An automatic seed finder for brachytherapy CT postplans based on the Hough transform

The purpose of the work is to describe a new algorithm for the automatic detection of implanted radioactive seeds within the prostate. The algorithm is based on the traditional Hough transform. A method of quality assurance is described as well as a quantitative phantom study to determine the accuracy of the algorithm.

An algorithm is described which is based on the Hough transform. The Hough transform is a well known transform traditionally used to automatically segment lines and other well defined geometric objects from images. The traditional Hough transform is extended to three-dimensions and applied to CT images of seed implanted prostate glands. A method based on digitally reconstructed radiographs is described to quality assure the determined three-dimensional positions of the detected seeds. Two phantom studies utilizing eight seeds and nine seeds are described. All eight seeds form a contiguous a square while the nine seed phantom describes seeds which are placed side-by-side in groups of two and three. The algorithm is applied to the CT scans of both phantoms and the seed positions determined.

The algorithm has been commercially developed and used to perform postsurgical dosimetric assessment on approximately 1000 patients. Using the described quality assurance tool it was determined that the algorithm accurately determined the seed positions in all 1000 patients. The algorithm was also applied to the eight seed phantom. The algorithm successfully found all eight seeds as well as their seed coordinates. The average radial error was determined to be 0.9 mm. For the nine seed phantom, the algorithm correctly identified all nine seeds, with an average radial error of 3 mm.

The described algorithm is a robust, accurate, automatic, three-dimensional application for CT based seed determination.