Lung Nodule and Vessel Visualization from 3D CT Data

Multiscale blob and line filters were applied to chest CT images for nodule enhancement and vessel enhancement to detect early-stage lung cancers and visualize them with relation to peripheral vessels [33-35]. The CT dataset consisted of 60 slices with a 512 x 512 matrix (Fig. 10.7(a)); the pixel dimensions were 0.39 mm2. The slice thickness and reconstruction pitch were 2 mm and 1 mm, respectively. The matrix was reduced to half in the x^-plane, and thus the pixel interval was 0.78 mm. The data were then interpolated along the 2-axis using sinc interpolation so that the voxel was isotropic. While nodules, vessels, and other soft tissues have similar CT values in original images, the nodules and vessels have blob and line structures, respectively. Multiscale blob filtering was applied to the interpolated images using y23 = y12 = 0.5, a = 0.25, a1 = 2.0 pixels, s = V2, and n = 3. Multiscale line filtering was applied using y23 = y12 = 1.0, a = 0.25, a1 = 1.0 pixels, s = V2, and n = 3.

Figure 10(b) shows the color volume renderings of nodules (green), vessels (red), lung (violet), and bone tissues (white). In the left frame, the opacity and color functions were adjusted using only CT values of the original images. In the right frame, the original, blob, and line filtered images were used, where voxels having high intensities in the blob-filtered images were assigned as nodules (green), those having high in the line-filtered images as vessels (red), those having low in the original and two filtered images as lung tissues (violet), and those having high in the original but low in the two filtered images as bone tissues (white). The nodules and vessels were clearly depicted with different colors using blob and line enhancement filtering, while it was difficult to discriminate soft tissues into different categories using only original intensity values.

0 0

Post a comment