Design and Implementation - Segmentation of Hand Bone for Bone Age Assessment - page 94

Fig. 3.25 The final
detected hand bone edges of
Fig. 3.20 a, b after denoising
using k-mean clustering
Fig. 3.26 The Combination
of hand bone and detected
edges
3.5.3 The Area Restoration and Elimination Analysis
Similar to any existing segmentation frameworks, the proposed framework is by
no means a perfect segmentation framework because it is the inherent weakness of
conventional clustering method that it contains no pixel spatial information. These
imperfections or artifacts after segmentation are always due to two reasons:
1. False segmentation of spongy bone and therefore a restoration process was set up to
identify unfilled bounded areas (lost data) and fill it with original bone pixel value.
2. False labeling of soft-tissue regions as bone and therefore a elimination process
was set up to identify these over-segmentation artifacts followed by eliminating
it and replacing it back to soft-tissue region.
The steps in the BARNAE: Step in part (A) demonstrates the labeling process in
each direction. Step in part (B) explains the stopping criteria. Step in part (C) defines
the recognition of bounded area, for it a noise or lost data. The entire process men-
tioned above is repeated in step in part (D). Last step involves the filling in the lost
data or elimination of noise. The overview of the BARNAE is illustrated in Fig. 3.27 .