Yao Lu PhD, Presenter: Nothing to Disclose

Heang-Ping Chan PhD, Abstract Co-Author: Nothing to Disclose

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

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

Ravi Kumar Samala PhD, Abstract Co-Author: Nothing to Disclose

Image quality is an important factor that will affect breast cancer detection in digital breast tomosynthesis (DBT). The high-attenuation metal clips embedded in the breast marking a previous biopsy site cause errors in the estimation of attenuation along the ray paths intersecting the clips during reconstruction, which result in interplane and inplane metal artifacts (MAs). Because of the small number of projection views (PVs) acquired in a limited angular range, the voxel value errors in the artifact region cannot be compensated for. This causes stronger MAs for DBT than those for CT reconstruction. We developed a new MA reduction (MAR) method to improve image quality.

Our MAR method uses iterative detection and segmentation to automatically generate a clip location map for each PV. Correlation among different PVs is used to reduce false positive detections. Iterative diffusion-based inpainting is designed to replace the labeled clip pixels with estimated tissue intensity from the neighboring regions in each PV. The inpainted PVs are then used for DBT reconstruction. A voting technique is used to determine the location and shape of the clips and label them in the reconstructed volume. The MAR method does not depend on specific reconstruction techniques. With IRB approval and informed consent, DBT of human subjects was acquired with a GE prototype system(60º arc, 21 PVs, 3º increments). 20 DBT views from 10 breasts of various densities with clips were reconstructed with and without MAR. Five breasts had multiple large clips from lumpectomy, two of which and five other breasts had microclips from core biopsy. The improvement in MAs was visually assessed.

The clip detection rate in the PVs was 100% with no false positives. The interplane and inplane MAs were reduced to a level that was not visually apparent in the reconstructed slices regardless of the size and number of clips in the breast. The visibility of microcalcifications and breast tissues along the ray paths of the clips was improved.

The inpainting-based MAR method reduced the MAs while preserving the structured background and microcalcifications. The visibility of breast lesions obscured by the MAs was improved.

DBT has strong potential to improve breast cancer detection. Reducing the MAs in DBT can improve detection and assessment of subtle breast lesions, especially recurrence near the biopsy site.

Lu, Y, Chan, H, Wei, J, Hadjiiski, L, Samala, R, Improving Image Quality for Digital Breast Tomosynthesis: Automatic Detection and Inpainting Method for Metal Artifact Reduction.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13017471.html