Room: Exhibit Hall | Forum 6
Purpose: To investigate a fast low-dose 4D digital tomosynthesis image reconstruction method in order to effectively reduce the artifacts in a sparse imaging tomosynthesis condition within faster acquisition time compared to conventional cone-beam based radiotherapy guidance.
Methods: A flat-panel detector mounted on a LINAC system was used to acquire the moving phantom images around a whole gantry rotation. We selected the several projections which were measured in the tomosynthesis angular ranges of anterior-posterior (AP) and lateral views. The projections were retrospectively sorted into five different phase bins and each respiratory phase dataset was reconstructed using a conventional filtered back-projection, simple gradient decent-based iterative scheme, and the proposed compressed-sensing (CS) methods. All reconstruction schemes were self-coded using CUDA programming with a single GPU card.
Results: The proposed reconstruction method showed higher contrast-to-noise ratios (CNRs) of spherical inserted target compared to the other conventional methods by factors of 1.2--3.4 for both AP and lateral tomosynthesis views. The streak reduction ratio (SRR) in the region of interest of the target also demonstrated that the proposed method could effectively reduce the streaking levels which were due to the under-sampled projections. The computation time taken for each algorithm to reconstruct whole five phase bin images using the proposed method was only ~2 min, which could demonstrate that GPU-accelerated programming may resolve the current limitation of large demanding computational burden.
Conclusion: The clinical use of 4D digital tomosynthesis is still controversial, but the authors believe that the proposed algorithm could play a significant role in low-dose application because our proposed low-dose reconstruction scheme may provide better outcomes compared to conventional reconstructions.
Respiration, Reconstruction, Tomosynthesis
IM/TH- RT X-ray Imaging: limited angle CBCT/Digital tomosynthesis reconstruction