Segmenting biological structures from microscopy images
A method performed by one or more processors, includes: receiving an image to be segmented into one or more representations of one or more biological structures; accessing data representing a set of biological structures that are derived from other biological structures delineated in a training image, wherein the training image is associated with a level of modality that corresponds to a level of modality associated with the image to be segmented; computing a normalized cross correlation of the received image against one or more of the biological structures in the set of biological structures; generating, based on computing, seed data representing an estimate of a spatial organization of the one or more biological structures in the received image; and segmenting, based on a normalized cross correction of the received image to the seed data, the received image into the one or more representations of the one or more biological structures.
1. A method performed by one or more processors, comprising:
receiving an image to be segmented into one or more representations of one or more biological structures;
accessing data representing a set of biological structures that are derived from other biological structures delineated in a training image, wherein the training image is associated with a level of modality that corresponds to a level of modality associated with the image to be segmented;
computing a normalized cross correlation of the received image against one or more of the biological structures in the set of biological structures;
generating, based on computing, seed data representing an estimate of a spatial organization of the one or more biological structures in the received image; and
segmenting, based on a normalized cross correlation of the received image to the seed data, the received image into the one or more representations of the one or more biological structures.
2. The method of claim 1 , wherein segmenting comprises:
using the normalized cross correlation of the received image to the seed data to spatially register one or more items of seed data to the one or more biological structures in the received image.
3. The method of claim 1 , wherein generating comprises:
computing quantity γ p (u) in accordance with:
γ
p
(
u
)
=
∑
x
J
(
x
)
W
p
(
x
-
u
)
J
_
(
u
)
W
_
p
wherein J(x) represents an approximate segmentation of the received image;
wherein each biological structure in the set of biological structures is represented as W p (x), with p=1, . . . , K;
wherein J (u) represents a sum of squared values of image J under a template positioned at u;
wherein W p represent a sum of squared values of a template p; and
wherein the method further comprises:
computing a detection probability image M(u) in accordance with M(u)=max p γ(u); and
identifying the seed data based on identification of pixel locations in detection probability image M that are greater than a threshold μ.
4. The method of claim 1 , wherein the received image comprises a microscopy image.
5. The method of claim 1 , wherein the one or more biological structures in the received image comprise one or more of a nucleus, a cell, a tissue, and an organ.
6. The method of claim 1 , further comprising:
receiving the training image; and
receiving data specifying a delineation of biological structures in the training image.
7. The method of claim 6 , further comprising:
generating, based on the delineation, a template representing a biological structure that is derived from features of the delineated biological structures;
applying a principal component analysis technique to the template and to variations of the template;
generating, based on application of the principal component analysis technique, a statistical model of spatial transformations representing possible variations in shape of the delineated biological structures in the training image; and
generating, based on application of the statistical model to the delineated biological structures in the received image, the set of biological structures.
8. The method of claim 1 , wherein the one or more biological structures represented in the received image comprises at least one of:
a nucleus of a U2OS cell;
a nucleus of a NIH3T3 cell; and
a nucleus of a synthetic cell.
9. One or more non-transitory machine-readable media configured to store instructions that are executable by one or more processors to perform operations comprising:
receiving an image to be segmented into one or more representations of one or more biological structures;
accessing data representing a set of biological structures that are derived from other biological structures delineated in a training image, wherein the training image is associated with a level of modality that corresponds to a level of modality associated with the image to be segmented;
computing a normalized cross correlation of the received image against one or more of the biological structures in the set of biological structures;
generating, based on computing, seed data representing an estimate of a spatial organization of the one or more biological structures in the received image; and
segmenting, based on a normalized cross correlation of the received image to the seed data, the received image into the one or more representations of the one or more biological structures.
10. The one or more non-transitory machine-readable media of claim 9 , wherein segmenting comprises:
using the normalized cross correlation of the received image to the seed data to spatially register one or more items of seed data to the one or more biological structures in the received image.
11. The one or more non-transitory machine-readable media of claim 9 , wherein generating comprises:
computing quantity γ p (u) in accordance with:
γ
p
(
u
)
=
∑
x
J
(
x
)
W
p
(
x
-
u
)
J
_
(
u
)
W
_
p
wherein J(x) represents an approximate segmentation of the received image;
wherein each biological structure in the set of biological structures is represented as W p (x), with p=1, . . . , K;
wherein J (u) represents a sum of squared values of image J under a template positioned at u;
wherein W p represent a sum of squared values of a template p; and
wherein the method further comprises:
computing a detection probability image M(u) in accordance with M(u)=max p γ(u); and
identifying the seed data based on identification of pixel locations in detection probability image M that are greater than a threshold μ.
12. The one or more non-transitory machine-readable media of claim 9 , wherein the received image comprises a microscopy image.
13. The one or more non-transitory machine-readable media of claim 9 , wherein the one or more biological structures in the received image comprise one or more of a nucleus, a cell, a tissue, and an organ.
14. The one or more non-transitory machine-readable media of claim 9 , wherein the operations further comprise:
receiving the training image; and
receiving data specifying a delineation of biological structures in the training image.
15. The one or more non-transitory machine-readable media of claim 14 , wherein the operations further comprise:
generating, based on the delineation, a template representing a biological structure that is derived from features of the delineated biological structures;
applying a principal component analysis technique to the template and to variations of the template;
generating, based on application of the principal component analysis technique, a statistical model of spatial transformations representing possible variations in shape of the delineated biological structures in the training image; and
generating, based on application of the statistical model to the delineated biological structures in the received image, the set of biological structures.
16. The one or more non-transitory machine-readable media of claim 9 , wherein the one or more biological structures represented in the received image comprises at least one of:
a nucleus of a U2OS cell;
a nucleus of a NIH3T3 cell; and
a nucleus of a synthetic cell.
17. An electronic system comprising:
one or more processors; and
one or more machine-readable media configured to store instructions that are executable by the one or more processors to perform operations comprising:
receiving an image to be segmented into one or more representations of one or more biological structures;
accessing data representing a set of biological structures that are derived from other biological structures delineated in a training image, wherein the training image is associated with a level of modality that corresponds to a level of modality associated with the image to be segmented;
computing a normalized cross correlation of the received image against one or more of the biological structures in the set of biological structures;
generating, based on computing, seed data representing an estimate of a spatial organization of the one or more biological structures in the received image; and
segmenting, based on a normalized cross correlation of the received image to the seed data, the received image into the one or more representations of the one or more biological structures.
18. The electronic system of claim 17 , wherein segmenting comprises:
using the normalized cross correlation of the received image to the seed data to spatially register one or more items of seed data to the one or more biological structures in the received image.
19. The electronic system of claim 17 , wherein generating comprises:
computing quantity γ p (u) in accordance with:
γ
p
(
u
)
=
∑
x
J
(
x
)
W
p
(
x
-
u
)
J
_
(
u
)
W
_
p
wherein J(x) represents an approximate segmentation of the received image;
wherein each biological structure in the set of biological structures is represented as W p (x), with p=1, . . . , K;
wherein J (u) represents a sum of squared values of image J under a template positioned at u;
wherein W p represent a sum of squared values of a template p; and
wherein the method further comprises:
computing a detection probability image M(u) in accordance with M(u)=max p γ(u); and
identifying the seed data based on identification of pixel locations in detection probability image M that are greater than a threshold μ.
20. The electronic system of claim 17 , wherein the received image comprises microscopy image.
21. The electronic system of claim 17 , wherein the one or more biological structures in the received image comprise one or more of a nucleus, a cell, a tissue, and an organ.
22. The electronic system of claim 17 , wherein the operations further comprise:
receiving the training image; and
receiving data specifying a delineation of biological structures in the training image.
23. The one electronic system of claim 22 , wherein the operations further comprise:
generating, based on the delineation, a template representing a biological structure that is derived from features of the delineated biological structures;
applying a principal component analysis technique to the template and to variations of the template;
generating, based on application of the principal component analysis technique, a statistical model of spatial transformations representing possible variations in shape of the delineated biological structures in the training image; and
generating, based on application of the statistical model to the delineated biological structures in the received image, the set of biological structures.
24. The electronic system of claim 17 , wherein the one or more biological structures represented in the received image comprises at least one of:
a nucleus of a U2OS cell;
a nucleus of a NIH3T3 cell; and
a nucleus of a synthetic cell.