Active shape models—their training and application
Computer Vision and Image Understanding
An Active Contour Model without Edges
SCALE-SPACE '99 Proceedings of the Second International Conference on Scale-Space Theories in Computer Vision
| Hi-index | 0.00 |
Prostate segmentation in transrectal ultrasound (TRUS) images is highly desired in many clinical applications. However, manual segmentation is difficult, time consuming and irreproducible. In this paper, we present a novel automatic approach using narrow band contrast pattern to segment prostates in TRUS images. Implicit representation of the segmenting level sets curve is firstly trained via principal component analysis, which also constraints the shape of prostate into a linear subspace. Then the model evolves to segment the prostate by maximizing the contrast in a narrow band near the segmenting curve. Many experimental results demonstrate the performance of the proposed algorithm, whose favorableness is validated by comparing to the state-of-the-art algorithms. Especially, the shape of prostate segmented by our algorithm is close to the one manually obtained by expert, and the mean absolute distance is only 1.07±0.77mm, which is quite promising.