# US 2006/0217925 A1 · Methods for entity identification

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2006/0217925 A1
- Document ID: US20060217925A1
- Status: Published application
- Kind: A1
- Application number: 11/366,236
- Filed: 2006-03-02
- Published: 2006-09-28
- Earliest claimed: 2005-03-23
- Inventors: Maxime Taron; Nikolaos Paragios; Marie-Pierre Jolly
- Web page: https://patentgrep.com/patent/US20060217925A1

## CPC

- G06T 7/0012: {Biomedical image inspection}
- G06F 17/10: Complex mathematical operations {(function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544)}
- G06T 7/12: Edge-based segmentation
- G06T 7/143: involving probabilistic approaches, e.g. Markov random field [MRF] modelling
- G06V 30/1904: {involving a deformation of the sample or reference pattern; Elastic matching}
- G06V 30/19087: {based on parametric eigenspace representations, e.g. eigenspace representations using pose or illumination parameters; Shape manifold}
- G06T 2207/10072: Tomographic images
- G06T 2207/20081: Training; Learning
- G06T 2207/30016: Brain
- G06V 30/10: Character recognition
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Certain exemplary embodiments comprise a method, which can comprise automatically determining a probability that an entity belongs to a representation set. The representation set can be associated with a set of vectors of parameters and associated covariance matrices. Each covariance matrix can be associated with uncertainties of values comprised in the vector of parameters.

## Claims

### Claim 1

A method comprising a plurality of activities comprising:
providing a system configured to automatically determine a probability that an entity belongs to a representation set, said representation set associated with a plurality of vectors of parameters and a plurality of covariance matrices computed from correspondences at zero isosurfaces and associated with each of said plurality of vectors of parameters, said plurality of covariance matrices associated with uncertainties of registration values comprised in said plurality of vectors of parameters, said probability based upon said plurality of covariance matrices and said plurality of vectors of parameters, said plurality of covariance matrices and said plurality of vectors of parameters determined based upon a plurality of exemplary entities corresponding to said representation set.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
rendering a user interface indicative of said probability.

### Claim 3 (depends on 1)

The method of claim 1, further comprising:
determining at least one of said plurality of covariance matrices.

### Claim 4 (depends on 1)

The method of claim 1, further comprising:
determining at least one of said plurality of covariance matrices:
ΣΘ=σ2({circumflex over (χ)}T{circumflex over (χ)}+γI)−1
where:
ΣΘ is said covariance matrix;
σ2 is a scalar, which is a scaling factor for said covariance matrix;
I is an identity matrix;
γ is an arbitrarily small positive parameter; and
χ
^
=
(
η
1
T
⁢
⁢
χ
⁢
⁢
(
x
1
)
⋮
η
K
T
⁢
⁢
χ
⁢
⁢
(
x
K
)
)
where:
χ(x1) is a matrix of dimensionality 2×N based on B-spline basis functions,
with N being a size of Θ; and
ηi=∇φT(x′i), 2×1 column vector of image gradient:
where
∇ is a mathematical gradient;
φT is a Euclidean distance transform; and
xi is a point coordinate for a particular representation.

### Claim 5 (depends on 1)

The method of claim 1, further comprising:
determining a probability density function associated with said plurality of vectors of parameters, said probability density function determined responsive to a registration of said plurality of exemplary entities corresponding to said representation set.

### Claim 6 (depends on 1)

The method of claim 1, further comprising:
registering said entity via a transformation using free form deformation to match said entity to said representation set via energy minimization, said probability based upon said registering activity.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
registering said entity based upon a cubic B-spline, said probability based upon said registering activity.

### Claim 8 (depends on 1)

The method of claim 1, further comprising:
registering said entity utilizing a free form deformation model according to a topology preservation algorithm, said probability based upon said registering activity.

### Claim 9 (depends on 1)

The method of claim 1, further comprising:
attempting to minimize an energy function with a retrieval of a principal mode of a probability density function of a form α exp(E/β)
where:
E is an energy function to be minimized;
α is an unknown factor so the density sums to one; and
β is a selected bandwidth scaling parameter.

### Claim 10 (depends on 1)

The method of claim 1, further comprising:
registering said entity, said probability based upon said registering activity, said entity register via one or more attempts to optimize an objective function:
Eα∞((Θ))+wEsmooth((Θ))]
where:
E is a global data-based energy function to be minimized;
is a registration transform ((Θ): R2→R2);
w is a weight factor;
Θ is said vector of parameters; and
Esmooth((Θ))=∫∫Ω(|xx|2+2|xy|2+|yy|2)dΩ
where:
x is a coordinate of a point partially describing said entity;
y is a coordinate of a point partially describing said entity;
xx is a second derivative of registration transform; and
Ω is a domain of an image of said entity.

### Claim 11 (depends on 1)

The method of claim 1, further comprising:
registering said entity, said probability based upon said registering activity; and
continuously recalculating a distance map associated with said entity responsive to an iterative minimization of an energy function, said energy function associated with said registering activity.

### Claim 12 (depends on 1)

The method of claim 1, further comprising:
registering each of said plurality of exemplary entities via an affine transformation.

### Claim 13 (depends on 1)

The method of claim 1, further comprising:
warping a model of each of said plurality of exemplary entities to a shape of said representation set, said model constrained in a normal direction, said model unconstrained in a tangential direction.

### Claim 14 (depends on 1)

The method of claim 1, further comprising:
finding a plurality of transformations for each of said plurality of exemplary entities to shapes of said representation set, each of said plurality of transformations associated with a weight and a covariance matrix of said plurality of covariance matrices.

### Claim 15 (depends on 1)

The method of claim 1, further comprising:
evaluating said probability based upon a hybrid estimator according to an equation:
f
^
H
⁡
(
x
,
∑
)
=
∑
(
x
i
,
∑
i
⁢
,
w
i
)
∈
ℨ
K
⁢
w
i
·
𝒦
⁡
(
x
,
∑
,
x
i
,
∑
i
)
where:
{circumflex over (ƒ)}H is said hybrid estimator;
K is a number of kernels extracted from said representation set;
κ is a normal probability density function: N(x−xi, (Σ++Σi+)+)
w is a weight factor;
x is an element associated with said representation set; and
Σ is an indexed element from said covariance matrix.

### Claim 16 (depends on 1)

The method of claim 1, further comprising:
determining at least one of said plurality of vectors of parameters.

### Claim 17 (depends on 1)

The method of claim 1, further comprising:
iteratively determining a most likely probability density function associated with said entity, said probability based upon said most likely probability density function.

### Claim 18 (depends on 1)

The method of claim 1, further comprising:
reducing a number of kernels associated with said vector of parameters via a selection of a subset of kernels from a set of kernels via a maximum likelihood criterion, said probability based upon said subset of kernels.

### Claim 19 (depends on 1)

The method of claim 1, further comprising:
via an iterative suboptimal algorithm, reducing a number of kernels associated with said vector of parameters via a selection of a subset of kernels from a set of kernels.

### Claim 20 (depends on 1)

The method of claim 1, further comprising:
testing a validity of a model of a set of representations can be tested by determining a log likelihood via evaluating an expression:
C
K
=
∑
i
=
1
M
⁢
log
⁢
⁢
(
1
K
⁢
∑
(
x
j
,
∑
j
)
∈
ℨ
K
⁢
𝒦
⁡
(
x
j
,
∑
j
⁢
,
x
i
,
∑
i
)
)
where:
CK is a log likelihood;
K is a number of kernels extracted from said representation set;
M is a total number of kernels in said representation set;
xi is an element associated with said representation set; and
Σi is an indexed element from said covariance matrix.

### Claim 21 (depends on 1)

The method of claim 1, further comprising:
building a statistical estimator via kernels and Parzen Window density estimation, said probability based upon said statistical estimator.

### Claim 22

A method comprising:
automatically determining a probability that an entity belongs to a representation set, said representation set associated with a plurality of vectors of parameters and a plurality of covariance matrices computed from correspondences at zero isosurfaces and associated with each of said plurality of vectors of parameters, said plurality of covariance matrices associated with uncertainties of registration values comprised in said plurality of vectors of parameters, said probability based upon said plurality of covariance matrices and said plurality of vectors of parameters, said plurality of covariance matrices and said plurality of vectors of parameters determined based upon a plurality of exemplary entities corresponding to said representation set.

### Claim 23

A machine-readable medium comprising machine instructions for activities comprising:
automatically determining a probability that an entity belongs to a representation set, said representation set associated with a plurality of vectors of parameters and a plurality of covariance matrices computed from correspondences at zero isosurfaces and associated with each of said plurality of vectors of parameters, said plurality of covariance matrices associated with uncertainties of registration values comprised in said plurality of vectors of parameters, said probability based upon said plurality of covariance matrices and said plurality of vectors of parameters, said plurality of covariance matrices and said plurality of vectors of parameters determined based upon a plurality of exemplary entities corresponding to said representation set.

### Claim 24

A method comprising a plurality of activities comprising:
providing a system configured to automatically determine a probability that an entity belongs to a representation set, said representation set associated with a plurality of vectors of parameters and a plurality of covariance matrices computed from correspondences at zero isosurfaces and associated with each of said plurality of vectors of parameters, said plurality of covariance matrices associated with uncertainties of registration values comprised in said plurality of vectors of parameters, said probability based upon said plurality of covariance matrices and said plurality of vectors of parameters, said plurality of covariance matrices and said plurality of vectors of parameters determined based upon a plurality of exemplary entities corresponding to said representation set; and
a means for causing a user interface to be rendered, the user interface indicative of said probability.
