Publication Date:
2024-03-21
Description:
Background: In computed tomography (CT) quality assurance, alignment of image quality phantoms is
crucial for quantitative and reproducible evaluation and may be improved by alignment correction. Our goal was
to develop an alignment correction algorithm to facilitate geological sampling of sediment cores taken from a
cold-water coral mount.
Methods: An alignment correction algorithm was developed and tested with a CT acquisition at 120 kVp and
150 mAs of an image quality phantom. Random translation (maximum 15 mm) and rotation (maximum 2.86°)
were applied and ground-truth was compared to parameters determined by alignment correction. Furthermore,
mean densities were evaluated in four regions of interest (ROIs) placed in the phantom low-contrast section,
comparing values before and after correction to ground truth. This process was repeated 1000 times. After
validation, alignment correction was applied to CT acquisitions (140 kVp, 570 mAs) of sediment core sections up
to 1 m in length, and sagittal reconstructions were calculated for sampling planning.
Results: In the phantom, average absolute differences between applied and detected parameters after alignment
correction were 0.01 ± 0.06 mm (mean ± standard deviation) along the x-axis, 0.11 ± 0.08 mm along the y-axis, 0.15 ± 0.07°
around the x-axis, and 0.02 ± 0.02° around the y-axis, respectively. For ROI analysis, differences in densities were
63.12 ± 30.57, 31.38 ± 32.10, 18.27 ± 35.57, and 9.59 ± 26.37 HU before alignment correction and 1.22 ± 1.40, 0.76 ± 0.9,
0.45 ± 0.86, and 0.36 ± 0.48 HU after alignment correction, respectively. For sediment core segments, average absolute
detected parameters were 3.93 ± 2.89 mm, 7.21 ± 2.37 mm, 0.37 ± 0.33°, and 0.21 ± 0.22°, respectively.
Conclusions: The alignment correction algorithm was successfully evaluated in the phantom and allowed a correct
alignment of sediment core segments, thus aiding in sampling planning. Application to other tasks, like image quality
analysis, seems possible.
Type:
Article
,
PeerReviewed
Format:
text
Permalink