Portal Vein Segmentation from 3D CT Data

Multiscale line filtering was applied to abdominal CT images taken by a helical CT scanner so as to segment the portal veins to localize a tumor with the relation to them for surgical planning. The CT dataset consisted of 43 slices of 512 x 512 pixels; the pixel dimensions were 0.59 mm2. The beam width was 3 mm and the reconstruction pitch was 2.5 mm. The CT data were imaged using CTAP

ยป

0

Figure 10.5: Liver vessel (portal vein) segmentation from abdominal CT images. (a) Original cross-sectional images (left) and segmented liver region (right). (b) Original (left) and line-filtered (right) surface-rendered images.

Figure 10.5: Liver vessel (portal vein) segmentation from abdominal CT images. (a) Original cross-sectional images (left) and segmented liver region (right). (b) Original (left) and line-filtered (right) surface-rendered images.

(CT arterial portography)2; the portal veins had high CT values due to the injection of contrast material. A region of 400 x 400 pixels from each slice was trimmed, which included the whole liver (the left frame of Fig. 10.5(a)), and further the image size was reduced to half using the Laplacian pyramid [31] to reduce a computational amount to a practical level. The liver regions were roughly hand-segmented by a radiology specialist and used as a mask (the right frame of Fig. 10.5(a)). The CT values were converted so that the image intensity f was zero for less than fmta, fmax - fmin for more than fmax, and f - fmta for between fmin and fmax (where fmin = 1000 and fmax = 1300). Line filtering was applied to the sinc-interpolated images using y23 = y12 = 1.0, a = 0.25, ai = 0.8 pixels,

2 The CT data were obtained by a helical CT scanner with the 20-sec delay following the administration of contrast material using a catheter inserted in the SMA (superior mesenteric artery). This method of portal vein imaging is called CTAP (CT arterial portography).

s = 1.5, and n = 2. We multiplied the mask images with the line-filtered images, thresholded the masked line-filtered images using an appropriate threshold value, and removed small connected components whose size was less than 10 voxels.

In Fig. 10.5(b), the left frame gives the rendered result of the original binary images, and the right frame shows a combination of the line-filtered binary images for small-vessel detection and the original binary images using relatively high threshold values for large-vessel detection. The two binary images were combined by taking the union of them. The CT data were scanned when the contrast material in the portal vein began to be absorbed by the liver tissues, as seen in the lower part of Fig. 10.5(b). Such a condition is quite common in CTAP for portal vein imaging. In the original images, the small vessels appear buried due to the contrast material absorbed by the liver tissue. In the combined result of the original and line-filtered images, not only is the nonuniformity of the contrast material canceled out, but also the recovery of small vessels is significantly improved over the entire liver area.

0 0

Post a comment