Various reconstruction algorithms for Tomosnythesis WEIHUA ZHOU, LINLIN CONG 11/07/2011DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING SOUTHERN ILLINOIS UNIVERSITY CARBONDALE
Current Technology of Breast Cancer Detection Early detection of breast cancer is viewed as the best hope to decrease breast cancer mortality. Mammography The most commonly accomplished screening tool for early detection Supplemental tools Breast Ultrasound Breast CT Breast MRI Biopsy
X-ray Attenuation Model I Reconstructed Volume x T Ie Voxel j T Projection Image Pixel iI : incident intensity T : transmitted intensity : attenuation coefficientx : the length of the path where x-ray passes through the voxel
Current Tomosynthesis Imaging Geometry X-Ray tube Compression paddleReconstructed Breastplanes Digital detector X-ray tube rotates along an arc path above the breast Series of low dosage images are acquired at different angles
Reconstruction Algorithms of Digital Breast Tomosynthesis Reconstruct 3D object volume based on acquired 2D projection images Categories Mathematic reconstruction methods (SAA / BP) Filter-based reconstruction methods (FBP / MITS) Statistical reconstruction methods (MLEM, MOSC, etc) Algebraic reconstruction methods (ART, SART, etc)
Reconstruction Algorithm: SAA (Shift-And-Add) X-ray tubes (1) Compression paddle Plane I OZ Pi DetectorNote: O is the center of the reconstructed plane I. The shift amount of any pixel on Plane I will be equal tothe shift amount of O.
Reconstruction Algorithm: BP (Back-Projection) X-ray tubes Compression paddle Plane I A O Z Pi DetectorNote: O is the center of the reconstructed plane I. The shift amount of any pixel A on Plane I varies with itslocation.
Reconstruction Algorithm: FBP (Filtered Back Projection) Take BP as the foundation Apply high-freq and low-freq filters to enhance signals and suppress the noise Filters Ramp Filter Hamming Filter Gaussian Filter
Reconstruction Algorithm: MLEM (Maximum Likelihood Expectation Maximization) Originally invented by Lange and Fessler ML: maximize the likelihood of getting the projection T from the incident X-ray intensity I and attenuation coefficient μ. EM: one of the methods to solve ML problem. Iterative equation*: i li , j I i e l .u ( n ) i Ti (n) (j n 1) j j j (n) (n) (n) j i l ,u (n) i l i, j l, (n) i I i e * Wu, T., et al, “Tomographic mammography using a limited number of low-dose cone-beam projection images,” Med. Phys. 30, 365-380 (2003).
Reconstruction Algorithm: SART (Simultaneous Algebraic Reconstruction Technique) Originated from ART and was invented by Henderson Iteratively solves the attenuation equations Iterative equation *: Di j l ij n i lij j j lij (n) (n) (n) ( n 1) i l i , j j j j j* Zhang, Y., et al.,“A comparative study of limited-angle cone-beam reconstruction methods for breast tomosynthesis,” Med. Phys. 33(10), 3781-3795 (2006).
Image Reconstruction with Breast Phantom Breast Biopsy Training Phantom Low Dosage Middle Projection• Standard breast biopsy training phantom with solid masses and two micro-calcification clusters.• 25 projection images at a total exposure of 100 mAs (4mAs/projection) were obtained. View angle 48 degrees.
Image Reconstruction with Breast Phantom (a) BP (b) FBP (c) MLEM (d) SARTReconstructed ROIs of a mass (z=45 mm above the detector). (a) BP (b) FBP (c) MLEM (d) SARTReconstructed ROIs of a micro-cluster / micro-calcification (z=45 mm above thedetector).
Comparisons of Reconstruction Methods MITS and FBP are fast-speed reconstruction methods with deblurring algorithms to remove out-of-plane blur MLEM and SART are iterative methods based on ray- tracing computation; they can provide good image quality, but need a long running time
AcknowledgementsAppreciate our collaborators at The University of North Carolina at Chapel Hill (UNC), North Carolina State University (NCSU), Southern Illinois University (SIU).o The related work is supported by U.S. National Institutes of Health (NIH/NCI R01 CA134598-01A1).