Related Experiment Videos
Intravascular ultrasound image segmentation: a three-dimensional fast-marching method based on gray level
Marie-Hélène Roy Cardinal1, Jean Meunier, Gilles Soulez
1Laboratory of Biorheology and Medical Ultrasonics, University of Montreal Hospital's Research Center, 2099 Alexandre de Sève, Montreal, QC H2L 2W5, Canada. roy-carmh@iro.umontreal.ca
IEEE Transactions on Medical Imaging
|May 13, 2006
Summary
A new 3D Intravascular Ultrasound (IVUS) segmentation model uses fast-marching and probability density functions (PDFs) for accurate vessel wall analysis. This method effectively processes complex IVUS data, improving atherosclerotic disease assessment.
Area of Science:
- Medical Imaging
- Biomedical Engineering
- Cardiovascular Research
Background:
- Intravascular ultrasound (IVUS) is crucial for studying atherosclerotic disease, providing detailed cross-sectional images of blood vessels.
- Quantitative assessment of vascular wall, lesion nature, plaque shape, and size are key applications of IVUS.
- Automatic processing of large IVUS datasets faces challenges from ultrasound speckle, catheter artifacts, and calcification shadows.
Purpose of the Study:
- To develop a novel 3D IVUS segmentation model for enhanced analysis of vascular structures.
- To address the challenges in automatic processing of IVUS data, particularly in complex or artifact-laden regions.
- To accurately segment lumen, intima plus plaque, and media layers of the vessel wall.
Main Methods:
- A new 3D IVUS segmentation model was developed, integrating the fast-marching method with gray level probability density functions (PDFs).
- The gray level distribution of IVUS data was modeled using a mixture of Rayleigh PDFs.
- Multiple interface fast-marching segmentation was employed to simultaneously compute lumen, intima plus plaque, and media layers.
Main Results:
- The PDF-based fast-marching method achieved accurate results on simulated IVUS data (average distance <0.072 mm).
- Applied to 9 in vivo IVUS pullbacks, the model showed good overall performance with average distances <0.16 mm compared to manual contours.
- Low Hausdorff distances (<0.40 mm) indicated robust performance even in regions with artifacts or missing information.
Conclusions:
- The developed segmentation model demonstrates significant potential for processing 3D IVUS images.
- Gray level PDF and fast-marching methods offer a promising approach for overcoming challenges in IVUS data analysis.
- This technique can improve the quantitative assessment of atherosclerotic disease using IVUS imaging.