Limits...
Semiautomatic, Quantitative Measurement of Aortic Valve Area Using CTA: Validation and Comparison with Transthoracic Echocardiography.

Tuncay V, Prakken N, van Ooijen PM, Budde RP, Leiner T, Oudkerk M - Biomed Res Int (2015)

Bottom Line: Time required for manual and semiautomatic segmentations was recorded.Intra- and interobserver variability were 8.4 ± 7.1% and 27.6 ± 16.0% for manual, and 5.8 ± 4.5% and 16.8 ± 12.7% for semiautomatic measurements, respectively.Newly developed semiautomatic segmentation provides an accurate, more reproducible, and faster AVA segmentation result.

View Article: PubMed Central - PubMed

Affiliation: Center for Medical Imaging North East Netherlands (CMI-NEN), Department of Radiology, University of Groningen, University Medical Center Groningen, P.O. Box 30001, 9700 RB Groningen, Netherlands.

ABSTRACT

Objective: The aim of this work was to develop a fast and robust (semi)automatic segmentation technique of the aortic valve area (AVA) MDCT datasets.

Methods: The algorithm starts with detection and cropping of Sinus of Valsalva on MPR image. The cropped image is then binarized and seed points are manually selected to create an initial contour. The contour moves automatically towards the edge of aortic AVA to obtain a segmentation of the AVA. AVA was segmented semiautomatically and manually by two observers in multiphase cardiac CT scans of 25 patients. Validation of the algorithm was obtained by comparing to Transthoracic Echocardiography (TTE). Intra- and interobserver variability were calculated by relative differences. Differences between TTE and MDCT manual and semiautomatic measurements were assessed by Bland-Altman analysis. Time required for manual and semiautomatic segmentations was recorded.

Results: Mean differences from TTE were -0.19 (95% CI: -0.74 to 0.34) cm(2) for manual and -0.10 (95% CI: -0.45 to 0.25) cm(2) for semiautomatic measurements. Intra- and interobserver variability were 8.4 ± 7.1% and 27.6 ± 16.0% for manual, and 5.8 ± 4.5% and 16.8 ± 12.7% for semiautomatic measurements, respectively.

Conclusion: Newly developed semiautomatic segmentation provides an accurate, more reproducible, and faster AVA segmentation result.

No MeSH data available.


The flowchart of the AVA segmentation algorithm.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4499628&req=5

fig3: The flowchart of the AVA segmentation algorithm.

Mentions: The flowchart in Figure 3 shows the details of the segmentation of the AVA. The main segmentation tool is the gradient vector flow (GVF) snake [29]. The snake algorithm is an active contour, which moves to the edges of the object in order to reach the boundaries of the object. In the binarized images (described above) the edges are clearer and the active contour can move towards the object boundaries easier compared to grayscale images. The user places three seed points on the grayscale image where the cusps are connected to each other to create the initial contour for the GVF snake. This initial contour creates a mask image, which is used on the binarized image. The active contour shrinks to cover the AVA region. However, the GVF snake result overestimates the AVA. Therefore this GVF snake contour is used to mask the image again. In case the resulting double-masked image contains more than 1 object, the size and location of these objects were determined. Objects smaller than 40 mm2 and the object most distant from the center were removed. The remaining object was identified as the AVA.


Semiautomatic, Quantitative Measurement of Aortic Valve Area Using CTA: Validation and Comparison with Transthoracic Echocardiography.

Tuncay V, Prakken N, van Ooijen PM, Budde RP, Leiner T, Oudkerk M - Biomed Res Int (2015)

The flowchart of the AVA segmentation algorithm.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4499628&req=5

fig3: The flowchart of the AVA segmentation algorithm.
Mentions: The flowchart in Figure 3 shows the details of the segmentation of the AVA. The main segmentation tool is the gradient vector flow (GVF) snake [29]. The snake algorithm is an active contour, which moves to the edges of the object in order to reach the boundaries of the object. In the binarized images (described above) the edges are clearer and the active contour can move towards the object boundaries easier compared to grayscale images. The user places three seed points on the grayscale image where the cusps are connected to each other to create the initial contour for the GVF snake. This initial contour creates a mask image, which is used on the binarized image. The active contour shrinks to cover the AVA region. However, the GVF snake result overestimates the AVA. Therefore this GVF snake contour is used to mask the image again. In case the resulting double-masked image contains more than 1 object, the size and location of these objects were determined. Objects smaller than 40 mm2 and the object most distant from the center were removed. The remaining object was identified as the AVA.

Bottom Line: Time required for manual and semiautomatic segmentations was recorded.Intra- and interobserver variability were 8.4 ± 7.1% and 27.6 ± 16.0% for manual, and 5.8 ± 4.5% and 16.8 ± 12.7% for semiautomatic measurements, respectively.Newly developed semiautomatic segmentation provides an accurate, more reproducible, and faster AVA segmentation result.

View Article: PubMed Central - PubMed

Affiliation: Center for Medical Imaging North East Netherlands (CMI-NEN), Department of Radiology, University of Groningen, University Medical Center Groningen, P.O. Box 30001, 9700 RB Groningen, Netherlands.

ABSTRACT

Objective: The aim of this work was to develop a fast and robust (semi)automatic segmentation technique of the aortic valve area (AVA) MDCT datasets.

Methods: The algorithm starts with detection and cropping of Sinus of Valsalva on MPR image. The cropped image is then binarized and seed points are manually selected to create an initial contour. The contour moves automatically towards the edge of aortic AVA to obtain a segmentation of the AVA. AVA was segmented semiautomatically and manually by two observers in multiphase cardiac CT scans of 25 patients. Validation of the algorithm was obtained by comparing to Transthoracic Echocardiography (TTE). Intra- and interobserver variability were calculated by relative differences. Differences between TTE and MDCT manual and semiautomatic measurements were assessed by Bland-Altman analysis. Time required for manual and semiautomatic segmentations was recorded.

Results: Mean differences from TTE were -0.19 (95% CI: -0.74 to 0.34) cm(2) for manual and -0.10 (95% CI: -0.45 to 0.25) cm(2) for semiautomatic measurements. Intra- and interobserver variability were 8.4 ± 7.1% and 27.6 ± 16.0% for manual, and 5.8 ± 4.5% and 16.8 ± 12.7% for semiautomatic measurements, respectively.

Conclusion: Newly developed semiautomatic segmentation provides an accurate, more reproducible, and faster AVA segmentation result.

No MeSH data available.