(a) Close-up photograph of the Fletcher–Weeks CT-compatible afterloadable colpostats and one of the tandems used in this study. (b) Computed CBCT projections of the 3D tandem and right colpostat models, where the image background represent uniform elliptical water cylinder. The image intensity values represent an arbitrary integer number assigned to each material in the model.
An example illustrating postprocessing of experimentally acquired ACUITY projection images. (a) Raw projection image, (b) top-hat filtered image, (c) binary image, and (d) blurred grayscale image used as an input to the generalized IFPM algorithm.
An illustration of the iterative convergence process for a simulated implant consisting of tandem and bilateral colpostats for a 25 mm colpostat separation (see convergence rate graph in Fig. 4). (a) Initial estimate of the applicator configuration, (b) computed images at convergence, (c) the true/synthetic measured images, and (d) difference between (b) and (c), where the rows represents different gantry angles. The horizontal line in the third row shows the qualitative comparison of an initial estimate in part (a) with computed and measured images in part (b) and (c), respectively. The generalized IFPM algorithm was able to reproduce each applicator pose, as well as overlapping components.
The similarity metric score vs iteration number for the generalized IFPM algorithm for the three simulated full ICB applicator configurations.
Point-by-point overlay of the reconstructed applicators with the true/synthetic measured applicators, demonstrating near coincidence achieved by the 3D/2D gIFPM registration and applicator reconstruction as a unified process.
Superposition of experimentally acquired binary images (white) with automatically reconstructed applicators positions (black) projected onto the detector planes. (a) 0° gantry angle, (b) −30° gantry angle, and (c) gantry angle, respectively, when using 40 mm colpostat separation. The applicator registration error was less than 1 mm for the intrauterine tandem and about 1.5 mm for the bilateral colpostats on each image plane.
Accuracy of gIFPM reconstructed poses for three simulated full applicator system configurations. The difference for the each applicator component position and orientation coordinates is reported.
The residual 2D registration error between computed and measured applicator projection in terms of 2D radial difference in each image plane. The center of mass position of the entire tandem is reported, whereas the center of mass of the source position is empirically calculated for the right and left colpostats.
Article metrics loading...
Full text loading...