Literature Reviews - Segmentation of Hand Bone for Bone Age Assessment - page 16

Sharif et al. [ 22 ] have conducted a study on bone outlines detection to per-
form bone segmentation using edge detection depending on the intensity from
the Derivative of Gaussian (Drog) before implementing the thresholding tech-
nique. The technique of pre-processing proposed by Mahmoodi et al. [ 13 ] involve
converting the image into binary image and adopting the thresholding technique
determined by image histogram to acquire the ROI. The subsequent segmen-
tation of epiphysis residing in the ROI is performed via the active shape model
framework. As mentioned, the drawbacks of this method are associated with the
sensitivity in uneven illumination and the presence of soft-tissue region. The pre-
processing method used in Mahmoodi et al. [ 23 ] for segmentation of bone using
deformable models and a hierarchical bone localization scheme. The method back-
ground removing process is performed only after obtaining the ROI. Mahmoodi et
al. [ 14 ] adopt binary thresholding to acquire the delineation of the hand, followed
by location searching of concave-convex; finally the segmentation is performed by
the method of active shape models.
Sebastian et al. [ 15 ] segmented the carpal bones from CT images by deform-
able models, the pre-processing incorporates the strength of various segmentation
techniques such as snake models, region-based segmentation, global competition
in seeded region growing and also the local competition in region competition. The
drawback of this method is that it is complicated and involves intensive computing
consumption during the computation of the partial differential equation. Besides,
active contour model has been invariably adopted in partitioning the hand bones,
the methods c-means clustering algorithm, Gibbs random fields and estimation of
the intensity function have been also proposed by Pietka et al. [ 24 ]. Also, They sug-
gested to segment the hand bone using the histogram analysis during pre-processing
stage [ 25 ] by acquiring the histogram peak of pixels intensity followed by identify-
ing the background and soft-tissue region.
Hsieh et al. [ 26 ] incorporate adaptive segmentation method with Gibbs random
field at the pre-processing stage. Zhang et al. [ 27 ] suggest segmenting the carpal
by non-liner filter as pre-processing follow by adaptive image threshold setting,
binary image labelling and small object removal. However, it involves user-spec-
ified threshold and Canny edge detection which are not robust in segmentation.
Similarly, Somkantha et al. [ 5 ] segment the carpals bone using a combination of
vector image model and Canny edge detector. Han et al. [ 7 ] propose to implement
watershed transform and gradient vector flow (GVF) to perform the segmentation
where the performance of watershed transform and GVF depends heavily on edge
gradient strength. Tran Thi My Hue et al. [ 28 ] propose to implement watershed
transform with multistage merging for the segmentation task.
The utilization of the state-of-the-art technique of deformable model such as
Active Shape Model (ASM) and Active Appearance Model (AAM) of hand bone
segmentation has gained considerable attentions in recent years [ 29 - 31 ]. The
strength of this method is that it is well-founded on statistical learning theory.
However, the main drawback of this technique is that it is not yet developed into a
fully automated system. The initialization of the technique is to delineate the hand
bone shape and this thus far is accomplished manually.