Computer Aided Detection of Breast Masses in Digital Tomosynthesis
Abstract
The objective of this study is to investigate if digital tomosynthesis can fundamentally improve sensitivity of detecting breast masses compared to conventional mammography. Overlapping dense tissue in mammography is one of the most common causes for unnecessary callbacks as well as missed cancers. By removing such overlapping tissue, breast tomosynthesis can obviate unnecessary callbacks as well as missed cancers. The goal is to provide 3D information at high resolution, comparable dose to mammography, and with lower cost and hardware requirements compared to other common alternatives such as breast Computed Tomography (CT) or breast Magnetic Resonance (MR). In the first stage of this study we applied 2-D CAD algorithms to individual projection images of the tomosynthesis data set. We also reconstructed pre-processed projection images using filtered back projection algorithm, where suspicious regions were identified using a DoG filter. Lastly, we studied feasibility of implementing Laguerre-Gauss channelized hotelling observers on mammographic ROIs and compared their performance against that of another visual model proposed by Watson.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2006
- Accession Number
- ADA462451
Entities
People
- Joseph Y. Lo
- Swatee Singh
Organizations
- Duke University