14.08.2013 Views

Docteur de l'université Automatic Segmentation and Shape Analysis ...

Docteur de l'université Automatic Segmentation and Shape Analysis ...

Docteur de l'université Automatic Segmentation and Shape Analysis ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

140 Chapter 6 Conclusions<br />

6.2 Future works<br />

In the future work, new methodologies <strong>and</strong> algorithms for brain image segmenta-<br />

tion, shape mo<strong>de</strong>ling <strong>and</strong> analysis of neuroanatomical structures may be <strong>de</strong>veloped<br />

to extend the current work presented in this thesis.<br />

6.2.1 Non-local approach to the atlas based image segmentation<br />

Non-local approach has been <strong>de</strong>veloped to select patches of atlases similar to the<br />

target image. One advantage of this approach is its computational efficiency,<br />

since non-rigid registrations of the atlases images are no longer necessary in this<br />

approach. The label for each voxel is searched over the atlases, which provi<strong>de</strong>s a<br />

larger selection pool of c<strong>and</strong>idate labeling. The current strategies for atlas selection<br />

<strong>and</strong> fusion may be exten<strong>de</strong>d to the patch selection <strong>and</strong> fusion in the non-local<br />

approach. In addition to the propagation of the anatomical labeling, <strong>de</strong>mographic<br />

<strong>and</strong> diagnostic information of each atlas may also be propagated to the target<br />

image, thus localizing the effects of the pathology as well as normal physiology on<br />

a voxelwise basis.<br />

6.2.2 Groupwise shape correspon<strong>de</strong>nce by optimization<br />

The method for shape correspon<strong>de</strong>nce by optimization adopted <strong>and</strong> <strong>de</strong>veloped<br />

in this thesis applies a fluid regularization to the <strong>de</strong>formation, which is similar<br />

to the techniques wi<strong>de</strong>ly used image registration. The current state-of-the-art in<br />

the field of image registration provi<strong>de</strong>s some important insight to the solution<br />

of the correspon<strong>de</strong>nce problem by optimization. A general algorithmic framework<br />

accommodating a wi<strong>de</strong> range of registration techniques including fluid registration,<br />

elastic registration <strong>and</strong> Demons method may be implemented.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!