Room: AAPM ePoster Library
Purpose: C-arm cone-beam CT can acquire a volumetric image of the knee joint in a single short scan. However, the limited number of projections, limited tube output, and C-arm geometric “wobble” may induce noise and artifacts in the reconstructed images. In this work, we investigate the Tomographic Iterative GPU-based Reconstruction (TIGRE) toolbox to reconstruct high quality images of the knees.
Methods: The knees of human subjects were scanned in a non-weight-bearing (supine) position using a Siemens Artis Zeego C-arm CBCT system. We quantified the wobble of both the x-ray source and flat panel detector via geometric analysis of pre-calibrated projection matrices, and configured the acquisition geometry in the TIGRE toolbox accordingly. Then, we reconstructed the knee images using both Feldkamp (FDK) and an iterative algorithm (conjugate gradient least squares, CGLS), at both 1.0 mm and 0.5 mm isotropic voxel size. Finally, we examined the effect of additional detector binning (e.g., 4x4 binning) on the reconstructed images.
Results: The reconstructed images are degraded with severe geometric artifacts if a circular trajectory is assumed and no wobble correction is applied. However, using the projection matrices produced accurate reconstructions. In comparing the analytical FDK and iterative CGLS reconstruction algorithms, we found CGLS improved the reconstructed image quality by substantially reducing streak artifacts from the short scan acquisition, although reconstruction time was increased by >10x. Also, even though 4x4 binning is already smaller than the voxel size, using 2x2 binning yields higher resolution of bone structures but also increases reconstruction time.
Conclusion: We have proposed and validated a reconstruction framework for C-arm CBCT imaging based on the TIGRE toolbox. The improved image quality by the CGLS algorithm has potential to facilitate cartilage segmentation and knee osteoarthritis assessment. Ongoing work will extend it to horizontal CT acquisitions, to achieve high quality reconstructions of weight-bearing knees.
Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by NIH R01 AR065248 and Siemens Healthineers