Limits...
Feasibility of Quantification of Intracranial Aneurysm Pulsation with 4D CTA with Manual and Computer-Aided Post-Processing

View Article: PubMed Central - PubMed

ABSTRACT

Background and purpose: The analysis of the pulsation of unruptured intracranial aneurysms might improve the assessment of their stability and risk of rupture. Pulsations can easily be concealed due to the small movements of the aneurysm wall, making post-processing highly demanding. We hypothesized that the quantification of aneurysm pulsation is technically feasible and can be improved by computer-aided post-processing.

Materials and methods: Images of 14 cerebral aneurysms were acquired with an ECG-triggered 4D CTA. Aneurysms were post-processed manually and computer-aided on a 3D model. Volume curves and random noise-curves were compared with the arterial pulse wave and volume curves were compared between both post-processing modalities.

Results: The aneurysm volume curves showed higher similarity with the pulse wave than the random curves (Hausdorff-distances 0.12 vs 0.25, p<0.01). Both post-processing methods did not differ in intra- (r = 0.45 vs r = 0.54, p>0.05) and inter-observer (r = 0.45 vs r = 0.54, p>0.05) reliability. Time needed for segmentation was significantly reduced in the computer-aided group (3.9 ± 1.8 min vs 20.8 ± 7.8 min, p<0.01).

Conclusion: Our results show pulsatile changes in a subset of the studied aneurysms with the final prove of underlying volume changes remaining unsettled. Semi-automatic post-processing significantly reduces post-processing time but cannot yet replace manual segmentation.

No MeSH data available.


Related in: MedlinePlus

Workflow on the 3D+t model: On axial images the vasculature is defined with a threshold (green).To reduce computation time for the 3D+t model, the volume for post-processing is reduced by placing a VOI over the aneurysm (A). The resulting 3D+t model can be rotated, translated and zoomed (B). Segmentation points are placed on the aneurysms neck (C) and the aneurysm (green) is separated from the parent vessel (red) and the volume and surface is calculated (D).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5120820&req=5

pone.0166810.g001: Workflow on the 3D+t model: On axial images the vasculature is defined with a threshold (green).To reduce computation time for the 3D+t model, the volume for post-processing is reduced by placing a VOI over the aneurysm (A). The resulting 3D+t model can be rotated, translated and zoomed (B). Segmentation points are placed on the aneurysms neck (C) and the aneurysm (green) is separated from the parent vessel (red) and the volume and surface is calculated (D).

Mentions: All measurements were conducted by the same neuroradiologist with 8 years of experience (M.B.) with the in-house developed post-processing software AnToNIa (Analysis Tool for Neuroimaging Data) and a delay of three weeks [14–16]. As in the manual segmentation, the vasculature was segmented from the cerebrospinal fluid with a threshold of 160 to 890 HU. Based on this segmentation, a single surface model was generated for every time-point using the marching cube algorithm (3D+t model). The surface model can be rotated, moved and zoomed to enable the best view on relevant details. To segment the aneurysm from its parent vessel, segmentation points were manually placed on the 3D+t model at time point zero. These initial segmentation points were automatically transferred to all other time points using an iterative closest point approach and could be corrected manually (Fig 1). For each time point, the intraluminal segmentation plane was calculated by automatically connecting these manually placed segmentation points. The final segmentation was displayed on axial, coronal and sagittal planes. The time needed for segmentation was recorded.


Feasibility of Quantification of Intracranial Aneurysm Pulsation with 4D CTA with Manual and Computer-Aided Post-Processing
Workflow on the 3D+t model: On axial images the vasculature is defined with a threshold (green).To reduce computation time for the 3D+t model, the volume for post-processing is reduced by placing a VOI over the aneurysm (A). The resulting 3D+t model can be rotated, translated and zoomed (B). Segmentation points are placed on the aneurysms neck (C) and the aneurysm (green) is separated from the parent vessel (red) and the volume and surface is calculated (D).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5120820&req=5

pone.0166810.g001: Workflow on the 3D+t model: On axial images the vasculature is defined with a threshold (green).To reduce computation time for the 3D+t model, the volume for post-processing is reduced by placing a VOI over the aneurysm (A). The resulting 3D+t model can be rotated, translated and zoomed (B). Segmentation points are placed on the aneurysms neck (C) and the aneurysm (green) is separated from the parent vessel (red) and the volume and surface is calculated (D).
Mentions: All measurements were conducted by the same neuroradiologist with 8 years of experience (M.B.) with the in-house developed post-processing software AnToNIa (Analysis Tool for Neuroimaging Data) and a delay of three weeks [14–16]. As in the manual segmentation, the vasculature was segmented from the cerebrospinal fluid with a threshold of 160 to 890 HU. Based on this segmentation, a single surface model was generated for every time-point using the marching cube algorithm (3D+t model). The surface model can be rotated, moved and zoomed to enable the best view on relevant details. To segment the aneurysm from its parent vessel, segmentation points were manually placed on the 3D+t model at time point zero. These initial segmentation points were automatically transferred to all other time points using an iterative closest point approach and could be corrected manually (Fig 1). For each time point, the intraluminal segmentation plane was calculated by automatically connecting these manually placed segmentation points. The final segmentation was displayed on axial, coronal and sagittal planes. The time needed for segmentation was recorded.

View Article: PubMed Central - PubMed

ABSTRACT

Background and purpose: The analysis of the pulsation of unruptured intracranial aneurysms might improve the assessment of their stability and risk of rupture. Pulsations can easily be concealed due to the small movements of the aneurysm wall, making post-processing highly demanding. We hypothesized that the quantification of aneurysm pulsation is technically feasible and can be improved by computer-aided post-processing.

Materials and methods: Images of 14 cerebral aneurysms were acquired with an ECG-triggered 4D CTA. Aneurysms were post-processed manually and computer-aided on a 3D model. Volume curves and random noise-curves were compared with the arterial pulse wave and volume curves were compared between both post-processing modalities.

Results: The aneurysm volume curves showed higher similarity with the pulse wave than the random curves (Hausdorff-distances 0.12 vs 0.25, p<0.01). Both post-processing methods did not differ in intra- (r = 0.45 vs r = 0.54, p>0.05) and inter-observer (r = 0.45 vs r = 0.54, p>0.05) reliability. Time needed for segmentation was significantly reduced in the computer-aided group (3.9 ± 1.8 min vs 20.8 ± 7.8 min, p<0.01).

Conclusion: Our results show pulsatile changes in a subset of the studied aneurysms with the final prove of underlying volume changes remaining unsettled. Semi-automatic post-processing significantly reduces post-processing time but cannot yet replace manual segmentation.

No MeSH data available.


Related in: MedlinePlus