Program Information
Novel Strategies to Improve Image Quality in a Multi-Layer Imager (MLI)
Y Hu1*, R Fueglistaller2 , J Rottmann1 , M Myronakis1 , A Wang3 , P Huber2 , D Morf2 , J Star-Lack3 , R Berbeco1 , (1) Department of Radiation Oncology, Brigham and Women's Hospital, Dana-Farber Cancer Institute, and Harvard Medical School, Boston, MA, (2) Varian Medical Systems, CH-5405, Baden-Dattwil, Switzerland, (3) Varian Medical Systems, Palo Alto, California 94304-1030
Presentations
SU-K-FS4-10 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: Four Seasons 4
Purpose: We leverage the architecture of an MV multi-layer imager (MLI) to improve low-dose image quality. A novel processing technique called Shift-and-bin-addition (ShiBA) imaging, employs multiple-image offset-sampling and pixel-binning to increase signal gain with respect to electronic noise while maintaining sufficient resolution for high image quality.
Methods: Each layer readout of a prototype MLI detector was processed individually. Images from each layer underwent 4x4 pixel-binning (averaging) after an n-1 pixel shift in both detector directions, with n describing the layer number (n=1 being the top layer). Electronic noise power spectra (NPS) were measured from dark-fields (DF). Qualitative comparisons of resolution were assessed using a SeeDOS QC3 phantom and a head phantom. Benefits of hardware binning were analyzed by acquiring two sets of low-resolution (i.e. 2x2 binned) images from a standard, single-layer electronic portal imaging device (EPID).
Results: For the QC3 phantom, full-resolution (FR) images resolve the highest resolution (0.76 cycles/mm), 4x4 binning without shifting (BIN) the lowest (0.23 cycles/mm), while the ShiBA process was able to recover up to 0.43 cycles/mm. Visible pixelation observed in phantom images with BIN processing was not observed with ShiBA. Software binned images exhibit equivalent electronic noise to FR images. Electronic noise in up-sampled, binned images exhibit decreases in comparison to FR due to multiplication by the sampling (Sinc2) function. Improvement in the signal-to-noise-ratio (SNR) of a factor of 4 was observed for software binned treatments. Electronic NPS in hardware-binned images exhibited an 88% reduction compared to FR readout.
Conclusion: The ShiBA technique leverages the unique architecture of the MLI by enabling pixel binning without losses in image sampling. Improvement in SNR is observed while retaining or decreasing present levels of electronic noise. These techniques will significantly reduce the dose requirements for high quality MV-CBCT.
Funding Support, Disclosures, and Conflict of Interest: grant from Varian Medical Systems R01CA188446-01 (National Cancer Institute)
Contact Email: