SU‐GG‐I‐47: A New Limited‐Angle CT Reconstruction Method Based On Total‐Variation Minimization

Purpose: To develop a new CT reconstruction algorithm based on Total‐Variation (TV) minimization. The new algorithm is specifically designed for limited angle reconstruction to alleviate volume information loss caused by incomplete projections. Method and Materials: TV is defined as the summation of the gradients between adjacent pixels. A nonlinear conjugate gradient algorithm is used to minimize TV and projection error simultaneously. Pixel Signal‐to‐Noise Ratio (PSNR) was used to quantitatively evaluate reconstruction quality of a Shepp‐Logan phantom. Head and neck CT images were reconstructed from simulated projections to demonstrate reconstruction quality at different scan angle. The effects of weighting factors for TV and projection frequency were investigated. Results: The algorithm can reconstruct good quality images with the scan angle as low as 60 degree. With 90‐degree projections, the reconstructed Shepp‐Logan phantom has a high PSNR of 17 dB. Weighting factors for TV were determined through simulations. Image qualities were improved by weighting more high frequency of the projection data. In simulation tests using head/neck simulated projections, volume information can be recovered accurately from 60‐degree projections. Conclusion: The TV‐based reconstruction algorithm improves the quality of limited angle CT reconstruction. Research sponsored by a Varian research grant.