Program Information
A Comparison of Artifact Blur Using SAA and FBP in Endorectal Digital Prostate Tomosynthesis with a Simulated Phantom
J Steiner*, K Matthews , G Jia , Louisiana State University, Baton Rouge, LA
Presentations
MO-RAM-GePD-IT-2 (Monday, July 31, 2017) 9:30 AM - 10:00 AM Room: Imaging ePoster Theater
Purpose: To compare endorectal digital prostate tomosynthesis (endoDPT) out-of-plane (OOP) structure blur in images reconstructed with shift-and-add (SAA) and filtered backprojection (FBP) algorithms.
Methods: A 0.02 cm diameter wire phantom tilted 10° relative to the detector plane was modeled 1.5 cm above a stationary detector in MATLAB®. A tomosynthesis scan was simulated using point-by-point projection. 21 projection images in 2° increments were collected with the x-ray source motion occurring in the x-direction in a plane parallel to and 100 cm above the detector surface. Imaging parameters were determined from previous endoDPT studies of physical plastic pelvic phantoms. Image planes (0.05 cm plane separation) were reconstructed above and parallel to the detector plane using SAA and FBP. FBP utilized three filters: A modified Ram-Lak filter to invert the blur from the backprojection process, a von Hann window to reduce noise accentuated by the Ram-Lak filter, and a von Hann filter to reduce noise in the slice thickness direction due to limited angular sampling.
Results: The artifact spread function (ASF) assesses the OOP structure blur artifact of the tilted wire as function of perpendicular distance (z) from the reconstructed plane (z₀). FBP resulted in a faster fall-off of the ASF compared to the SAA. For an arbitrary threshold of ASF = 20% (z₀ = 1.5 cm), the OOP structure was located at z = 0.056 cm and 0.137 cm for FBP and SAA, respectively, indicating FBP more effectively and efficiently removed OOP structure blur.
Conclusion: The ASF demonstrated FBP improves removal of OOP structure blur compared to SAA in endoDPT. Averaging improves contrast and noise effects in SAA, but results in significant OOP structure blurring as there is no mechanism for blur removal. FBP results in an overshoot artifact, due to incomplete sampling of the volume of interest.
Funding Support, Disclosures, and Conflict of Interest: No conflicts of interest to declare. Work supported in part by: Leveraging Innovation for Technology Transfer (LIFT2) Grant and the LSU I-CORPS Site Program.
Contact Email: