images have very poor spatial resolution in one direction and, as a result, there is “residual” in the image from structures above and below the slice of interest. This has the potential for obscuring a cancer. The amount of residual in a slice can be controlled by the way the tomo image is acquired.

The goal of this project is to measure quantitatively the amount of residual structure in the tomo images for different acquisition methods. The approach is to measure the slope of the power spectrum of the tomo image for different acquisition conditions. The power spectrum is measure by taking the Fourier transform of the image after correcting the image for any biases that may exist. Then using images of test objects, the visibility of structures in the slices can be correlated with the amount of residual structure to find an acceptable level.

The student will learn about mammography, image reconstruction and digital tomosynthesis. In addition, the student will develop new techniques for analyzing medical images. The student should be familiar with Fourier transforms and be able to program in at least of one: C, C++, Matlab or IDL.

The project will be conducted in the Department of Radiology at the University of Chicago under the direction of Robert Nishikawa, Ph.D., a world leading researcher in Computer Aided Detection of Cancer.

This is part of a collaboration with the Medical Imaging Research group at KTH headed by Mats Danielsson.