Improved Visualization with High Resolution Breast CT Images
C Lai*, Y Shen, Y Zhong, C Shaw, UT MD Anderson Cancer Center, Houston, TXSU-D-116-6 Sunday 2:05PM - 3:00PM Room: 116
To investigate the visualization that could be improved with high resolution breast CT (BCT) images acquired with a high resolution BCT scanner.
A high resolution BCT scanner employing a CMOS flat panel detector (2923, Dexela Limited, London, UK) with a pixel size of 0.075 mm and 0.5 mm thick CsI scintillator was used to image mastectomy specimens with IRB approval. The flat panel digital detector was operated with non-binning mode for image acquisition to acquire high resolution projection views. 300 projection views were acquired over 360° for each scan. The radiation dose per scan was comparable to that of a two-view mammography (about 4 mGy). The high resolution projection views were reconstructed with Feldkamp-Davis-Kress (FDK) filtered backprojection algorithm and a ramp filter for image reconstruction to generate high resolution BCT image sets with voxel size of 0.06 mm. Following, the projection views were processed with 2x2 and 4x4 binning to simulate low resolution projection views and then reconstructed to obtain two BCT image sets with voxel sizes of 0.12 and 0.24 mm, respectively. The BCT image sets with all different voxel sizes were displayed on a LCD monitor and then visually compared the findings based on mammograms and the diagnostic reports.
Based on visual comparison, significantly improved visualization for small structures (either calcifications or irregular masses) in the 0.06 mm-voxel BCT images was observed compared to the 0.12 or 0.24 mm-voxel BCT images.
We have successfully demonstrated improved visualization for small structures with the use of the high resolution BCT images, which may have a potential for better detection sensitivity and characterization accuracy, leading to reduced recall rate in screening and reduced biopsy rate in diagnosis of breast cancers.
Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by grants CA139830, CA138502 and CA124585 from the NIH-NCI.