TABLE I.
Summary of notation
Symbol | Description | |
---|---|---|
(Ai, Li) | Intensity and label images of the i-th atlas | |
(Ãi, L̃i) | Intensity and label images of the i-th atlas after non-rigid warping to a particular target | |
(T, LT, ) | Target intensity image, estimated label image (by MAS) and ground-truth label image, respectively. | |
I | Index-set of the N atlases | |
t | Index to denote the atlas acting as target image in the training set | |
K | Scalar denoting the number of selected atlases to be used for MAS | |
|
Set of indexes of the K best atlases for segmenting target image At, according to ground-truth | |
ST | Set of indexes of the K best atlases for segmenting target image T | |
f (·,·) | Scoring function that maps a pair of affinely registered images to their expected segmentation performance | |
w⃗ | Weighting vector learned by SVM-Rank | |
θ⃗i | Vector with the HOG features extracted from i-th image in the training set | |
M | Scalar denoting the total number of features | |
Θ⃗k | Vector with the k-th pairwise feature from all atlas-target pairs in the training set | |
Ψ⃗ | Vector with the ground-truth DR from all atlas-target pairs in the training set | |
Φ(·,·) | Compact vector of pairwise features from a pair of target and atlas images | |
m′ | Scalar denoting the size of the compact set of features |