Algorithmic determination of source orientations for the CivaSheet directional brachytherapy device

The CivaSheet device uses multiple directionally shielded Pd-103 CivaDot sources to produce a directional planar dose distribution. In postplanning, manually digitizing the 3D source orientation is challenging because the 3D vector must be digitized by using 2D displayed images. The aim of this study is to develop an algorithm that will automatically determine the direction of each CivaDot source based on the location of sources adjacent to it.

The algorithm determines the source direction by averaging the normal directions of multiple local planes established by the adjacent sources. The algorithm was tested on a manually constructed CivaSheet-like device that was CT scanned in known flat geometries and two known curved geometries. Algorithmically determined source directions were compared with the known directions. The algorithm was also used on a postplan for a gynecological pelvic sidewall tumor bed implant and compared against manual digitization of the source directions.

For the flat and curved test geometries, the average angular difference between the algorithm determined and known orientation was 1.2° ± 0.8° (flat geometry), 1.7° ± 2.1° (curve about vertical axis), and 2.3° ± 2.4° (curve about horizontal axis). For the patient case, results showed on average a 23.1° ± 10.8° difference between the manual digitized orientation and the algorithm's predicted orientation.

The algorithm calculates the source orientation with accuracy better than 2.3° for the controlled experiments. In addition to its accuracy, the algorithm produces consistent results and lessens the difficult challenge of orienting the partially shielded sources.