Segmentation of prostate ultrasound images using an improved snakes model

We have applied an improved deformable 2D snakes modeling technique to the segmentation of prostate ultrasound images. Special measures were taken to deal with the high speckle noises and complex shapes of prostate boundaries. Median filtering proves to be effective in removing speckle noises, and by dynamically changing the rigidity parameters of the snakes model, our implementation has shown satisfactory preliminary segmentation results. We also have experimented with the gradient vector flow (GVF) snakes model with various edge detectors, and a combined LOG & Sobel operator is shown to perform the best in determining the edge map gradient field for the GVF.