Contrast or Drift?

Project Overview

X-ray computed tomography (CT) is used extensively at Lawrence Livermore National Laboratory to nondestructively characterize a variety of materials. An important goal at the laboratory is to generate CT images with quantitative accuracy on the order of 1%. Various physical phenomena, combined with approximations in reconstruction, cause deterministic uncertainties (errors) that are greater than our accuracy goal. One of these phenomena is that the spectrum emitted by an x-ray tube is spatially variant along the anode-cathode axis, called the heel effect. The changing spectrum leads to a drift in x-ray attenuation, which would be attributed to contrast in the material. We had not previously sufficiently characterized the heel effect on our test-bed scanner system, although this system is used for internal lab research and external customer projects; therefore, we did not sufficiently characterize its impact on reconstructed image quality. The scanner has an YXLON model Y.TU 450-D11 x-ray tube, with a 450 kV maximum rating.

We experimentally quantified the heel effect by scanning water, aluminum and rubidium bromide solution (RbBr), placed in two different locations in the field of view: near the top and bottom of the two-dimensional CT detector array. We used a collimator with an inexpensive modification to prevent scatter from contaminating the characterization of the heel effect. The scans showed that measured linear attenuation coefficient increased from top to bottom. The discrepancy increased with effective atomic number of the scanned material (as low as 5% for water, 15 % for RbBr). We also performed Monte Carlo simulations to validate and better understand our experimental data. We used the water scans to estimate the spectra at the scanned locations. Using the position-dependent estimated spectra in existing image reconstruction software, we compensated for the heel effect, reducing the discrepancy to nearly zero for water, and about 2% in RbBr. With additional experiments, we demonstrated that the amplitude of the heel effect was comparable to that of object-scattered radiation, for the objects we scanned. In this feasibility study, we measured the heel effect for two spectra, but recommend that we do this simple calibration and compensation for other scanning techniques, particularly when we need to use a large area of the panel.

Mission Impact

In this feasibility study, we quantified the impact of the heel effect on our images, and defined a simple compensation procedure to get us closer to our goal of quantitatively accurate CT. Quantitatively accurate CT will impact all programs that use CT scanning with a large field of view, including weapons research (such as aging studies), characterization of explosives, and the characterization of additively or conventionally manufactured industrial parts.

Publications, Presentations, and Patents:

Karimi, Seemeen, Hall, James and Ruelas, Joshua, "Heel Effect Measurement and Compensation" (Presentation, ASNT Research Symposium, Columbus OH, June 2023).