Bone Segmentation on X-ray
Posted: 2013-08-07T08:52:53-07:00
Hi everyone,
I'm kind a stuck with one thing. I hope that somebody will give me a hint about the direction that I should go. So here is a thing. I have a fluoroscopy video of Cervical Vertebra from which I need to extract vertebras and its motion. This is how it looks like:
http://www.youtube.com/watch?v=q0cKhkJk30A
Let me show you what I've already tried. Let's say that this is the starting frame of the movie.
At the beggining I've decided that system will be semi automaticly so I cut a square around every vertebra like this
and I've tried some edge filtering. You can see below the results for the best thresholds. From the top left: Sobel, Canny, Roberts, Prewitt, Log, Zero Crossing.
after that I've tried Hough transform for line and parabola
But I've noticed that It doesn't give any satisfactionary results. Although the algorithm found some lines or parabola I still don't see how can I solve this problem. I know that Active Shape Model could give good results but I am afraid that it will take a lot of computation time. Could you help me and give me a hint in which direction shoud I go? There already are some informations in net how to do that like this: http://www.hindawi.com/journals/ijbi/2011/621905/ but those people work on better image quality and the segmentation gives a totally different results. Help guys.
I'm kind a stuck with one thing. I hope that somebody will give me a hint about the direction that I should go. So here is a thing. I have a fluoroscopy video of Cervical Vertebra from which I need to extract vertebras and its motion. This is how it looks like:
http://www.youtube.com/watch?v=q0cKhkJk30A
Let me show you what I've already tried. Let's say that this is the starting frame of the movie.
At the beggining I've decided that system will be semi automaticly so I cut a square around every vertebra like this
and I've tried some edge filtering. You can see below the results for the best thresholds. From the top left: Sobel, Canny, Roberts, Prewitt, Log, Zero Crossing.
after that I've tried Hough transform for line and parabola
But I've noticed that It doesn't give any satisfactionary results. Although the algorithm found some lines or parabola I still don't see how can I solve this problem. I know that Active Shape Model could give good results but I am afraid that it will take a lot of computation time. Could you help me and give me a hint in which direction shoud I go? There already are some informations in net how to do that like this: http://www.hindawi.com/journals/ijbi/2011/621905/ but those people work on better image quality and the segmentation gives a totally different results. Help guys.