(Sunday, 7/29/2018) 1:00 PM - 1:55 PM
Room: Room 205
Purpose: To analyze noise and texture of computed tomography (CT) images reconstructed with iterative reconstruction kernels and investigate their potential application in CT protocol optimization.
Methods: A water phantom was scanned in a Siemens Definition Force at different exposures (25mAs to 300mAs). CT images were reconstructed with Filtered Back Projection (FBP) and Advanced Modeled Iterative Reconstruction (ADMIRE) at different strength settings (level 1 to 5). Unsubtracted and subtracted noise power spectra (NPS) were measured for the analysis of noise and texture properties. The sum of differences between NPS of different exposures and strength settings were calculated to identify the NPS similarity. The acquisition parameters that generated the most similar NPS were used to optimize CT protocols.
Results: Image noise strongly depends on the exposure settings, following a linear relationship for both FBP and ADMIRE. Image texture is independent of exposure. The normalized NPS across all exposures for any given reconstruction share a similar shape and peak spatial frequency. For ADMIRE, the increase in reconstruction strength (from 1 to 5) displays an increasing smoothing effect on the image which can be observed by a linear decrease in image noise and a shift of the NPS peak towards lower frequencies. Taking Br40 reconstruction kernel as an example of CT protocol optimization, the NPS of FBP at 300mAs, Br40(1) at 250mAs, Br40(2) at 200mAs, Br40(3) at 175mAs, Br40(4) at 125mAs, and Br40(5) at 75mAs show the comparable noise and texture properties. This indicates that CT images acquired with 75mAs and reconstructed using Br40 with strength 5 may match those acquired at 300mAs and reconstructed with FBP, leading to 75% radiation dose reduction.
Conclusion: The analysis of noise and texture properties provides a potential guideline for the optimization of CT protocols with iterative reconstruction algorithms. An in-vivo study is being conducted for further evaluation.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
