Exhibit Hall | Forum 9
Purpose: Image processing methods for correcting detector-module-boundary-line artifacts, detector-module-response inhomogeneity, and x-ray tube output-fluctuation effects during 1000 fps HSA sequence acquisitions.
Methods: The 100 micron pixel-pitch Aries PCD (Varex) has a 768 by 512 pixel FOV and is composed of 12 modules. Image sequences of a patient-specific carotid aneurysm phantom attached to a flow loop with and without omnipaque contrast injection were taken at 70 kVp. To correct for module-boundary artifacts and inter-module inhomogeneity, background images were subtracted from the phantom images; to quantify inhomogeneity, the coefficient of variation for the average intensity of the individual modules of flat-field images was calculated pre-correction and post-correction. At detector frame rates of 1000 fps, high-frequency x-ray tube output-fluctuations produce noticeable changes in image intensity throughout a sequence. Output-fluctuation correction factors were calculated as the ratio of the average intensity of the full sequence to the average intensity of each individual frame. Each frame’s intensity is then multiplied by the respective correction factor. To quantify the effects of image correction on velocity calculations, Optical Flow algorithm calculations were performed on contrast-injected model image sequences before and after output-fluctuation correction.
Results: Detector-module-boundary-line artifacts and inter-module inhomogeneity are minimized in flat-field corrected images with a reduction of inter-module response coefficient of variation from 0.023 to 0.00047. The output-fluctuation correction reduced intensity fluctuations between frames and improved the flow-quantification accuracy . Increases in calculated velocities due to output-fluctuations in the pre-correction sequences are not evident in post-correction sequences.
Conclusion: This work shows that simple image corrections for detector-module boundary-line artifacts, individual detector-module response differences, and x-ray tube output-fluctuation effects can improve 1000-fps HSA image appearance and flow quantification.