Fast geometric flows based white matter fiber tract segmentation in DT-MRI
View Patent ↗Segmentation of White Matter Fiber Tract in DT-MRI images is achieved by using level sets methods including a method which will increase the speed and quality of the surface evolvement. A method system representing a measure of coincidence of the tensor field with the normal of the surface front to increase the speed and quality of a level set method in DT-MRI is disclosed.
1. A method to segment white matter fiber tracts from MRI image data by diffusion tensor imaging wherein a surface of an object is evolved using a surface evolution velocity, comprising: a processor to perform the following steps:
determining the surface evolution velocity as a function of a normalized tensor scalar product plus a coincidence term that measures the coincidence of a normal direction and a tensor field; and
growing the object in accordance with the surface evolution velocity; wherein the surface evolution velocity is expressed
as F=mean(NTSP (D i , D i+1 ), NTSP(D i , D i+2 ))+βCOIN wherein:
NTSP
(
D
i
,
D
i
-
1
)
=
D
i
:
D
i
-
1
Trace
(
D
i
)
Trace
(
D
i
-
1
)
,
wherein D i is the diffusion tensor;
NTSP (D i , D i−1 ) is the normalized tensor scalar product;
COIN is the coincidence factor and β is a constant.
2. The method as claimed in claim 1 , wherein the measure of coincidence COIN is expressed as
COIN
3
=
(
FA
)
*
N
·
Di
*
N
max
(
λ
1
,
λ
2
,
λ
3
)
.
3. The method as claimed in claim 1 , wherein the measure of coincidence COIN is expressed as
COIN
2
=
(
FA
)
*
N
·
D
i
*
N
D
i
*
N
.
4. The method as claimed in claim 1 , wherein the measure of coincidence COIN is expressed as
COIN 1 =N ·PE.
5. The method as claimed in claim 1 , wherein β=10.
6. The method as claimed in claim 1 , wherein the surface evolution is terminated at locations where the similarity measure NTSP is less than a threshold or when a gradient of the measure of similarity NTSP is less than a threshold.
7. A system to process an image from image data, including:
a processor;
application software operable on the processor to segment white matter fiber tracts from MRI image data by diffusion tensor imaging wherein a surface of an object is evolved using a surface evolution velocity, comprising:
determining the surface evolution velocity as a function of a normalized tensor scalar product plus a coincidence term that measures the coincidence of a normal direction and a tensor field; and
growing the object in accordance with the surface evolution velocity; wherein the surface evolution velocity is expressed
as F=mean(NTSP (D i , D i+1 ), NTSP (D i , D i+2 ))+βCOIN wherein:
NTSP
(
D
i
,
D
i
-
1
)
=
D
i
:
D
i
-
1
Trace
(
D
i
)
Trace
(
D
i
-
1
)
,
wherein D i is the diffusion tensor;
NTSP (D i , D i−1 ) is the normalized tensor scalar product;
COIN is the coincidence factor and β is a constant.
8. The system as claimed in claim 7 , wherein the measure of coincidence COIN is expressed as
COIN
3
=
(
FA
)
*
N
·
Di
*
N
max
(
λ
1
,
λ
2
,
λ
3
)
.
9. The system as claimed in claim 7 , wherein the measure of coincidence COIN is expressed as
COIN
2
=
(
FA
)
*
N
·
D
i
*
N
D
i
*
N
.
10. The system as claimed in claim 7 , wherein the measure of coincidence COIN is expressed as
COIN 1 =N ·PE.
11. The system as claimed in claim 7 , wherein β=10.
12. The system as claimed in claim 7 , wherein the surface evolution is terminated at locations where the similarity measure NTSP is less than a threshold or when a gradient of the measure of similarity NTSP is less than a threshold.