Removal of bone in CT angiography by multiscale matched mask bone elimination
Illustration of bone removal with the MMBE method on a region of interest (ROI) of of axial images of the petrous segment of the right internal carotid artery. Window center: 100 HU. Window width: 500 HU. (a) Registered nonenhanced image, (b) CTA image, and (c) CTA image after masking of the bone voxels. The borders of the adjacent vessels are masked as well.
Flow diagram of the multiscale MMBE method. Below the images the symbolic notations are denoted that are used in the different stages of multiscale MMBE (see Secs. II A–II D).
Construction of the bone mask from a sharply reconstructed nonenhanced image (a) on a ROI of . Window center: 60 HU. Window width: 400 HU. The sharp mask (b) is obtained by thresholding (a) at 150 HU, the smooth mask (c) is obtained by thresholding a blurred version of (a) at 150 HU, and the preliminary mask (d) is obtained by taking all voxels that masks (b) and (c) have in common. Image (e) shows the voxels that are present both in the high-decrease mask (not shown) and in the sharp mask (b). The final bone mask (f) is obtained by adding the voxels of (e) to those in (d).
Fourier transforms of the two reconstruction kernels (smooth, lower curve) and (sharp, upper curve).
A diagram of the three configurations of the phantom: (a) the PVC cube with water-filled holes, (b) the PVC cube filled with three POM cylinders, and (c) the POM cylinders in a small PMMA holder. This represents three different situations of vessels in a CTA examination: nonenhanced in bone (a), contrast-enhanced in bone (b), and contrast-enhanced in tissue (as reference) (c).
Relation of and the standard deviation of the noise after Gaussian blurring with a number of values of . The two reconstruction filters sharp and smooth are denoted with a ◆ and ▴ marker, respectively. The resulting and after blurring with different values of are denoted with markers. The standard deviation of the measurements are shown as horizontal en vertical bars.
The averaged intensity profiles through the cylinder (diameter 5.0 mm) orientated in the direction and scanned with collimation. The profile in the reference situation is the dashed line without markers. The profiles after application of multiscale MMBE without, with four voxels and with ten voxels of dilation are the solid lines with ◯, ▵, and markers, respectively. The profile after application of the original MMBE method is the solid black line with ◻ markers.
The effect of blurring and dilation illustrated with a ROI of an axial image of the petrous segment of the internal carotid artery processed by multiscale MMBE (left and middle column) and the original MMBE method (right column). Window center: 50 HU. Window width: 300 HU. The first column shows the masked sharp CTA image without (a) and with (d) dilation by four adjacent in-plane voxels. The second column shows the same images after blurring [(b) and (e)]. The third column shows the result of the original MMBE method without (c) and with dilation (f).
Coronal MIP images of a CTA examination of the circle of Willis. Window center: 50 HU. Window width: 300 HU. Bone was removed in (a) with multiscale MMBE without dilation, (b) with four voxels dilation and (c) with ten voxels dilation. Image (d) was processed with the original MMBE method. The ROI where the SD is measured to quantify the bone removal is shown in the lower left.
Images of a ROI of the left internal carotid artery from Fig. 9. Window center: 50 HU. Window width: 300 HU. Bone was removed with (a) multiscale MMBE without dilation, (b) with four voxels dilation, and (c) with ten voxels dilation. Image (d) was processed with the original MMBE method.
Sagittal and coronal MIP images of a CTA examination of the circle of Willis. Window center: 200 HU. Window width: 400 HU. Bone was removed with [(a) and (b)] multiscale MMBE with narrow collimation, [(c) and (d)] multiscale MMBE with deconvolution and [(e) and (f)] classic MMBE. The ROI where the SD is measured to quantify the bone removal (in the region of the nasal cavity) is shown in the lower right [(a), (c), and (e)].
Measured values of for the kernel and , for collimation before and after deconvolution and for collimation.
The estimated diameters of the POM cylinders (diameter 5.0 mm) after application of multiscale MMBE, multiscale MMBE with four and ten voxels dilation and original MMBE (with ten voxels dilation). For comparison the estimated reference diameters are shown in the last column.
The calculated width of the masked strip adjacent to bone in multiscale MMBE, multiscale MMBE with four and ten dilation, and in the original MMBE method.
Quality of bone removal in two patients quantified by the standard deviation in a ROI without vessels after multiscale MMBE, multiscale MMBE with four and ten voxels dilation, and the original MMBE method.
Article metrics loading...
Full text loading...