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.

xvi LISTE DES FIGURES<br />

3.7 The average Dice similarity coefficient (DSC) of left <strong>and</strong> right hippocampi<br />

using locally weighted voting (LWV) on the normal control<br />

(NC) atlas set. The atlases selected by maximal marginal relevance<br />

(MMR) re-ranking. The parameter λ varies from 0.1 to 0.9. The<br />

case of λ = 1 is equivalent to the selection by NMI ranking. . . . . 68<br />

3.8 The average Dice similarity coefficient (DSC) of left <strong>and</strong> right hippocampi<br />

using locally weighted voting (LWV) on the normal control<br />

(NC) atlas set. The atlases are selected according to image similarity<br />

ranking, maximal marginal relevance (MMR) re-ranking <strong>and</strong><br />

least angle regression sequence. . . . . . . . . . . . . . . . . . . . . 69<br />

4.1 Parameterizations <strong>and</strong> reparameterizations of the shape surfaces<br />

<strong>and</strong> correspon<strong>de</strong>nces. . . . . . . . . . . . . . . . . . . . . . . . . . . 73<br />

4.2 Parameterization of quadrilateral. . . . . . . . . . . . . . . . . . . . 78<br />

4.3 Multi-resolution subsampling by icosahedron subdivision. . . . . . . 80<br />

4.4 Cutting the octahedron <strong>and</strong> unfold to a plane. . . . . . . . . . . . . 81<br />

4.5 The construction of a shape image, from shape to sphere to embed<strong>de</strong>d<br />

octahedron to final shape image. . . . . . . . . . . . . . . . . . 82<br />

4.6 Reparameterization on the spherical parameterization <strong>and</strong> on the<br />

shape image. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83<br />

4.7 Six possible choices of reorientation <strong>and</strong> unfolding of the octahedron. 88<br />

4.8 The compactness of the Statistical <strong>Shape</strong> Mo<strong>de</strong>l (SSM). . . . . . . 98<br />

4.9 The behavior of the Specificity(M) in the groupwise optimization of<br />

the Statistical <strong>Shape</strong> Mo<strong>de</strong>l (SSM), using the first M = 15 principal<br />

components, the number of trials in the simulation N = 1000. . . . 99<br />

4.10 The <strong>de</strong>scendance of Generalisability(M) in the groupwise optimization<br />

of the Statistical <strong>Shape</strong> Mo<strong>de</strong>l (SSM), using the first M = 15<br />

principal components. . . . . . . . . . . . . . . . . . . . . . . . . . 100<br />

4.11 The shape variation in the first 3 mo<strong>de</strong>s of the left hippocampal<br />

shape mo<strong>de</strong>l (±2 st<strong>and</strong>ard <strong>de</strong>viations). . . . . . . . . . . . . . . . . 101<br />

4.12 The precision of the mo<strong>de</strong>l reconstruction with the estimated shape<br />

parameter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104<br />

4.13 An example of reconstruction from estimated shape parameters. . . 105<br />

4.14 An example of regularization on symmetric shape parameter estimation.<br />

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106<br />

5.1 The pipeline of local shape <strong>de</strong>scriptor extraction. . . . . . . . . . . 112<br />

5.2 The significance map of local difference on hippocampal surface by<br />

Hotelling’s T 2 test. . . . . . . . . . . . . . . . . . . . . . . . . . . . 121<br />

5.3 Threshol<strong>de</strong>d significance map, by thresholding p-values in the significance<br />

map Figure 5.2. . . . . . . . . . . . . . . . . . . . . . . . . 122<br />

5.4 The percentage of l<strong>and</strong>marks selected, with varying thresholds in<br />

the l<strong>and</strong>mark selection on the significance map of Hotelling’s T 2 test.123<br />

5.5 The training accuracy of the disease classification using bagged support<br />

vector machines (SVMs) with varying thresholds in the l<strong>and</strong>mark<br />

selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

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

Saved successfully!

Ooh no, something went wrong!