Program Information
Band-Limited Noise Structure Analysis On Images With Adaptive Statistical Iterative Reconstruction (ASIR) in Abdominal CT
Y Zhou*, A Scott , J Allahverdian , C Lee , Cedars-Sinai Medical Center, Los Angeles, CA
Presentations
SU-F-207-15 (Sunday, July 12, 2015) 4:00 PM - 6:00 PM Room: 207
Purpose: To investigate the noise details in ASIR reconstructed images with regard to location and signal dependency and distribution normality in abdominal CT scans.
Methods: A realistically shaped abdomen phantom (CIRS TE-07). The phantom was customized to allow for the insert of a low contrast module (CIRS 700 QA). Helical CT scans (120 kV, pitch 1.375, and thickness 5 mm) were performed on a GE 750 HD at two dose levels of 8 mGy and 18 mGy. The images were reconstructed using different ASIR fractions (0 - 100%). The noise images, obtained using subtraction from adjacent slices, were partitioned into different scales using matrices of square cells sized differently. The noise granularities were quantified by the standard deviation of the matrix cell pixel means distribution at each cell size. For each ASIR blending fraction, the following noise granularity properties were examined: the dependency on the matrix location and the dependency on the low contrast signal. The Komogorov-Smirnov normality test was performed on the matrix cell pixel means distributions.
Results: For each ASIR blending fraction, the standard deviation of the matrix cell means distribution at each cell size was found independent of the matrix location. Furthermore, it does not vary due to the presence of the low contrast signal. The normality tests demonstrated that the cell means distributions are significantly drawn from Gaussian at the 0.05 level.
Conclusion: The noise granularity independency on matrix location with ASIR reconstructed images suggests bigger matrices can be used for a better statistical description of the noise. The noise granularity independency on low contrast signal suggests that noise detail analysis can be done in a bigger area in the absence of the signal. Finally, the Gaussian distribution of the matrix cell means lends itself to statistically defining the minimum detectable contrast.
Contact Email: