Segmentation of the common carotid artery with active shape models from 3D ultrasound images

Abstract

Carotid atherosclerosis is a major cause of stroke, a leading cause of death and disability. In this paper, we develop and evaluate a new segmentation method for outlining both lumen and adventitia (inner and outer walls) of common carotid artery (CCA) from three-dimensional ultrasound (3D US) images for carotid atherosclerosis diagnosis and evaluation. The data set consists of sixty-eight, 17 × 2 × 2, 3D US volume data acquired from the left and right carotid arteries of seventeen patients (eight treated with 80mg atorvastain and nine with placebo), who had carotid stenosis of 60% or more, at baseline and after three months of treatment. We investigate the use of Active Shape Models (ASMs) to segment CCA inner and outer walls after statin therapy. The proposed method was evaluated with respect to expert manually outlined boundaries as a surrogate for ground truth. For the lumen and adventitia segmentations, respectively, the algorithm yielded Dice Similarity Coefficient (DSC) of 93.6% ± 2.6%, 91.8% ± 3.5%, mean absolute distances (MAD) of 0.28 ± 0.17mm and 0.34 ± 0.19mm, maximum absolute distances (MAXD) of 0.87 ± 0.37mm and 0.74 ± 0.49mm. The proposed algorithm took 4.4 ± 0.6min to segment a single 3D US images, compared to 11.7 ± 1.2min for manual segmentation. Therefore, the method would promote the translation of carotid 3D US to clinical care for the fast, safety and economical monitoring of the atherosclerotic disease progression and regression during therapy.

Bimodal Automated Carotid Ultrasound Segmentation Using Geometrically Constrained Deep Neural Networks

“…The results noted in Table III show a comparative assessment of performance between our technique and the techniques of other studies whose work similarly addressed the problem of MAB and LIB segmentation. Table III shows that our average results yield similar performance to that of Ukwatta et al in their study in [23] and [10], and superior performance to that of Yang et al in [16] and that of Hossain et al in their work of [9]. In their work, Ukwatta et al propose a semiautomated technique based on a level-set method to segment the MA and LI interfaces, whereby the operator was asked to provide anchor points as high-level domain knowledge.…”

“…Furthermore, we demonstrate that albeit the technique being fully automated and having a larger and more generalisable dataset, it retains a good performance of 94% and 91% for MAB and LIB borders respectively, and that this retains comparable performance to techniques from other studies which test on a much smaller dataset. [10] 0.954 ± 0.016 -± -0.931 ± 0.031 -± -Yang et al [16] 0.918 ± 0.035 -± -0.936 ± 0.026 -± -Ukwatta et al [23] 0.95 ± 0.017 -± -0.92 ± 0.042 -± -Hossain et al [9] 0.915 ± 0.035 0.25 ± -0.735 ± 0.169 0.25 ± -Azzopardi et al [28] 0.988 ± 0.035 0.05 ± --± --± -This study 0.940 ± 0.040 0.24 ± 0.16 0.910 ± 0.070 0.33 ± 0.33…”

“…In 2009, Seabra et al [15] proposed a semi-automatic technique for plaque segmentation in transverse images based on a manually initialised 2D active contour algorithm. Another study by Yang et al [16] proposed to use active shape models to segment both the MA and LI interfaces. Ukwatta et al [10] proposed a novel semi-automated technique based on a level-set method to segment the MA and LI interfaces.…”

“…One notes further that, particularly in the case of transverse segmentation, studies have aimed to segment either the LI interface alone [7], [11], [17], [20]- [22], or else both the LI and the MA interfaces [9], [10], [16], [23]. The latter approach has increasingly gained interest, due to new volumetric parameters such as vessel wall volume (VWV) and total plaque volume (TPV) which have been proposed to characterise plaque burden [24].…”

Three‐dimensional segmentation of three‐dimensional ultrasound carotid atherosclerosis using sparse field level sets

“…Although, VWV has been shown to be more sensitive to temporal changes in carotid atherosclerosis than the more widely used intima-media thickness (IMT), 13 3DUS VWV is not used because manual segmentation of the carotid MAB and LIB is tedious and timeconsuming. Currently, VWV measurements are mainly generated using manual segmentation, [8][9][10] and there have been a few attempts 11,14,15 to develop and use semiautomated algorithms. Previously developed semiautomated algorithms 14,16 required user interactions on every individual slice.…”