Fig 1.
Generalized block diagram of proposed LCM based automatic Cobb angle measurement method.
Fig 2.
Region of interest cropping from the original image.
Fig 3.
Method of CM computation on an image, Cn = {Cn,1, Cn,2, …Cn,4} of the nth pixel of k = 4iterations.
Fig 4.
Principle of calculated centre of mass: (a) Intensity profile of an image (fn, blue) calculated local CM (Cn, red) and the identity line (n, dotted green) (b) Calculated weighting (Wm,n) of centre of mass.
Fig 5.
Cobb angle calculation procedure: (a) Centre curve, marking centre point and point of intersection; (b) Cobb angle calculation method on spine image.
Fig 6.
Algorithm flow chart to find Cobb angle for image directory.
Fig 7.
X-ray Image Segmentation of Scoliosis in (a) Thoracic region (b) Thoraco-Lumbar region.
Fig 8.
Left: Intensity profile of the horizontal blue line in Fig 7 (top, left), indicated in blue. Additionally, the local CM in red and the identity line as a dotted green line are presented. Right: Depiction of the weighting (wm,n).
Table 1.
Optimal values of the parameters chosen for the two methods.
Table 2.
Quantitative analysis of LCM and GMM-HMRF segmentation methods.
Fig 9.
Background subtraction process to separate spinal region: (a) input spine image; (b) LCM segmented image with predefined colors; (c) binary matrix of the mask; (d) cropped image after background subtraction; (e) after canny edge detection; (f) center line marking.
Fig 10.
Measurement of Cobb angle using the proposed method on various scoliosis cases.
Fig 11.
The bias of measured Cobb angle from the actual angle obtained from ground truth images for 50 subjects.