Program Information
Sliding Motion Compensated Simultaneous 4D-CBCT Reconstruction
J DANG1*, F Yin2 ,T You1 , C Dai1 ,J Wang3 , (1) The Affiliated Hospital of Jiangsu University, Zhenjiang, Jiangsu, (2) Duke University Medical Center, Durham, NC, (3) UT Southwestern Medical Center, Dallas, TX
Presentations
TH-CD-303-12 (Thursday, July 16, 2015) 10:00 AM - 12:00 PM Room: 303
Purpose:Current deformable-vector-field(DVF) based 4D-CBCT reconstruction estimates anisotropic smoothed DVFs between each respiration phase, which is not true for all patients because organs usually move anisotropically. We investigated an anisotropic motion model by extracting organ boundary local motions(e.g. sliding) and incorporating them into 4D-CBCT reconstruction to optimize the motion modeling and reconstructions.
Methods:Initially, a high quality reference phase is reconstructed by a modified Simultaneous Algebraic Reconstruction Technique(mSART) using all phase projections. Then sliding motion was estimated by matching the measured projection to the forward projection of the deformed reference phase. Meanwhile the moving organ boundary was semi-automatically segmented and the normal vectors of the boundary were extracted and incorporated for DVF estimation optimization. The regularization term in the objective function adaptively regularizes the DVF by 1)isotropically smoothing the DVF for intra-organ sites, 2)smoothing the DVF at boundary sites along each normal vector’s normal direction; and 3)leaving the tangent directions unsmoothed. A non-linear conjugate gradient optimizer was used for optimization. The DVF solution deforms the reference phase to the final corresponding reconstructed phase. A sliding motion phantom is simulated to test the local motion incorporated 4D-CBCT reconstruction. It contains two square tubes with increasing intensities: one static bright tube and another medium intensity tube slides against the static one. The relative reconstruction error(RE), defined by the square-root-of the ratio between the sum of squared absolute error(the difference between one reconstructed phase to its corresponding truth, namely the phantom itself) and the sum of squared truth value for all pixels, was used to evaluate the reconstruction accuracy improvement.
Results:RE is 16.20% for sliding motion modeling based result and 27.04% for results without sliding modeling. Final DVF results also preserve better tube boundaries with sliding modeling.
Conclusion:Preliminary phantom results indicated the local sliding motion incorporated 4D-CBCT motion modeling improves final DVF estimation accuracy.
Contact Email: