Sifakis et al. [98]
|
Vertical intensity profile based signal selection, segment filtering,
Statistical methods for lumen centre point identification
|
N = 2149 (100)
|
SR =100%
Mean FC = 95.76±9.61%
Error = 0.43±0.26 mm
|
CCA recognition in: poor image quality, presence of plaque, jugular
vein, intensity variations and moderate arterial curvature
|
Araki et al. [88]
|
Spectral analysis for peak detection,
K-means classifier
|
N = 404 (202)
|
DS ~1, JI=92.1
Mann-Whitney U= 63,356.5
Mean error, PoM
Cross-correlation
SSI
|
Segmented result close to GT. Consistent LD assessment. Automated and
manual results are consistent if SSI>40%. Susceptible to
noise, motion artefacts and noise.
|
Molinari et al. [99]
|
Geometric feature extraction,
Line fitting, classification
|
N = 200
|
mean distance errors ± SD in near-wall = 1.05 ± 1.04 pixels in far wall
2.68 ± 3.94 pixels
|
Automatic delineation of CCA.
|
Kumar et al. [100]
|
Vertical spectral analysis to trace adventitial border
Spatial transformation to straighten curved vessels. Scale-space
segmentation.
|
N = 404
(202)
|
DS and JI were 94.2 (89.1) and 93.9 (88.6).
Mean LD error = 0.27±0.25 mm
Mean IAD error = 0.24±0.24 mm
|
Delineation lumen and adventitial borders even in curved vessels.
Validated with expert readings
|
Carvalho et al. [95]
|
Centreline estimation,
Graph-based segmentation using DP.
|
N = 21
(17): 2 datasets of size 11 and 10
|
Dataset 1(RMSE = 191±43 µm) and Dataset 2 (RMSE 351 ± 176 µm)
carotids.
|
Both contrast enhanced-US and B-mode were used for lumen
segmentation.
|
Rocha et al. [101]
|
Auto ROI,
DP to delineate longitudinal paths, Linear Bayes classifier.
|
N=199 Two datasets
|
SR=99.5%
|
Real-time processing, fully automated.
|