(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose:
Automated detection of image artifacts caused by air bubbles in a CT scanner cooling system when routine quality control (QC) passes.
Methods:
Two CT scanners with air bubbles in their cooling systems were selected for testing; one was used for air bubble artifact identification algorithm (ABAIA) development and the other was treated as novel. Water images using the vendor’s daily QC protocol and phantom from the training set were obtained before and after bubble removal. ABAIA was then tested on the novel dataset.
ABAIA was designed to assess the circular symmetry of slices of the images. First, each image was divided into 7 annular slices and the mean of the voxel values were calculated. Each slice was further divided into sections 25 voxels wide. The mean of each section was compared against the mean for the annular slice. Air bubbles present in the cooling system caused a reduction in attenuation of between 0.6-0.9 HU across a section, so any section that had a mean more than 0.6 HU lower than the slice mean was identified by the algorithm as containing an artifact.
Results:
ABAIA was tested against a novel dataset and achieved 91% accuracy, and correctly classified all of the post-bubble removal images as artifact free. Additionally, ABAIA identified the location and severity (variance from the slice mean) of each artifact present in each image.
Conclusion:
This work has the potential to immediately alert medical staff when quality control images are compromised by image artifacts. Further research could show whether this technique generalizes across manufacturers and other types of artifacts.
Funding Support, Disclosures, and Conflict of Interest: AV is employed by Advanced Quality Systems, who designs CT QC software.
Keywords
Taxonomy
IM- CT: Quality Control and Image Quality Assessment
Contact Email
