# US 2013/0259340 A1 · Voxel-based transformation method for transforming diffusion mri data and groups test method using the same

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

- Number: US 2013/0259340 A1
- Document ID: US20130259340A1
- Status: Published application
- Kind: A1
- Application number: 13/524,704
- Filed: 2012-06-15
- Published: 2013-10-03
- Earliest claimed: 2012-04-03
- Assignees: National Taiwan University
- Inventors: Isaac Wen-Yih Tseng; Fang-Cheng Yeh
- Web page: https://patentgrep.com/patent/US20130259340A1

## CPC

- G01R 33/56341: {Diffusion imaging}
- G01R: MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES (indicating correct tuning of resonant circuits H03J3/12)

## Abstract

A voxel-based transformation method includes: a) obtaining a MRI dataset in a subject space associated with subject voxel coordinates, subject sampling directions, and subject voxel spin amounts, and a dataset of a co-registration template associated with template voxel coordinates, each subject voxel coordinate corresponding to a template voxel coordinate according to a mapping function; b) through an inverse of the mapping function, obtaining subject voxel coordinates and a Jacobian matrix; and c) obtaining template voxel spin amounts, each being a function of a template sampling direction and a template voxel coordinate, using the Jacobian matrix and image data.

## Claims

### Claim 1

A voxel-based transformation method for a magnetic resonance imaging (MRI) dataset, comprising:
a) obtaining a subject MRI dataset in a subject space and a dataset of a co-registration template in a template space, the subject MRI dataset being associated with a plurality of subject voxel coordinates, a plurality of subject sampling directions, and a plurality of subject voxel spin amounts, each of which is a function of one of the subject voxel coordinates and one of the subject sampling directions, the dataset of the co-registration template being associated with a plurality of template voxel coordinates, each of the subject voxel coordinates corresponding to one of the template voxel coordinates in the template space according to a mapping function of a co-registration method;
b) through an inverse of the mapping function, obtaining one of the subject voxel coordinates of the subject MRI dataset that corresponds to one of the template voxel coordinates of the co-registration template, and obtaining a Jacobian matrix associated with the obtained one of the subject voxel coordinates and the corresponding one of the template voxel coordinates; and
c) through a q-space distribution reconstruction (QSDR) function, obtaining a template voxel spin amount which is a function of a template sampling direction and the corresponding one of the template voxel coordinates using the Jacobian matrix and image data at said one of the subject voxel coordinates obtained in step b).

### Claim 2 (depends on 1)

The method as claimed in claim 1, further comprising:
d) performing step a) to step c) on a plurality of subjects, followed by averaging the template voxel spin amounts at the respective template voxel coordinate and the respective template sampling direction of the subjects to obtain a generated dataset, and constructing a generated template according to the generated dataset.

### Claim 3 (depends on 2)

The method as claimed in claim 2, wherein the generated template is calculated according to:
ψ
_
t

(
r
t
,
v
)
=
1
n

∑
i
=
1
n

ψ
t
,
i

(
r
t
,
v
)
,
where ψt(rt,v) is the template voxel spin amount at the respective template voxel coordinate (rt) and the respective template sampling direction (v), i denotes an ith subject, and n is a total number of the subjects.

### Claim 4 (depends on 1)

The method as claimed in claim 1, wherein, in step c), the template voxel spin amount is obtained for each of the template voxel coordinates, the method further comprising the following steps after step c):
e) calculating fiber orientation at each of the template voxel coordinates according to results obtained in step c); and
f) calculating quantitative anisotropy (QA), which is defined as amount of spins that undergo diffusion along the fiber orientation.

### Claim 5 (depends on 4)

The method as claimed in claim 4, further comprising the following step after step f):
g) calculating accumulated QA for each fiber orientation.

### Claim 6 (depends on 5)

The method as claimed in claim 5, wherein the accumulated QA is calculated as:
∫
R

Q


A

(
r
,
a
^

(
r
)
)



r
where R is the region occupied by the fiber and â(r) is the fiber orientation at coordinate r.

### Claim 7 (depends on 1)

The method as claimed in claim 1, wherein the template voxel spin amount satisfies:
ψ
t

(
r
t
,
v
)
=

J
ϕ
-
1

(
r
t
)


ψ
s

(
ϕ
-
1

(
r
t
)
,
J
ϕ
-
1

(
r
t
)

v

J
ϕ
-
1

(
r
t
)


v

)
where ψt(rt,v) is the template voxel spin amount at the respective template voxel coordinate (rt) and the respective template sampling direction (v), φ−1 denotes an inverse function of diffeomorphic mapping φ, |Jφ−1(rt)| is a Jacobian determinant of φ−1 at rt, ψs is an original spin distribution function (SDF) in the subject space, and ∥Jφ−1(rt)v∥ is a vector norm of a transformed vector Jφ−1(rt)v.

### Claim 8 (depends on 1)

The method as claimed in claim 1, wherein the QSDR function is:
ψ
t

(
r
t
,
v
)
=

J
ϕ
-
1

(
r
t
)


A

∑
i

W
i

(
ϕ
-
1

(
r
t
)
)

sin


c
(
σ

6

D


b
i

〈
g
i
,
J
ϕ
-
1

(
r
t
)

v

J
ϕ
-
1

(
r
t
)


v

〉
)
where ψt(rt,v) is the template voxel spin amount at the respective template voxel coordinate (rt) and the respective template sampling direction (v), φ−1 denotes an inverse function of diffeomorphic mapping φ, |Jφ−1(rt)| is a Jacobian determinant of φ−1 at rt, A is a constant term for a spin distribution function (SDF) quadrature, Wi(φ−1(rt)) is MR data at φ−1(rt) acquired when b-value=bi and diffusion gradient direction=gi, σ is a diffusion length ratio that controls percentage of coverage of diffusion spins, D is diffusivity of the diffusion spins, ∥Jφ−1(rt)v∥ is a vector norm of transformed vector Jφ−1(rt)v, and
<
g
i
,
J
ϕ
-
1

(
r
t
)

v

J
ϕ
-
1

(
r
t
)

v

>
is an inner product of gi and
J
ϕ
-
1

(
r
t
)

v

J
ϕ
-
1

(
r
t
)

v

.

### Claim 9

A group test method for magnetic resonance imaging (MRI) analysis, comprising:
(A) performing transformation on two groups, the two groups including a plurality of subjects, the transformation performed on each of the two groups including the following sub-steps:
(i) obtaining a subject MRI dataset in a subject space and a dataset of a co-registration template in a template space, the subject MRI dataset being associated with a plurality of subject voxel coordinates, a plurality of subject sampling directions, and a plurality of subject voxel spin amounts, each of which is a function of one of the subject voxel coordinates and one of the subject sampling directions, the dataset of the co-registration template being associated with a plurality of template voxel coordinates, each of the subject voxel coordinates corresponding to one of the template voxel coordinates in the template space according to a mapping function of a co-registration method;
(ii) through an inverse of the mapping function, obtaining one of the subject voxel coordinates of the subject MRI dataset that corresponds to one of the template voxel coordinates of the co-registration template, and obtaining a Jacobian matrix associated with the obtained one of the subject voxel coordinates and the corresponding one of the template voxel coordinates;
(iii) through a q-space distribution reconstruction (QSDR) function, obtaining a template voxel spin amount which is a function of a template sampling direction and the corresponding one of the template voxel coordinates using the Jacobian matrix and image data at said one of the subject voxel coordinates obtained in sub-step (ii), wherein the template voxel spin amount is obtained for each of the template voxel coordinates;
(iv) calculating fiber orientation at each of the template voxel coordinates according to results obtained in sub-step iii); and
(v) calculating quantitative anisotropy (QA), which is defined as amount of spins that undergo diffusion along the fiber orientation; and
(B) performing QA test of the two groups to obtain a p-value, which represents significance of QA difference, at each of the template voxel coordinates.

### Claim 10 (depends on 9)

The group test method as claimed in claim 9, wherein, in sub-step i), the co-registration template is one of a publicly available template and a generated template obtained from calculation of a population including a plurality of subjects, the calculation including:
performing sub-steps i) to iii) on the plurality of subjects, followed by averaging the template voxel spin amounts at the respective template voxel coordinate and the respective template sampling direction of the subjects to obtain a generated dataset, and constructing the generated template according to the generated dataset.

### Claim 11 (depends on 9)

The group test method as claimed in claim 9, wherein, in sub-step iv), the fiber orientation is calculated based on one of a publicly available template, a generated template obtained by averaging the template voxel spin amounts obtained for all of the subjects among the two groups, and a generated template obtained by:
performing sub-steps i) to iii) on a plurality of subjects, followed by averaging the template voxel spin amounts at the respective template voxel coordinate and the respective template sampling direction of the subjects to obtain a generated dataset, and constructing the generated template according to the generated dataset.

### Claim 12 (depends on 9)

The group test method as claimed in claim 9, further comprising:
(C) calculating a positive segment from results obtained in step (B), the positive segment representing fiber orientations that have p-values below a p-value threshold;
(D) calculating supra-threshold clusters formed by continuous positive segments using a fiber tracking algorithm; and
(E) determining clusters having significant difference using false discovery rate (FDR) control.
