(a) Photograph of the CBCT-capable mobile C-arm in the laboratory, showing a typical setup for a cadaveric head specimen supported by a carbon-fiber head frame. (b) Illustration of the system geometry. CBCT and tomosynthesis acquisitions differ in the extent of the orbital arc, . Relative to the cadaver setup in (a), the domains corresponding to axial, sagittal, and coronal image planes are (, ), (, ), and (, ), respectively.
Illustration of image quality in coronal and sagittal planes for tomosynthesis (acquisition angle, , ranging ) and CBCT . The vertical white line in the coronal image marks the location of the corresponding sagittal image and vice versa.
Clutter phantom. (a) Photograph of the phantom including a central solid water slab incorporating six inserts (A–F) ranging in electron density from polypropylene ( HU) to Teflon ( HU) surrounded by heterogeneous “clutter” (A–G) on each side of the slab. (b) Coronal and (c) sagittal CBCT images of the phantom. Images of the slab, as in (b), present tissue-simulating materials within a uniform background with noise attributable to quantum noise, view aliasing, and out-of-plane clutter.
Illustration of 3D localization tasks. CBCT images are shown, with crosshairs marking the average 3D position as indicated by six expert observers (defined as the “true” position of the landmark).
Tomosynthesis reconstructions of a 6.4-mm-diam sphere illustrating asymmetric blur and distortion associated with the angular range of source-detector orbit. (a) Triplanar views of the sphere reconstructed at various tomosynthesis arcs. (b) Signal profiles in the -direction through the center of the sphere, illustrating the loss in depth resolution at smaller tomosynthesis angles. Profiles were detrended by a linear fit to the background for purposes of illustration.
Coronal tomosynthesis reconstructions of the clutter phantom (Fig. 3) reconstructed at various settings of and such that the total dose was fixed in each case.
Surgical localization results for Task 1 (tip of K-wire). Triplanar views are shown for various settings of tomosynthesis angle. The crosshairs identify the “true” location and colored symbols indicate the individual responses of six observers.
Surgical localization results for Task 6 (sphenopalatine foramen). Triplanar views are shown for various settings of tomosynthesis angle. The crosshairs identify the “true” location and colored symbols indicate the individual responses of six expert observers. Note the error and posterior bias in localization in the -direction associated with limited depth resolution at smaller arcs.
Localization error plotted as a function of for all localization tasks. The measurement plotted at each value of represents the average response of six observers with error bars given by standard deviations.
Localization error averaged over all tasks and observers. (a) Absolute difference in individual (, , ) coordinates from “truth.” (b) In-plane localization error, characterized by the 2D distance in orthogonal planes.
Summary of dose and imaging/reconstruction time for 3D tomosynthesis and cone-beam CT. The terms “HiRes” and “LoRes” refer to voxel sizes 0.4 and (isotropic), respectively. Protocols for which imaging time is less than (roughly consistent with intraoperative time constraints) are highlighted in gray. Reconstruction times correspond to the (unoptimized) modified FDK algorithm used in our laboratory computed on a desktop PC (Dell Precision 470, dual Xeon CPU, RAM). Significant reduction in can be expected with high-speed reconstruction software and use of parallel processing.
Article metrics loading...
Full text loading...