41
(a) (b)
(c) (d)
(e) (f)
Proximal wall
Distal wall Lumen axis
Transformed ROI Adventitial borders
LD bordersBinary lumen
42
(g) (h)
Fig. 4: Illustrating the algorithmic steps on a single image having curvature. (a) Curved ROI, (b)
Lumen axis on ROI, (c) Transformed ROI, (d) Adventitial borders on the transformed ROI, (e)
Binary lumen from classifier, (f) LD borders on transformed ROI, (g) LD borders on inverse
transformed ROI, (h) LD borders on original ROI.
LD bordersLD borders
43
(a)
(b)
(c)
Reader
Reader
Reader
Reader
Reader
Auto SS
Auto SS
Auto SST
Auto SST
Auto SS Auto SST
Reader
44
(d)
Fig. 5: Carotid Auto LD borders compared against the manual tracings on the grayscale images
of four patients for both simple scale-space and transformation based iterative methods. Carotid
Auto LD borders are shown in solid white; while manual LD borders are shown in dashed white
(Auto SS – Automated Simple scale-space, Auto SST – Automated scale-space with
Transformation, Reader – Manual Reader).
Reader
Auto SS Auto SST
Reader
45
(a) (b)
(c) (d)
Fig. 6: Carotid Auto LD/IAD borders drawn on the gray scale images of patients having jugular
vein interference (a) and thick plaque can be seen in (b). (a) Carotid Auto LD borders (solid
white) Vs. Manual-1 LD borders (dashed white); (b) Auto LD borders (solid white) Vs. Manual-
2 LD borders (dashed white); (c) Auto IAD borders (solid black) Vs. Manual-1 IAD borders
(dashed black); (d) Auto IAD borders (solid black) Vs. Manual-2 IAD borders (dashed black).
Auto, LD Borders (solid)
Auto, LD Borders (solid)
Manual-1, LD Borders (dashed)Manual-1, LD Borders (dashed)
Manual-1, IAD Borders (dashed)
Auto, IAD Borders (solid)
Manual-1, IAD Borders (dashed)
Auto, IAD Borders (solid)