Developing a New Automatic Fast Algorithm for Recognizing the Seed Positions in Permanent Brachytherapy of Prostate
M Gholamian1*, m Yazdi2, R Faghihi3, M Mohammadi4, (1) Shiraz University, Shiraz, Fars, (2) Shiraz University, Shiraz, Fars, (3) Shiraz University, Shiraz, Fars, (4) Royal Adelaide Hospital, Adelaide, SASU-E-J-104 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose:Permanent prostate seed implant brachytherapy is one of the effective treatments for prostate cancer.Using three-film technique in which radiographic planar images are taken from three directions is a common method in evaluating the location and the number of seeds implanted in the prostate.The aim of this study is to validate the seeds' location and estimate accurately the number of implanted I-125 seeds using an automatic method.
Methods:We first enhance the resolution of seeds by applying some image preprocessing techniques on radiographic images. Then for recognizing automatically all seeds in the images we extract certain features and design filters based on mean and standard deviation of gray values for removing noises from images. Eventually we can segment the image into some distinguished regions. After filtering, without cropping any region of the image (for instance prostate) mean and standard deviation are measured for each region in the image which determine if this region belongs to seeds or not. At last we analyze histogram of seeds' region to still determine if this region belongs to an isolated seed or some overlapped seeds.
Results:Post-implant fluoroscopic images of 27 patients that were taken from three different directions were used in our study. Our algorithm effectively processed properties of three images of each patient and quickly and automatically recognized the seeds. The algorithm was able to recognize one hundred percent of seeds without missing any image properties.
Conclusion:previous works crop region of prostate that includes majority of seeds and then perform noise removal and seeds recognition. So they ignore seeds out of that region. In contrast, the merit of our presented algorithm is not only in recognizing automatically seeds in a short time by using adaptive filters but also not in not missing any seeds even out of prostate region.