Result and Discussion - Segmentation of Hand Bone for Bone Age Assessment - page 101

Chapter 4

Result and Discussion

Abstract In this chapter, the functionality of proposed MBOBHE and the modi-
fied anisotropic diffusion will be justified. Both techniques create a suitable envi-
ronment for the subsequent segmentation modules by reducing the disturbance
of uneven illumination, sharpening the edge and smoothing the texture; at the
same time, MBOBHE is capable of enhancing the features of ossification sites to
improve the performance of computerized BAA. After that, an analytical compari-
son of AAM segmentation framework with the proposed segmentation framework
is performed.

4.1 Introduction

As mentioned earlier, the radiograph used in all the experiments in this chap-
ter were adopted from hand bone online database http://ipilab.org which com-
prises both genders in four populations which are Caucasian, African American,
Hispanic, and Asian, ages ranging between 0 and 18, collected from Children's
Hospital Los Angeles (CHLA). To increase the validity of the experimental results,
a large set of radiographs from the mentioned database were used in experi-
ments to dampen down the possibilities of low accuracy and skewing results due
to extreme data. Besides, all experiments adopted hand bone radiographs which
range from different ages, genders and races to prevent problems arising from
incomplete data and bias data selection. Furthermore, the standard database is
available online to enable other researchers repeat the conducted experiments
using the same set of data to facilitate future comparisons. All the experiments
in assessing computational algorithms were conducted using identical ver-
sion of software and in identical personal computer to minimize random errors.
Qualitative and Quantitative experiments were set up to justify the previously dis-
cussed arguments in this topic:

(1) The first experiment was set up to justify the motivation of equalizing the
image before the system and to justify the strengths of the proposed Multiple
Beta Optimized Histogram Equalization (MBOHE) in comparison to other
current histogram equalizations.