Purpose. Detection and measurement accuracy using industrial CT is limited by image quality (IQ). The reconstruction algorithm is a strong contributor to IQ. Image quality factors includes spatial resolution, image contrast, and image variations from random noise and artifacts. Artifacts include cupping, shading, and streaks. We propose and assess a method (phantoms, IQ metrics, and analysis algorithms) designed to measure IQ in industrial CT images. Methods. Four different annular phantoms (4-inch OD; 2-inch ID) were constructed from stainless steel. Each phantom was scanned in an industrial cone-beam CT system at four scatter levels. Images were reconstructed using five reconstruction algorithms (filtered backprojection (FBP) and four advanced reconstruction algorithms with different artifact correction and noise suppression methods), producing 80 datasets (4 phantoms, 4 scatter levels, 5 reconstruction algorithms). For each dataset, spatial resolution was measured using the edge response function (ERF), its derivative, the point spread function (PSF), and the PSF’s Fourier transform, the modulation transfer function (MTF) with methods proposed in the ASTM standards, with several modifications. Also for each dataset, image contrast (C), random noise (R), and artifact level (A) were computed in relevant regions of interest; from these, total variations (T) and contrast-to-variation ratios (C/R, C/A, C/T) were computed. Results. All IQ metrics showed trending results for each reconstruction algorithm over the 20 phantom/scatter conditions but different results for different reconstruction algorithms. PSF full-width-at-half-maximum (FWHM) ranged from 29.7 mils (0.0297 inch) to 15.0 mils. C/T ranged from 4.3 to 15.9. Conclusion. The proposed method was able to quantitatively evaluate the quality of industrial CT images in terms of spatial resolution, contrast, noise, and artifacts.

Related Publications