Zhou et al. [108]
|
Modified DP
|
N=200 (32)
videos
|
Overall error,
Total time,
Bland-Altman plots
|
Accuracy improved when the algorithm was refined using snake.
|
Faita et al. [109]
|
Gradient-based edge detection
|
N=150
|
Bland-Altman plots
|
variation in mean bias ± SD of 0.001 ± 0.035.
High accuracy and real-time application
|
Rossi et al. [110]
|
Adventitial delineation using sustain attack filter, intimal delineation
using MAB.
|
N=36 (12)
|
Bland-Altman plots,
Measurement error
|
Intra-observer variability = 0, suitable for clinical trials, comparison
with synthetic ultrasound images. Radiofrequency envelopes
analysed.
|
Golemati et al. [112]
|
Hough transform,
Canny edge detection. Both B-mode and M-mode ultrasound
|
N=5
|
Radial displacement of ROI
|
Accurate for non-stenotic.
Atherosclerotic plaque affected the result.
|
Loizou et al. [113]
|
Snake algorithm
Speckle reduction
|
N=100
|
Bland-Altman plots
|
intra-observer error = 0.08 Hausdorff distance = 5.2
|
Petroudi et al. [90]
|
Active contour,
Speckle removal
|
N=100
|
Mean absolute distance. Polyline distance. Hausdorff distance.
|
Mean absolute distance error = 0.095±0.0615 mm, Polyline distance =
0.096±0.034 mm.
Hausdorff distance = 0.176±0.047 mm.
|
Santhiyakumari et al. [114]
|
Active contour segmentation. Semi-automatic ROI identification
|
N=100
63 normal
|
Coefficient of variation, Pearson’s CC, Wilcoxon metric
|
Inter-method error = 0.09 mm
CV = 18.9%
|
Destrempes et al. [115]
|
Expectation maximisation algorithm, Nakagami distribution
|
N=30
|
Mean distance,
Hausdorff distance
|
Error in:
LI = 0.46 mm
MA = 0.41 mm
|
Molinari et al. [117]
|
Local statistics.
Integrated approach (Greedy algorithm).
|
N=200
|
Polyline distance,
Mean system error
|
Error in:
LI = 26.3±55.6 µm
MA = 16.2±31.3 µm
IMT = 83.1±61.8 µm
|
Ilea et al. [119]
|
Unsupervised
IMC video segmentation
|
40 and 772 frames
|
coefficient of variation, Bland-Altman plot
|
Auto tracking of IMT variations in a cardiac cycle.
|