Program Information
Time Stability and Repeatability of CT Texture Measurements On Longitudinal CTs
Y Hao1*, J Paul1 , G Noid1 , L Court2 , D Mackin2 , Y Liu1 , X Li1 , (1) Medical College of Wisconsin, Milwaukee, WI, (2) The University of Texas MD Anderson Cancer Center, Houston, TX
Presentations
SU-K-601-14 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 601
Purpose: We have previously reported the feasibility of using radiation-induced CT texture changes during radiotherapy (RT) delivery as an imaging biomarker for RT response. This work aims to investigate the repeatability and uncertainty of CT texture changes over time measured from longitudinal CT sets of phantoms.
Methods: CT data for anthropomorphic, catphan, credence cartridge radiomics (CCR), and solid-water phantoms were acquired repeatedly at different time points, simulating longitudinal CT data sets, using a CT scanner installed inside RT room (Definition AS Open from Siemens). Standard protocol consisting of 120 kVp, 126 Effective mAs, 0.6 pitch, 1.2 mm focal-spot size, 17 mGy CTDIvol, 3 mm slice thickness, pixel spacing 0.799/0.799 mm, and B30f reconstruction kernel was used. For each phantom, various structure contours were created on the first CT and were populated to the CTs from other time points based on rigid image registration. A series of CT textures, such as mean Hounsfield Unit (HU), HU histogram, noise, local entropy, kurtosis, and coarseness, were extracted from CTs. The variations of these textures over time and their impact on the measurement repeatability of the texture changes were analyzed.
Results: The longitudinal changes of mean HU in the solid-water phantom were within 0.15, 0.5 and 2.0 HU, over time periods of one day, 8 weeks, and 5 years, respectively. The changes of noise, local entropy, kurtosis, and coarseness over 8 weeks (typical RT delivery duration) were 0.2, 0.06, 0.05, and 0.003, respectively. The repeatability coefficients (RC) of mean HU, local entropy, kurtosis, and coarseness in the ten cartridges of CCR phantom over 4 weeks were 1.35 HU, 0.21, 0.23 and 0.21, respectively.
Conclusion: The CT texture measurements of the phantoms from longitudinal CTs are highly repeatable and stable over time, indicating the machine related-uncertainty in longitudinal texture extraction is minimal.
Contact Email: