Two improved segmentation algorithms for whole cardiac CT sequence images

Abstract

The segmentation of the whole heart is quite necessary in computer-aided diagnosis. However, traditional medical image segmentation algorithms cannot achieve whole heart segmentation of CT sequence images accurately. In this study, two improved segmentation algorithms for whole cardiac CT sequence images are developed, as the correlation among adjacent slices of CT sequence images is considered, the segmentation result of the prior slice is used as the initial contour of the next slice, and combined with a Gauss filter, a morphological operation, and other image processing operations to improve the traditional region-based and edge-based segmentation algorithm. The improved algorithms are then applied in whole heart segmentation of CT sequence images. Experimental results show that: the improved region-based algorithm can avoid leaked segmentation and error segmentation appeared in the segmentation results based on traditional region growing algorithm, and improved edge-based segmentation algorithm can be applied in the segmentation of images without closed contour, which cannot be segmented by traditional edge-based algorithm. Besides, these two improved segmentation algorithms proposed in this paper have significant effect on cardiac location, cardiac surgical navigation, interventional treatment, computer-aided diagnosis applications and so on.