Biomedical Engineering Reference
In-Depth Information
Chapter 5
Conclusion and Future Works
5.1 Conclusion
The main topic objective has been achieved by designing a segmentation frame-
work that is invariant to both the limitations but at the same time exhibiting accu-
rate and consistent in comparisons to basic segmentation approaches. As a result,
empirical and analytical experiments showed that the proposed segmentation
framework was able to achieve relatively high accuracy despite the low number
of user-specified parameters and no training sample is needed. This property of
invariance, automaticity and other desired properties of segmentation have been
considered and incorporated in the designed framework in order to make it becom-
ing more practical in many daily applications such as the segmentation stage in
computer-aided skeletal age scoring system. This proposed framework minimizes
the labor-intensity, time consumption and high dependency on operator expertise
and subjectivity; automaticity also implies perfect repeatability and consistency
advantage.
The possible segmentation techniques and the current hand bone segmentation
techniques by analyzing the techniques applicability in automated hand bone seg-
mentation have been critically reviewed and evaluated to achieve the first objec-
tive. The evaluation comprises of computational complexity, number and nature of
user-specified parameters, the requirement for expert interventions during execu-
tion phase, the consistency of performance, and dependency on certain features,
the effect of inherent property of hand bone radiograph on each segmentation
technique.
The eventual purpose of existing histogram equalization has been redefined to
extend the capability to solve variability problem to achieve the second objective.
This solves the variability problem across different radiographs or radiographs
of patients from different age group by proposing a holistic histogram equaliza-
tion technique that considers luminance changes, the detail loss and the enhanced
contrast.
A fully automated anisotropic diffusion has been proposed, designed and
implemented to achieve the third objective. This automates conventional aniso-
tropic diffusion that requires manual intervention to specify the diffusion strength
and to select the scale. The anisotropic diffusion is able to smooth the irregular
Search WWH ::




Custom Search