Visualizing and analyzing the morphological structure of carotid bifurcations are important for understanding the etiology of carotid atherosclerosis, which is a major cause of stroke and transient ischemic attack. For delineation of vasculatures in the carotid artery, ultrasound examinations have been widely employed because of the non-invasive procedure without ionizing radiation. However, conventional 2D ultrasound imaging has technical limitations in observing the complicated 3D shapes and asymmetric vasodilation of bifurcations.
The 3D ultrasound imaging system has been employed in medical diagnosis and treatment because it provides comprehensive spatial information on soft tissues and organs. In a 3D ultrasound system, 3D volume information is usually acquired by scanning with a transducer and tracking the position and orientation of a transducer over the regions of interest (ROIs).
Recently, 3D ultrasound imaging system was employed to assess the progression of atherosclerotic plaques. However, the temporal resolution should be further improved for precise visualization of 3D vessels. In order to improve 3D visualization of vessels, several 3D vessel segmentation techniques using multi-slice images have been introduced.
Among these reconstruction techniques, considerable efforts went into the segmentation of ultrasound images. However, manual segmentation requires highly intensive labor and the results are operator dependent. As an automated segmentation technique, a gradient vector flow (GVF) algorithm has been widely used to detect the boundaries of carotid artery. However, the GVF method should set the initial contour close to the boundary for more precise detection.
Weak intensity gradients caused by acoustic artifacts also considerably restrict the segmentation process for 3D reconstruction. In order to compensate the weak intensity gradients caused by acoustic artifacts, several 3D semi-automated algorithms were proposed.
There was a recent study supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. 2008-0061991) which aims to propose image-processing techniques for better 3D reconstruction of a carotid bifurcation in a rat by using 2D cross-sectional ultrasound images. Recently, it is possible to monitor blood vessels in small animal models such as rats and mice with the high spatial resolution using commercially available HFUS systems.
The main objective of this study is to develop digital image-processing techniques for the easy 3D reconstruction of a rat carotid bifurcation by using multi-slice images acquired by a HFUS system. The arterial lumen boundaries in 2D sectional images were segmented by employing three image-processing methods. The first method detects a lumen boundary by fitting ultrasound images as an ellipsoidal shape. The second method applies the thresholding and edge detection methods to correlation maps that represent the degree of blood decorrelation. The final method adopts the ellipse-fitting method to the correlation maps.
Segmented lumen boundaries were depicted in the xyz coordinate, and the cross-sectional areas and relative diameters were measured and statistically analyzed for evaluating the segmentation performance of the three methods. To reconstruct the 3D geometry of the carotid bifurcation, the arterial lumen boundaries in 2D sectional images were processed by additional post processing such as smoothing and 3D rendering.
A high-resolution ultrasound imaging system with a probe centered at 40 MHz was employed to obtain 2D transversal images. The lumen boundaries in each transverse ultrasound image were detected by using three different techniques; an ellipse-fitting, a correlation mapping to visualize the decorrelation of blood flow, and the ellipse-fitting on the correlation map. When the results are compared, the third technique provides relatively good boundary extraction.
The incomplete boundaries of arterial lumen caused by acoustic artifacts are somewhat resolved by adopting the correlation mapping and the distortion in the boundary detection near the bifurcation apex was largely reduced by using the ellipse-fitting technique. The 3D lumen geometry of a carotid artery was obtained by volumetric rendering of several 2D slices. For the 3D vasodilatation of the carotid bifurcation, lumen geometries at the contraction and expansion states were simultaneously depicted at various view angles. The present 3D reconstruction methods would be useful for efficient extraction and construction of the 3D lumen geometries of carotid bifurcations from 2D ultrasound images.
Sources:
- Yeom et al. / Ultrasonics 54 (2014) 2184–2192
- Seabra, J. Sanches, L. Pedro, J. F. Fernandes, Three-dimensional ultrasonic assessment of atherosclerotic plaques, in: Iberian Conf. Patt. Recognition and Image Analysis, 2007, pp. 523–531.
- Stoitsis, S. Golemati, S. Kendros, K.S. Nikita, Automated detection of the carotid artery wall in B-mode ultrasound images using active contours initialized by the Hough transform, in: 30th Annual International IEEE EMBS Conference, 2008, pp. 3146–3149.