A biomechanical modeling-guided simultaneous motion estimation and image reconstruction technique (SMEIR-Bio) for 4D-CBCT reconstruction

Abstract

Reconstructing four-dimensional cone-beam computed tomography (4D-CBCT) images directly from respiratory phase-sorted traditional 3D-CBCT projections can capture target motion trajectory, reduce motion artifacts, and reduce imaging dose and time. However, the limited numbers of projections in each phase after phase-sorting decreases CBCT image quality under traditional reconstruction techniques. To address this problem, we developed a simultaneous motion estimation and image reconstruction (SMEIR) algorithm, an iterative method that can reconstruct higher quality 4D-CBCT images from limited projections using an inter-phase intensity-driven motion model. However, the accuracy of the intensity-driven motion model is limited in regions with fine details whose quality is degraded due to insufficient projection number, which consequently degrades the reconstructed image quality in corresponding regions. In this study, we developed a new 4D-CBCT reconstruction algorithm by introducing biomechanical modeling into SMEIR (SMEIR-Bio) to boost the accuracy of the motion model in regions with small fine structures. The biomechanical modeling uses tetrahedral meshes to model organs of interest and solves internal organ motion using tissue elasticity parameters and mesh boundary conditions. This physics-driven approach enhances the accuracy of solved motion in the organ's fine structures regions. This study used 11 lung patient cases to evaluate the performance of SMEIR-Bio, making both qualitative and quantitative comparisons between SMEIR-Bio, SMEIR, and the algebraic reconstruction technique with total variation regularization (ART-TV). The reconstruction results suggest that SMEIR-Bio improves the motion model's accuracy in regions containing small fine details, which consequently enhances the accuracy and quality of the reconstructed 4D-CBCT images.

Dosimetric evaluation of 4D‐CBCT reconstructed by Simultaneous Motion Estimation and Image Reconstruction (SMEIR) for carbon ion therapy of lung cancer

“…Because of the iterative, interleaved, and alternate nature of motion estimation and image reconstruction within SMEIR, the overall speed of our workflow may not yet be fast enough for real‐time clinical implementation. It needs about 30 min on a standard central processing unit to obtain a satisfactory image within a single iteration of SMEIR with a total of 20 iterations required for the overall convergence . The GPU dose calculation typically takes about an hour to optimize and about 10 min for dose calculation on each image.…”

“…Recently, we have developed a new technique for motion‐compensated 4D‐CBCT image reconstruction based on simultaneous motion estimation and motion‐compensated image reconstruction (SMEIR) . Simultaneous motion estimation and motion‐compensated image reconstruction obtains the deformation vector fields (DVF) by warping the projection of the reference phase image to those of all other phases in an iterative process . Hence, SMEIR can reconstruct any phase 4D‐CBCT by explicitly considering the motion model between different phases, effectively suppressing the view aliasing artifacts caused by the limited number of projections.…”

A biomechanical modeling guided simultaneous motion estimation and image reconstruction technique (SMEIR-Bio) for 4D-CBCT reconstruction

“…We developed the SMEIR-Bio technique based on the SMEIR algorithm for 4D-CBCT reconstruction but added biomechanical modeling to the workflow to improve DVF and image reconstruction accuracy in low-contrast regions or regions with small fine details. Lung surface DVFs solved by 2D-3D deformation, whose accuracy benefits from the highcontrast lung boundary, served as the boundary condition to drive finite element analysisbased biomechanical modeling to fine-tune the intra-lung DVFs [54]. Based on our evaluation using 11 lung cancer patient cases, the proposed SMEIR-Bio algorithm enhances the reconstructed 4D-CBCT quality, as well as the accuracy of inter-phase DVFs.…”

Advanced 4-dimensional cone-beam computed tomography reconstruction by combining motion estimation, motion-compensated reconstruction, biomechanical modeling and deep learning

“…1, 2), especially for under-sampled acquisition scenarios. To improve the motion estimation accuracy of SMEIR of these fine details, we further introduced biomechanical modeling into the SMEIR algorithm (SMEIR-Bio), to solve the DVF via a physics-driven approach [41]. Biomechanical modeling-based deformable registration has been found effective, especially at solving the deformation at low-contrast regions with minimal intensity variation signals [47][48][49][50][51].…”

“…The 4D-CBCT images at other phases are deformed from the mCBCT via the inverse DVFs simultaneously-optimized by the 2D-3D deformation algorithm. In this study, we detailed the general philosophy and workflow of the SMEIR algorithm, and introduced two new developments based on the original SMEIR algorithm: the biomechanical modeling-guided SMEIR (SMERI-Bio) and the SMEIR algorithm with artificial intelligence (AI)driven DVF fine-tuning (SMEIR-Unet) [41]. In comparison to the original SMEIR algorithm, the SMEIR-Bio algorithm introduced biomechanical modeling to improve the intra-lung DVF accuracy to better reconstruct the fine details in lung, and better capture their motion.…”