A spleen segmentation method based on PCA-ISO

Spleen segmentation is one of the most important steps of computer-aided diagnosis. As the impact of pseudo sampling, local body effect, space aliasing and other factors in medical images, it is difficult for traditional methods to separate the spleen area from abdominal CT image accurately. In this paper, a novel spleen segmentation method based on PCA and ISO is proposed. This method can not only remove the abdominal muscles which is adjacent to the spleen area effectively, but also can separate the spleen area accurately, in the meanwhile, it can overcome the over-segmentation problem with effect. Firstly, we use the PCA method to do the training for entire CT image sequence containing the spleen to get the projection space. Over-segmentation problems may occur in some images when processed by the traditional ISO method. We can do a PCA mapping firstly, then select a fewer principal components to reconstruct the image. In this way we can remove the abdominal muscles adjacent to the spleen. Finally, the ISO method can accurately separate the spleen area in a short period of time from the reconstructed image. The abdominal CT image spleen segmentation experiments demonstrate the method proposed in this paper can overcome the over-segmentation problem arisen from the grayscale similarity and separate the spleen area of interest rapidly and accurately.