Access to SPIE eBooks is limited to subscribing institutions. Access is not available as part of an individual subscription. However, books can be purchased on SPIE.Org
Chapter 12:
Computerized Mass Detection for Digital Breast Tomosynthesis
Editor(s): Jasjit Singh Suri; Rangaraj M. Rangayyan
Author(s): Reiser, Ingrid; Nishikawa, Robert
Published: 2006
DOI: 10.1117/3.651880.ch12
Digital breast tomosynthesis (DBT) is an emerging technique for 3D breast imaging. The main advantage of DBT is that lesion conspicuity is improved because overlaying tissue structures are resolved. In DBT, a finite number of tomographic projection views are acquired over a limited range of x-ray tube angles. A volume image is then reconstructed from the sequence of projections. Conceptually the most simple reconstruction algorithm is "€œshift and add,"€ where projections are shifted to bring one plane into focus. The amount of shift to be added depends on the depth of the in-focus plane. Structures located at this depth are brought into focus, while structures located at other depths are blurred. As a result, in depth resolution is lower than in-plane resolution, which depends on object size. Computer-aided detection (CADe) is being used clinically in mammography screening. The benefit of CADe supporting the radiologist's decision making has been proven in laboratory studies. Therefore, computerized lesion-detection algorithms for DBT are being developed. Because DBT is an x-ray imaging technique, there are similarities with conventional projection mammography. The appearance of the reconstructed breast slices is similar to that of a conventional mammogram. As a result, algorithms developed for computerized lesion detection in projection mammograms can be used in DBT, or can serve as guidelines. However, the resolution properties of DBT present a challenge to CADe. These authors' experience agrees with others in that algorithms from projection mammography may not be effective when applied to the reconstructed breast slices. Other investigators have explored computerized detection methods for mass detection. For computerized lesion-detection, both representations of the DBT image data can be exploited, the reconstructed volume image as well as the sequence of projections. In the following, the DBT prototype unit that was used for patient image acquisition is described. Then, two approaches to CADe for DBT is presented. The first is based on the reconstructed image volume. The second is based on analyzing the sequence of projection images. DBT is a new technique, with probably fewer than five machines being used clinically worldwide. In collaboration with Drs. Daniel Kopans and Elizabeth Rafferty, radiologists at Massachusetts General Hospital, the authors have used an initial database of 40 images. This database will be expanded as more patients are imaged with DBT. The methods presented below, then, are still preliminary because they have not been properly optimized nor have they been properly validated and evaluated. This will be done when a larger database is acquired. The small database allows examination of different approaches and techniques to provide a base on which CADe schemes for DBT can be developed.
Online access to SPIE eBooks is limited to subscribing institutions.

Back to Top