Heang-Ping Chan PhD, Presenter: Nothing to Disclose

Jun Wei PhD, Abstract Co-Author: Nothing to Disclose

Yiheng Zhang PhD, Abstract Co-Author: Nothing to Disclose

Mark Alan Helvie MD, Abstract Co-Author: Institutional grant, General Electric Company

Lubomir M. Hadjiiski PhD, Abstract Co-Author: Nothing to Disclose

Berkman Sahiner PhD, Abstract Co-Author: Nothing to Disclose

To study the effect of the number of projection views (PVs) on mass detection in DBT.

DBT mammograms of 50 breasts containing biopsy-proven masses (14 malignant, 36 benign) were obtained with IRB approval and informed consent using a GE GEN2 DBT system which acquires 21 PVs over a 60º arc in 3º increments. The total dose of the 21 PVs is about 1.5 times of that of a conventional mammogram. The DBTs were reconstructed with SART in two conditions: 21 PVs in 3º increments and 11 PVs in 6º increments. The latter therefore used a subset of the PVs and about half the dose of the former. We compared the mass detection accuracy in these two DBT sets using three approaches. In the first approach (3D), mass candidate identification and feature analysis were performed in the 3D DBT volume. A mass likelihood score (MKS) was estimated for each mass candidate using a linear discriminant analysis (LDA) classifier. A free response receiver operating characteristic (FROC) curve was generated by varying the decision threshold on the MKS. In the second approach (2D), mass candidate detection and feature analysis were performed on the individual PVs. An MKS was estimated for each mass candidate using an LDA classifier trained for the 2D features. The mass likelihood images derived from the PVs were then backprojected to the breast volume to identify high MKS locations. An FROC curve was again generated using the 3D MKS merged by backprojection. In the third approach (3D+2D), the two MKSs from the 3D and 2D approaches at the corresponding 3D location were averaged. Leave-one-out resampling was used to train the LDA classifiers in each approach.

With the 21-PV DBT set, the 3D, 2D and 3D+2D approaches achieved a test sensitivity of 85% at an average of 2.73, 4.04, and 1.56 FPs/volume, respectively. With the 11-PV DBT set, the FP rates were 3.02, 4.10, and 2.51, respectively, at the same sensitivity. The differences in the FROC curves between the 21-PV and 11-PV DBT sets were statistically significant (p<0.05) for all three approaches.

The number of PVs and dose affect image quality and mass detection. Further studies are needed for other conditions of dose and PV distributions.

DBT has the potential to improve breast cancer detection. Studies of factors affecting image quality and lesion detection will be important for optimization of the design of DBT systems.

Chan, H, Wei, J, Zhang, Y, Helvie, M, Hadjiiski, L, Sahiner, B, Digital Breast Tomosynthesis (DBT) Mammography: Effect of Number of Projection Views on Computerized Mass Detection Using 2D and 3D Approaches.  Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL. http://archive.rsna.org/2008/6021064.html