Breast tomosynthesis is a promising new technique that produces a 3-dimensional image set. Initial clinical results indicate that the technique can overcome the major problem with conventional mammography. A conventional mammography, because it produces a 2-D image, suffers from the problem of overlapping tissue either obscuring a cancer (false negative) or mimicking a cancer (false positive). While the technique demonstrates much promise, it has not yet been properly optimized. Our long-term goal is to optimize fully breast tomosynthesis. As part of that optimization, we believe that computer-aided diagnosis will play an important role in breast tomosynthesis, because each patient will have up to 80 images of each breast. Our goal for this proposal is to develop computer-aided detection (CADe) schemes for breast tomosynthesis. The specific aims are (1-3 are for R21 and 4-7 for R33): 1. To develop a database of clinical breast tomosynthesis cases; 2. To develop CADe schemes (clustered calcifications and masses) using projection images; 3. To develop a comprehensive model of breast tomosynthesis system to produce synthetic tomosynthesis images. 4. To study the effect of image acquisition on the performance of CADe schemes; 5. Use CADe results on projection images to improve reconstructed image set; 6. Develop CADe schemes using reconstructed image set; 7. Pilot observer study comparing tomosynthesis with and without CAD. If we are successful, tomosynthesis will have greater clinical utility and, therefore, acceptability. This should improve the detection of breast cancer - improved sensitivity and specificity - resulting in a reduction in breast cancer mortality and morbidity.

R33CA109963

2006-09-22

2009-08-31