Archives

  • 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • br labelling br Extract harris corner

    2020-08-18


    labelling
    Extract harris corner
    in label Lb in slice s
    L= the distance of pair
    corner feature (P) in
    label Lb;
    H=Alpha*L
    Find the position (x,y)
    at a distance H that is
    perpendicular to L in
    label Lb
    N
    Put value 1 throughout the
    interpolation in corner
    feature pair (P) on the
    label -Lb
    Is the last pair?
    Fill the holes
    Is the last label in slice s?
    N
    Y
    Is it the last slice? N S=S+1
    Y
    Store the lung
    border correction
    Finish
    Fig. 14. The flowchart of lung boundary correction with corner feature.
    2.7. Testing and evaluating
    Performances of the proposed method are evaluated in terms of computational time, the nodules area that unsuccessful covered as
    Fig. 15. The error in segmenting the 5Azacytidine without lung separation stage in lung boundary correction (a) ABM conventional; (b) ABM with local min-max detection; (c) ABM with corner detection; (d) morphological closing.
    Fig. 16. The result of lung extraction steps. (a) Before and (b) after finding the maximum area.
    Fig. 17. An example result of lung extraction process, left for 2D and right for 3D. (a and b) The result of lung extraction stage. (c and d) The result of tracheal extraction stages. (e and f) The result of lung fusion separation.
    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
    R. Nurfauzi et al. / Journal of King Saud University – Computer and Information Sciences xxx (xxxx) xxx 7
    Fig. 18. The result of the proposed method to separate the lung fusion. (a) there are lung fusion area with curve pattern.
    Fig. 19. The results of all lung boundary method (a) before and; (b) after using trachea elimination and lung fusion separation.
    called the under segmentation of nodule (US), the over region of nodule as called the over segmentation of nodule (OS) and false positive of nodule.
    US and OS can be calculated by Eq. (3) and Eq. (4) respectively. S is the segmented area resulting by the proposed method and SG is the GT area. The false positive is a condition when the system incorrectly detect an object as a nodule. These parameters are used compare performance of the proposed method with three others method, i.e., ABM- conventional (Pu et al., 2008), ABM-minmax feature (Nurfauzi et al., 2017); and morphological closing (Javaid et al., 2016; Teramoto and Fujita, 2013; Gupta et al., 2015).
    3. Result and discussion
    The separation of lung fusion is important stage that must be taken before correction of the lung boundary step. The mediastinal region will be recognized as a nodule basin when the separation of lung fusion is not used. Fig. 15 shows errors in lung segmentation resulting from all methods used to correct lung boundaries.
    The aim of lung extraction stage is to eliminate the region that is not attached to the lungs as shown in Fig. 16. This result still has a trachea and lung fusion as shown in Fig. 17(a). To overcome this limitation, tracheal extraction and lung fusion separation are applied. The results of these stages are shown in Fig. 17(c-d), and (e-f) respectively. At the tracheal extraction stage, a lot of data is unsuccessful for extracting the trachea. This limitation happens because the trachea and the bronchioles have similar features.
    The proposed method on separating of lung fusion by using multi-threshold is successfully applied. The smooth separation is produced as shown in Fig. 18.
    For reducing the computational time, the modified ABM with extracted corner feature is used. The result is shown in Fig. 19.
    As can be seen from Fig. 19, the lung boundary can be clearly defined without any error segmentation which generally occurs in the middle area of 2D CT slice. This result shows that the pro-posed method is suitable to avoid error segmentation in the medi-astinal region.
    The other limitations of the previous studies, such as long com-putational time (Pu et al., 2008) and a large number of under-
    Fig. 20. Segmented lung area in four lung boundary correction methods: (a) the characteristic of each method to repair the lung boundary; (b) unregulated area of morphological closing.
    The result of lung boundary correction with using 57 3D CT images containing juxta pleural nodules.
    Component ABM ABM-Corner ABM Morphological closing (Javaid et al., 2016;
    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