Image distortion and mechanical inaccuracies inherent to C-arm fluoroscopes pose obstacles to their use in planning for interstitial and intracavitary HDR brachytherapy. We present our implementation of post-processing algorithms which correct these problems, in which emphasis is placed on accuracy, speed, ease of calibration, and cost minimization. Images are corrected using the local technique using a fine (4 mm) triangular calibration grid. The pixel coordinates of the grid points as obtained by placing the grid on the fluoroscope image intensifier are forward-mapped to the ideal corrected coordinates, taken from a scanned film image of the calibration grid.
Extrapolation algorithms are employed to correct images to the very edge of the field of view. Even after correcting for image distortion, mechanical inaccuracies inherent to the C-arm cause mismatch between AP and lateral images.
An original parametrization of the C-arm camera geometry is presented. By obtaining distortion-corrected AP and lateral fluoroscope images of the calibration grid, placed obliquely at isocenter, and performing a multidimensional fit, we obtain data-constrained values for these camera geometry parameters. These results are subsequently used to quickly and easily correct AP and lateral images for distortion and camera geometry; these final, doubly-corrected images reduce internal distortion and AP/lateral mismatch to less than 1 mm.