Date(s) - 10/29/2012
9:30 am

Stephanie Leon, BME-Medical Physics PhD Candidate

Abstract:Mammography, a successful screening mechanism, still has room for improvement, particularly in false negative exams. Image quality improvements could be made by reducing the presence of scattered radiation through the image processing technique evaluated here. The study aim was to develop a point spread function (PSF), then use a spatially-variant Wiener filter to remove the scatter PSF from the image. Selection of the PSF used for scatter correction was performed on a pixel-by-pixel basis to account for apparent tissue thickness over each pixel.The PSF was developed from phantom measurements using Hologic Selenia mammography units and dependent on target/filter combination, tube potential, tissue thickness, and grid presence. The parameters used to describe the PSF, scatter fraction (SF) and mean radial extent (MRE) were obtained with physical measurements and verified with Monte Carlo simulations. The SF’s measured ranged from 0.05-0.17 with grid and from 0.25-0.52 without grid. The MRE’s ranged from 3.2-15 mm with grid and from 19-47 mm without grid. Differences in SF between physical and Monte Carlo measurements were -0.04 to +0.07. Monte Carlo MRE and physical data agreed with goodness-of-fit values of 0.96-1.00 with grid and 0.65-0.86 without grid.The image noise decreased after scatter correction approximately 60-70%. Correlation of CNR with MRE was statistically significant before scatter correction (r =  0.67 with grid and -0.89 without grid), but not significant after scatter correction (r = -0.24 with grid and -0.37 without grid). Spatial resolution appeared to have improved, as assessed qualitatively. This method surpassed simpler forms of processing including simple deconvolution and histogram manipulation.Fifty clinical images representative of parameters common to screening mammography were collected from Mo-target Selenia units and scatter-corrected. Three radiologists assessed their preference for the original or scatter-corrected images in eight categories related to image quality and anatomical features visibility. Scatter-corrected images were significantly preferred in all categories, with p ranging from 0.01 to nearly zero. These results indicate the potential to improve clinical image quality using scatter correction. Possibilities for future research include clinical imaging without a grid and applications to magnification imaging and digital breast tomosynthesis.