Limits...
Segmentation of Pulmonary Vascular Trees from Thoracic 3D CT Images.

Shikata H, McLennan G, Hoffman EA, Sonka M - Int J Biomed Imaging (2009)

Bottom Line: A quantitative validation was performed with more than 1000 manually identified points selected from inside the vessel segments to assess true positives (TPs) and 1000 points randomly placed outside of the vessels to evaluate false positives (FPs) in each case.On average, for both the high and low volume lung images, 99% of the points was properly marked as vessel and 1% of the points were assessed as FPs.Our hybrid segmentation algorithm provides a highly reliable method of segmenting the combined pulmonary venous and arterial trees which in turn will serve as a critical starting point for further quantitative analysis tasks and aid in our overall goal of establishing a normative atlas of the human lung.

View Article: PubMed Central - PubMed

Affiliation: Iowa Institute for Biomedical Imaging, The University of Iowa, Iowa City, IA 52242, USA.

ABSTRACT
This paper describes an algorithm for extracting pulmonary vascular trees (arteries plus veins) from three-dimensional (3D) thoracic computed tomographic (CT) images. The algorithm integrates tube enhancement filter and traversal approaches which are based on eigenvalues and eigenvectors of a Hessian matrix to extract thin peripheral segments as well as thick vessels close to the lung hilum. The resultant algorithm was applied to a simulation data set and 44 scans from 22 human subjects imaged via multidetector-row CT (MDCT) during breath holds at 85% and 20% of their vital capacity. A quantitative validation was performed with more than 1000 manually identified points selected from inside the vessel segments to assess true positives (TPs) and 1000 points randomly placed outside of the vessels to evaluate false positives (FPs) in each case. On average, for both the high and low volume lung images, 99% of the points was properly marked as vessel and 1% of the points were assessed as FPs. Our hybrid segmentation algorithm provides a highly reliable method of segmenting the combined pulmonary venous and arterial trees which in turn will serve as a critical starting point for further quantitative analysis tasks and aid in our overall goal of establishing a normative atlas of the human lung.

No MeSH data available.


Related in: MedlinePlus

Visualization of the phantom data. (a) Surface display of the simulated vascular tree. (b) Cross-section of a phantom subjected to Gaussian noise of standard deviation σ = 30. The model contains 62 branchpoints and 125 segments. Intensity values at the center of the segments vary depending on the radius of the segment to mimic vessels in typical thoracic CT images. Gaussian noise with different standard deviations was added.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2801012&req=5

fig5: Visualization of the phantom data. (a) Surface display of the simulated vascular tree. (b) Cross-section of a phantom subjected to Gaussian noise of standard deviation σ = 30. The model contains 62 branchpoints and 125 segments. Intensity values at the center of the segments vary depending on the radius of the segment to mimic vessels in typical thoracic CT images. Gaussian noise with different standard deviations was added.

Mentions: The algorithm was first applied to a set of computer-generated tree structures. Figure 5 shows the surface model of the tree structure and a cross-sectional image of one of the noisy phantom instances. This model was originally developed to represent an airway tree [20], yet it is appropriate also for use as a model for simulating a pulmonary vascular tree since the pulmonary arterial tree follows the airway tree out into the lung periphery and thus has the same general geometric relationships as the airway tree. The model contained 62 branchpoints and 125 segments. Each segment was characterized by a starting point, an end point and the associated radius. Branching angles were different for each bifurcation and their average was 83.9 degrees. The segments become thinner every time they bifurcate. 3D Gaussian spheres were moved along the linear segments to generate the tree structure in a 3D image whose dimension was 250 × 250 × 250 voxels. The radius of a segment decreases from the starting point to the end point of a segment so that it becomes the same as the radius of a child segment in order to allow smooth connection. Standard deviation of the 3D Gaussian spheres σr changes as the radius varies. In this phantom setting, maximum σr was six voxels and minimum σr was approximately one voxel in the volume. We also rotated the model in 11 different angles in 3D space and generated the structure in 3D images. Intensity value at the center of a segment mildly decreased from the starting point of a segment toward its end point along the tangent direction. This made thick segments brighter compared to the thin segments. The thickest segment had a value of −200 [H.U.] at the center whereas the thinnest had a value of −700 [H.U.] at the center. The background value was −900 [H.U.] to simulate a typical background intensity in the parenchymal region in lung CT images. Gaussian noise with a standard deviation σ = 20, 30, 40 was added to each image. After the result was obtained, a thinning method [19] was applied to obtain its graph representation so as to evaluate how many branches were correctly extracted and how many false branches were extracted by the segmentation algorithm. Both missing and extra branches were counted manually.


Segmentation of Pulmonary Vascular Trees from Thoracic 3D CT Images.

Shikata H, McLennan G, Hoffman EA, Sonka M - Int J Biomed Imaging (2009)

Visualization of the phantom data. (a) Surface display of the simulated vascular tree. (b) Cross-section of a phantom subjected to Gaussian noise of standard deviation σ = 30. The model contains 62 branchpoints and 125 segments. Intensity values at the center of the segments vary depending on the radius of the segment to mimic vessels in typical thoracic CT images. Gaussian noise with different standard deviations was added.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2801012&req=5

fig5: Visualization of the phantom data. (a) Surface display of the simulated vascular tree. (b) Cross-section of a phantom subjected to Gaussian noise of standard deviation σ = 30. The model contains 62 branchpoints and 125 segments. Intensity values at the center of the segments vary depending on the radius of the segment to mimic vessels in typical thoracic CT images. Gaussian noise with different standard deviations was added.
Mentions: The algorithm was first applied to a set of computer-generated tree structures. Figure 5 shows the surface model of the tree structure and a cross-sectional image of one of the noisy phantom instances. This model was originally developed to represent an airway tree [20], yet it is appropriate also for use as a model for simulating a pulmonary vascular tree since the pulmonary arterial tree follows the airway tree out into the lung periphery and thus has the same general geometric relationships as the airway tree. The model contained 62 branchpoints and 125 segments. Each segment was characterized by a starting point, an end point and the associated radius. Branching angles were different for each bifurcation and their average was 83.9 degrees. The segments become thinner every time they bifurcate. 3D Gaussian spheres were moved along the linear segments to generate the tree structure in a 3D image whose dimension was 250 × 250 × 250 voxels. The radius of a segment decreases from the starting point to the end point of a segment so that it becomes the same as the radius of a child segment in order to allow smooth connection. Standard deviation of the 3D Gaussian spheres σr changes as the radius varies. In this phantom setting, maximum σr was six voxels and minimum σr was approximately one voxel in the volume. We also rotated the model in 11 different angles in 3D space and generated the structure in 3D images. Intensity value at the center of a segment mildly decreased from the starting point of a segment toward its end point along the tangent direction. This made thick segments brighter compared to the thin segments. The thickest segment had a value of −200 [H.U.] at the center whereas the thinnest had a value of −700 [H.U.] at the center. The background value was −900 [H.U.] to simulate a typical background intensity in the parenchymal region in lung CT images. Gaussian noise with a standard deviation σ = 20, 30, 40 was added to each image. After the result was obtained, a thinning method [19] was applied to obtain its graph representation so as to evaluate how many branches were correctly extracted and how many false branches were extracted by the segmentation algorithm. Both missing and extra branches were counted manually.

Bottom Line: A quantitative validation was performed with more than 1000 manually identified points selected from inside the vessel segments to assess true positives (TPs) and 1000 points randomly placed outside of the vessels to evaluate false positives (FPs) in each case.On average, for both the high and low volume lung images, 99% of the points was properly marked as vessel and 1% of the points were assessed as FPs.Our hybrid segmentation algorithm provides a highly reliable method of segmenting the combined pulmonary venous and arterial trees which in turn will serve as a critical starting point for further quantitative analysis tasks and aid in our overall goal of establishing a normative atlas of the human lung.

View Article: PubMed Central - PubMed

Affiliation: Iowa Institute for Biomedical Imaging, The University of Iowa, Iowa City, IA 52242, USA.

ABSTRACT
This paper describes an algorithm for extracting pulmonary vascular trees (arteries plus veins) from three-dimensional (3D) thoracic computed tomographic (CT) images. The algorithm integrates tube enhancement filter and traversal approaches which are based on eigenvalues and eigenvectors of a Hessian matrix to extract thin peripheral segments as well as thick vessels close to the lung hilum. The resultant algorithm was applied to a simulation data set and 44 scans from 22 human subjects imaged via multidetector-row CT (MDCT) during breath holds at 85% and 20% of their vital capacity. A quantitative validation was performed with more than 1000 manually identified points selected from inside the vessel segments to assess true positives (TPs) and 1000 points randomly placed outside of the vessels to evaluate false positives (FPs) in each case. On average, for both the high and low volume lung images, 99% of the points was properly marked as vessel and 1% of the points were assessed as FPs. Our hybrid segmentation algorithm provides a highly reliable method of segmenting the combined pulmonary venous and arterial trees which in turn will serve as a critical starting point for further quantitative analysis tasks and aid in our overall goal of establishing a normative atlas of the human lung.

No MeSH data available.


Related in: MedlinePlus