Multiresolution iterative reconstruction in high-resolution extremity cone-beam CT

Abstract

Application of model-based iterative reconstruction (MBIR) to high resolution cone-beam CT (CBCT) is computationally challenging because of the very fine discretization (voxel size <100 µm) of the reconstructed volume. Moreover, standard MBIR techniques require that the complete transaxial support for the acquired projections is reconstructed, thus precluding acceleration by restricting the reconstruction to a region-of-interest. To reduce the computational burden of high resolution MBIR, we propose a multiresolution Penalized-Weighted Least Squares (PWLS) algorithm, where the volume is parameterized as a union of fine and coarse voxel grids as well as selective binning of detector pixels. We introduce a penalty function designed to regularize across the boundaries between the two grids. The algorithm was evaluated in simulation studies emulating an extremity CBCT system and in a physical study on a test-bench. Artifacts arising from the mismatched discretization of the fine and coarse sub-volumes were investigated. The fine grid region was parameterized using 0.15 mm voxels and the voxel size in the coarse grid region was varied by changing a downsampling factor. No significant artifacts were found in either of the regions for downsampling factors of up to 4×. For a typical extremities CBCT volume size, this downsampling corresponds to an acceleration of the reconstruction that is more than five times faster than a brute force solution that applies fine voxel parameterization to the entire volume. For certain configurations of the coarse and fine grid regions, in particular when the boundary between the regions does not cross high attenuation gradients, downsampling factors as high as 10× can be used without introducing artifacts, yielding a ~50× speedup in PWLS. The proposed multiresolution algorithm significantly reduces the computational burden of high resolution iterative CBCT reconstruction and can be extended to other applications of MBIR where computationally expensive, high-fidelity forward models are applied only to a sub-region of the field-of-view.

Three‐dimensional mapping of the joint space for the diagnosis of knee osteoarthritis based on high resolution computed tomography: Comparison with radiographic, outerbridge, and meniscal classifications

“…46,47 Highresolution cone-beam CT systems with voxel size <100 mm for assessment of bone and joint health are under development. 48 We have demonstrated that morphological analysis of the joint space performed on CT images with sufficiently high resolution is pertinent to describe indirectly tibial cartilage and meniscal impairments. The parameters JS_min, JS_asym, and JS_ent are the most interesting to characterize OA, and JS_asym despite a low reproducibility is a good predictor of the cartilage thickness.…”

Evaluation of the geometric accuracy of computed tomography and microcomputed tomography of the articular surface of the distal portion of the radius of cats

“…Other CT methods, including conventional CT with a voxel size < 0.6 mm and cone beam CT, would have provided greater accuracy but were not available in our teaching hospital. High-resolution cone beam CT images have slices as thin as 0.1 mm combined with low radiation, relative to slice thickness and radiation for conventional CT. 29,30 The distal portion of the radius is a common site of limb-sparing surgery in companion animals 31 and a site of hemiarthroplasty in humans. 32 Accurate assessment of the geometry of joint surfaces is a prerequisite for optimization of articulating custom orthopedic implants (eg, custom arthroplasties or hemiarthroplasties).…”

Multi-resolution statistical image reconstruction for mitigation of truncation effects: application to cone-beam CT of the head

“…The resulting boundary between the fine and coarse regions is outside the cranium (in air, presumably not of diagnostic interest), so downsampling / upsampling voxels in the other region is not considered when calculating neighboring voxel differences in the subsequent image reconstruction. In the current work, we investigate an implementation specifically with two voxel sizes (coarse and fine), but the term “multi-resolution” (c.f., “dual resolution”) is used for consistency with previous work (Delaney and Bresler 1995, Cao et al 2016) and for generality in anticipation of future work in which voxel size is more continuously varied from a fine value in the SFOV to progressively coarser values outside the SFOV.…”

“…In the regularization part, taking the fine region as an example, K F is the number of neighboring voxels in the fine region, and Ḣ and ω are the gradients and curvatures of the Huber penalty function H , respectively. While the pseudocode in (Cao et al 2016) used a small detector pixel size for the projection of a high-resolution region-of-interest and a large detector pixel size for the rest of the projection data, the pseudocode here used a single detector pixel size.…”