Three-dimensional reconstruction of blood vessel lumen from a single x-ray projection image
MetadataShow full item record
Stroke and heart attack are the leading causes of death and morbidity in developed countries. The underlying cause is the occlusion of blood vessel lumen by atherosclerotic plaque. Vascular intervention is the dominant method used for treating structural abnormalities in the blood vessels. The success of these minimally invasive procedures depends on accurate determination of lumen and plaque characteristics (e.g., length of diseased vessel segment, percentage luminal narrowing). Clinicians often obtain this information qualitatively from projection images of the blood vessels (angiograms). To provide more quantitative information, other investigators have generated 3D lumen and plaque distributions from two or more views. In this work, we present a novel method for three-dimensional (3D) reconstruction of blood vessel lumens deformed by atherosclerotic plaque from a single x-ray projection image. This 3D-from-a-single-view technique has several key components. We have developed a novel method to estimate the constant of proportionality relating x-ray path lengths inside the contrast-filled vessel lumen and the pixel values of the x-ray projection image. We used a model-based technique to estimate lumen (pixel value) profiles prior to and after plaque formation, and thereby extract quantitative information regarding the plaque distribution in the vessel. The proportionality constant is then used together with the profile data to generate lumen cross sections. To generate a 3D centerline from a single view, we take advantage of the change in x-ray path length with angle. The angle between vessel medial axis and imaging plane is thereby computed along the indicated segment, and a 3D centerline is constructed. The generated cross sections are positioned on the centerline, and a 3D model is formed. Intravascular ultrasound (IVUS) imaging is the gold-standard for determination of vessel lumen and plaque size. We have developed a method for generation of 3D vessel lumen from IVUS image sequences based on lumen boundaries indicated in the IVUS images. Our method incorporates techniques to correct for various discrepancies present in the original 3D lumen. The 3D models reconstructed from single-view angiograms compared very well with 3D models reconstructed from IVUS image sequences. Using this approach, interventionalists will have quantitative 3D plaque distributions for their decision process. These data will also be useful for treatment assessment and patient follow-up. Since the 3D model can be reconstructed from a single image acquired from any view, the required image acquisition/processing time and radiation dose are significantly reduced compared to other techniques that utilize x-rays.