Archives

  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • Start br Load the D lung Load the

    2020-08-18

    Start
    Load the 3D lung Load the initial
    binary image trachea position
    Threshold with 0-
    Label A=label
    Label B = List the labels
    in slice z+1 that intersect
    with label A(j) in slice z
    j = number label in A
    Y
    Calculate circularity each label
    in label B and the different area
    Circularity >0.8 and
    Y
    Label A=label B
    Z=total slice
    Y
    Create 3D area
    base on all label
    Store as trachea
    area
    Finish
    Fig. 8. The flowchart to find the initial point used for 3D controlled labeling.
    Finish
    Fig. 9. The flowchart of 3D controlled labeling.
    Please cite this 5-Azacytidine article as: R. Nurfauzi, H. A. Nugroho, I. Ardiyanto et al., Autocorrection of lung boundary on 3D CT lung cancer imagesq, Journal of King Saud University – Computer and Information Sciences, https://doi.org/10.1016/j.jksuci.2019.02.009
    R. Nurfauzi et al. / Journal of King Saud University – Computer and Information Sciences xxx (xxxx) xxx 5
    Start
    Load the sorted
    Load the 3D lung Load the initial
    3D lung CT
    binary image trachea position
    images
    CT= the largest V = label each region
    in each slice of lung
    label in slice i
    binary image i=z
    A= the largest label
    B= the largest label
    Y
    Y
    N
    N
    Store V as 3D lung
    Y
    binary images
    N
    Finish
    do the morphology
    dilated in each 2 biggest
    area of the slice i
    area intersection of
    morphology
    operation
    Normalize aRoi to 0-1
    Threshold aRoi in T-d
    labeled of RT.*B
    B= the largest label in n=1
    the labeled RT.*B
    Fig. 10. The flowchart of lung fusion separation.
    The presence of lung fusion in 3D lung CT images can be detected by scanning and finding the abnormality growth of the biggest lung area in each slice from the top to the bottom. The intersection areas resulting from the morphological dilation in the two lung areas in the previous slice of the incision are detected as the RoI.
    The morphological dilation uses a disk kernel with size of
    Fig. 11. The Region of Interest (ROI) of lung fusion.
    Fig 12. Border correction steps. a) input of 2D images, b) labelling, c) selecting label areas, (d) corner feature extraction, (e) connecting the pair of features point, (f) filling holes.
    find the pleural cavity area in the RoI. The pleural cavity is used to separate the lung fusion in binary image.
    2.6. Lung boundary correction
    Juxta-pleural and juxta-vascular are two types of nodules that have similar properties. The presence of these nodules affects the basin at lung boundary as shown in Fig. 12(a). Adaptive border marching (ABM) method modified by corner detection is proposed to cover the nodule basin with shorter computational time than that of the original one.
    There are four steps to cover nodule basin. The first step is 2D labeling. The second step is corner feature extraction by using Har-ris method (Harris et al., 1988) in each label. The corner feature extraction aims to reduce the computational time of ABM conven-tional. The connected corner points in the border lung form the curve. One of the corner points is caused by the nodules. The corner points in border lung are shown in Fig. 12(d).
    The third step is selection of corner point pairs in lung boundary which is assumed as nodule basin. Support vector machine (SVM) is applied to select the corner points in the nodule basin. SVM rules is adopted from Pu et al. (2008). The SVM design is shown in Fig. 13 with Eq. (2). Alpha is the division of curvature depth (Hmax) with length of two points (W). SVM selects the corner points depending on alpha value. Alpha value is set up to 0.33 (Pu et al., 2008). The selected curvature or nodule basin by SVM depends on the pixel value in length of H, which is perpendicular to both sides of the corner pair. The selected pairs are shown in Fig. 12(e). The process of these steps is depicted in Fig. 14.
    The fourth step aims to fill the result of previous step. Morpho-logical filling holes is applied as shown in Fig. 12(f).
    a ¼ Hmax ð2Þ
    W
    Please cite this article as: R. Nurfauzi, H. A. Nugroho, I. Ardiyanto et al., Autocorrection of lung boundary on 3D CT lung cancer imagesq, Journal of King Saud University – Computer and Information Sciences, https://doi.org/10.1016/j.jksuci.2019.02.009
    6 R. Nurfauzi et al. / Journal of King Saud University – Computer and Information Sciences xxx (xxxx) xxx
    Start
    lung binary image
    Label each slice