Automatic detection and identification of brain sulci in MRI
View Patent ↗Disclosed is a method for determining a position of a brain sulcus in the brain of a patient, the method comprising executing, on at least one processor of at least one computer, steps of: acquiring, at the at least one processor, patient image data describing a digital medical image of at least part of the brain; acquiring, at the at least one processor, atlas data describing an image-based model of the at least part of the brain; determining, by the at least one processor and based on the patient image data and the atlas data, mapping data describing a transformation between the patient image data and the atlas data; and determining, by the at least one processor and based on the patient image data and the atlas data and the mapping data, sulcus position data describing the position of the brain sulcus in the digital medical image.
1. A method for determining a position of a brain sulcus in the brain of a patient, the method comprising executing, on at least one processor of at least one computer, steps of:
a) acquiring, at the at least one processor, patient image data describing a digital medical image of at least part of the brain, wherein the digital medical image is represented by a data set representing a curved surface which is defined in three dimensions;
b) acquiring, at the at least one processor, atlas data describing an image-based model of the at least part of the brain;
c) determining, by the at least one processor and based on the patient image data and the atlas data, mapping data describing a transformation between the patient image data and the atlas data, wherein the atlas data describes color values defining the image-based model and the patient image data describes color values defining the digital medical image, and wherein the mapping data is determined by comparing the color values of the atlas data to the color values of the patient image data, wherein each color value defining the digital medical image is associated with a type of anatomical tissue and each color value defining the image-based model is associated with information about a type of anatomical tissue;
d) determining, by the at least one processor and based on the patient image data, the atlas data, and the mapping data, sulcus position data describing the position of the brain sulcus in the digital medical image, wherein the sulcus position data is determined in a projection of the digital medical image onto a planar surface which is defined in two dimensions;
e) determining, based on the mapping data, the atlas data, and the patient image data, tissue type data describing at least one type of anatomical tissue described by the digital medical image and the position of the at least one type of anatomical tissue in the digital medical image;
f) determining, based on the tissue type data, sulcus mask data describing a binary mask which can be applied to the digital medical image for masking it in order to differ, in the digital medical image, between the union of cortical grey brain matter and cortical cerebrospinal fluid on the one hand and white brain matter on the other hand; and
g) filtering the sulcus position data by applying a two-dimensional filament detection filter to the sulcus position data to enhance image contrast of the sulcus.
2. The method according to claim 1 , wherein the atlas data serves as a basis for estimating the position of the brain sulcus.
3. The method according to claim 1 , wherein the image-based model and the digital medical image have been generated with the same imaging modality, or wherein the image-based model and the digital medical image have been generated with different imaging modalities and the atlas data comprises a mapping between a color value range associated with the imaging modality used for generating the digital medical image and a color value range associated with the imaging modality used for generating the image-based model.
4. The method according to claim 1 , wherein the projection of the digital medical image is determined by projecting a curved surface in the patient image data, which is at least substantially parallel to an outer surface of the brain, onto a planar surface, wherein the curved surface is defined based on a spherical harmonics expansion of the outer brain surface up to a predetermined order.
5. The method according to claim 4 , wherein the curved surface of a given distance to the outer surface is defined by its radius r(θ,φ) as a function of the angular coordinates θ and φ in the digital medical image and wherein r(θ,φ) is expanded in real-valued spherical harmonics up to a given order l max so that
r
(
θ
,
φ
)
=
∑
l
=
0
l
ma
x
∑
m
=
-
l
l
r
l
m
Y
l
m
(
θ
,
φ
)
where the real-valued coefficients r lm are obtained by projecting r(θ,φ) onto the corresponding spherical harmonics:
r lm =∫ θ=0 π ∫ φ=0 2π r (θ,φ) Y lm (θ,φ)sin θ dθdφ,
and the real-valued spherical harmonics Y lm are defined as
Y
l
m
=
{
2
(
2
l
+
1
)
(
l
-
m
)
!
4
π
(
l
+
m
)
!
P
l
m
(
cos
θ
)
sin
(
m
φ
)
for
m
<
0
(
2
l
+
1
)
4
π
P
l
m
(
cos
θ
)
for
m
=
0
2
(
2
l
+
1
)
(
l
-
m
)
!
4
π
(
l
+
m
)
!
P
l
m
(
cos
θ
)
cos
(
m
φ
)
for
m
>
0
with P l m (cos θ) being the associated Legendre polynomials, and wherein the curved surface r(θ,φ) is projected onto a planar surface defined by the coordinates x and y, and for example points towards the right lateral side of the brain and the anterior part of the brain, respectively, and wherein r(θ,φ) and (x, y) are related by
x
=
r
(
φ
-
φ
0
)
and
y
=
r
ln
[
tan
(
π
2
-
θ
2
)
]
,
wherein the y-axis is defined to be at φ 0 , and
θ
=
π
-
2
tan
-
1
(
e
y
/
r
)
φ
=
x
r
+
φ
0
,
wherein positions in the planar surface are projected onto the curved surface in the digital medical image by determining r(θ,φ) for each pair of (x, y).
6. The method according to claim 5 , further comprising:
determining, based on the sulcus mask data and the relationship between r(θ,φ) and (x, y), projected mask data describing a projection of the binary mask onto the planar surface described by (x, y).
7. The method according to claim 6 , further comprising:
determining the sulcus position data for the projected digital medical image based on the projected mask data and the projected patient image data.
8. The method according to claim 7 , wherein the sulcus position data for the projected digital medical image is determined by laying the projection of the binary mask over the projected digital medical image.
9. The method according to claim 8 , wherein determining the sulcus position data comprises applying a watershed transform to the masked projected digital medical image, and conducting data processing which represents merging the resulting candidate sulcal basins.
10. The method according to claim 9 , further comprising:
determining the sulcus position data for the projected digital medical image based on determining candidate sulcal basins described by the sulcus position data, wherein determining the sulcus position data comprises sorting candidate basins being candidates for the sulcus basin of which the position is to be determined, wherein the candidate basins are sorted according to a criterion including at least one of an average distance of a candidate basin from the position of the brain sulcus, the elongatedness of a candidate basin, or the directionality of the candidate basin.
11. The method according to claim 10 , wherein the projected and masked patient image data is slice image data generated by application of a tomographic imaging modality, and wherein merging the candidate sulcal basins comprises application of an algorithm for calculating a minimal cost path between at least two extremal points of a candidate sulcal basin, and wherein the position of a central surface of a sulcus is determined as the union of such minimal cost paths for at least two slice images of the projected and masked patient image data.
12. The method according to claim 11 , further comprising:
determining, based on the union of minimal cost paths and the projected and masked patient medical image data, re-projected sulcus position data describing the position of the brain sulcus in the three-dimensional digital medical image by applying the relationship between r(θ,φ) and (x, y).
13. The method according to claim 1 , wherein the mapping data is determined by applying an image fusion algorithm to the patient image data and the atlas data.
14. The method according to claim 1 , further comprising:
acquiring sulcus identification data describing the identity of at least one brain sulcus, wherein the mapping data is determined based on the sulcus identification data by determining the position of the brain sulcus described by the sulcus identification data in the image-based model.
15. The method according to claim 1 , wherein the two-dimensional filament detection filter is a Frangi vesselness filter.
16. A non-transitory computer-readable storage medium having stored thereon computer-executable instructions that, when executed, configure a processor to:
acquire patient image data describing a digital medical image of at least part of the brain, wherein the digital medical image is represented by a data set representing a curved surface which is defined in three dimensions;
acquire atlas data describing an image-based model of the at least part of the brain;
determine, based on the patient image data and the atlas data, mapping data describing a transformation between the patient image data and the atlas data, wherein the atlas data describes color values defining the image-based model and the patient image data describes color values defining the digital medical image, and wherein the mapping data is determined by comparing the color values of the atlas data to the color values of the patient image data, wherein each color value defining the digital medical image is associated with a type of anatomical tissue and each color value defining the image-based model is associated with information about a type of anatomical tissue;
determine, based on the patient image data, the atlas data, and the mapping data, sulcus position data describing the position of the brain sulcus in the digital medical image, wherein the sulcus position data is determined in a projection of the digital medical image onto a planar surface which is defined in two dimensions;
determine, based on the mapping data, the atlas data, and the patient image data, tissue type data describing at least one type of anatomical tissue described by the digital medical image and the position of the at least one type of anatomical tissue in the digital medical image;
determine, based on the tissue type data, sulcus mask data describing a binary mask which can be applied to the digital medical image for masking it in order to differ, in the digital medical image, between the union of cortical grey brain matter and cortical cerebrospinal fluid on the one hand and white brain matter on the other hand; and
filter the sulcus position data by applying a two-dimensional filament detection filter to the sulcus position data to enhance image contrast of the sulcus.
17. The non-transitory computer-readable storage medium of claim 16 wherein the two-dimensional filament detection filter is a Frangi vesselness filter.