Method of segmenting single neuron images with high-dynamic-range thresholds and computer readable storage medium thereof
The method of segmenting single neuron images with high-dynamic-range thresholds of the present invention includes (a) preparing a biological tissue sample containing neurons and performing imaging to this sample to obtain a three-dimensional raw neuroimage; (b) deleting voxels in the three-dimensional raw neuroimage with signal intensities below a first signal intensity threshold to obtain a first thresholded image; (c) tracing the first thresholded image to obtain a first traced image; (d) calculating a structural importance score of every voxel in the first traced image to obtain a first structural importance score of every voxel; (e) gradually increasing the signal intensity threshold and repeating (b), (c) and (d) n−1 times; (f) summing up all the n structural importance scores of every voxel; (g) deleting voxels with summed structural importance score smaller than a pre-determined value from the raw image to obtain the segmented single neuron.
1. A method of segmenting single neuron images with high-dynamic-range thresholds, comprising:
preparing a biological tissue sample containing neurons and performing single three-dimensional imaging to said biological tissue sample containing neurons to obtain a single three-dimensional raw neuroimage;
filtering out voxels in said three-dimensional raw neuroimage whose signal intensities are below a first signal intensity threshold, a second signal intensity threshold, . . . , a n-th signal intensity threshold respectively by a filtering module to obtain a first filtered image, a second filtered image, . . . , a n-th filtered image;
tracing said first filtered image, said second filtered image, . . . , said n-th filtered image for their skeleton by a tracing module respectively to obtain a first traced image, a second traced image, . . . , a n-th traced image;
calculating structural importance scores of every voxel in said first traced image, said second traced image, . . . , said n-th traced image by a score calculating module respectively by utilizing an equation to obtain a first structural importance score, a second structural importance score, . . . , a n-th structural importance score of every voxel;
summing up said first structural importance score, said second structural importance score, . . . , and said n-th structural importance score of every voxel by a summing module to obtain a summed structural importance score of every voxel;
determining a structural importance score threshold; and
filtering out voxels in said three-dimensional raw neuroimage whose summed structural importance scores are below said structural importance score threshold by said filtering module to obtain a single neuron three-dimensional image, wherein n is a positive integer which is larger than 3.
2. The method of claim 1 , before the step of filtering out voxels in said three-dimensional raw neuroimage whose signal intensities are below a first signal intensity threshold, a second signal intensity threshold, . . . , a n-th signal intensity threshold respectively, further comprising detecting soma position in said three-dimensional raw neuroimage.
3. The method of claim 1 , wherein said equation comprises:
a
i
(
j
)
=
max
(
[
G
i
(
j
)
-
G
0
(
j
)
]
,
0
)
+
⌊
N
i
(
j
)
N
0
(
j
)
⌋
+
⌊
L
i
(
j
)
L
0
(
j
)
⌋
+
λ
i
(
j
)
wherein
G
0
(
j
)
=
max
(
⌈
G
0
(
1
)
×
(
1
-
t
j
t
max
)
⌉
,
1
)
,
wherein └ ┘ and ┌ ┐ are respectively Gaussian floor and ceil functions, L i (j) is length of i-th branch, N i (j) is number of offspring branches of the i-th branch, G i (j) is number of offspring generation of the i-th branch, G i (1) is at minimal signal intensity threshold equaled to a 75 percentile of a number of non-zero G i (j=1) , if this value is less than 20, G 0 (1) =20 is set, a i (j) is structural importance score of the i-th branch after j-th segmentation, t j is j-th signal intensity threshold, t max is maximal signal intensity threshold for all signal intensity thresholds t, N 0 (j) =3G 0 (j) , L 0 =6 micrometers.
λ
i
(
j
)
=
max
(
⌊
L
p
(
j
)
L
0
(
j
)
⌋
,
for
p
∈
Λ
i
(
j
)
,
G
i
(
j
)
<
G
0
(
j
)
and
N
i
(
j
)
<
N
0
(
j
)
)
is a score obtained from a length of longest downstream branch of the i-th branch under the j-th signal intensity threshold t j , where Λ i (j) is a set formed by all downstream branches of the i-th branch under the j-th signal intensity threshold t j , p is any one branch belonging to the Λ i (j) set.
4. The method of claim 1 , wherein said tracing comprises fast automatically structural tracing.
5. The method of claim 1 , wherein the step of determining a structural importance score threshold is based on said summed structural importance score of every voxel.
6. The method of claim 1 , wherein intervals among said first signal intensity threshold, said second signal intensity threshold, . . . , said n-th signal intensity threshold are determined from a quality of said single three-dimensional raw neuroimage.
7. The method of claim 1 , wherein n includes 50.
8. A non-transitory computer readable storage medium having stored thereon program instructions that, when executed by a processor, cause said processor to perform the steps of:
loading a single three-dimensional raw neuroimage;
filtering out voxels in said three-dimensional raw neuroimage whose signal intensities are below a first signal intensity threshold, a second signal intensity threshold, . . . , a n-th signal intensity threshold respectively to obtain a first filtered image, a second filtered image, . . . , a n-th filtered image;
tracing said first filtered image, said second filtered image, . . . , said n-th filtered image for their skeleton respectively to obtain a first traced image, a second traced image, . . . , a n-th traced image;
calculating structural importance scores of every voxel in said first traced image, said second traced image, . . . , said n-th traced image respectively by utilizing an equation to obtain a first structural importance score, a second structural importance score, . . . , a n-th structural importance score of every voxel;
summing up said first structural importance score, said second structural importance score, . . . , and said n-th structural importance score of every voxel to obtain a summed structural importance score of every voxel;
determining a structural importance score threshold; and
filtering out voxels in said three-dimensional raw neuroimage whose summed structural importance scores are below said structural importance score threshold to obtain a single neuron three-dimensional image, wherein n is a positive integer which is larger than 3.
9. The non-transitory computer readable storage medium of claim 8 , before the step of filtering out voxels in said three-dimensional raw neuroimage whose signal intensities are below a first signal intensity threshold, a second signal intensity threshold, . . . , a n-th signal intensity threshold respectively, further comprising detecting soma position in said three-dimensional raw neuroimage.
10. The non-transitory computer readable storage medium of claim 8 , wherein said equation comprises:
a
i
(
j
)
=
max
(
[
G
i
(
j
)
-
G
0
(
j
)
]
,
0
)
+
⌊
N
i
(
j
)
N
0
(
j
)
⌋
+
⌊
L
i
(
j
)
L
0
(
j
)
⌋
+
λ
i
(
j
)
wherein
G
0
(
j
)
=
max
(
⌈
G
0
(
1
)
×
(
1
-
t
j
t
max
)
⌉
,
1
)
,
wherein └ ┘ and ┌ ┐ are respectively Gaussian floor and ceil functions, L i (j) is length of i-th branch, N i (j) is number of offspring branches of the i-th branch, G i (j) is number of offspring generation of the i-th branch, G 0 (1) is at minimal signal intensity threshold equaled to a 75 percentile of a number of non-zero G i (j=1) , if this value is less than 20, G 0 (1) =20 is set, a i (j) is structural importance score of the i-th branch after j-th segmentation, t j is j-th signal intensity threshold, t max is maximal signal intensity threshold for all signal intensity thresholds t, N 0 (j) =3G 0 (j) , L 0 =6 micrometers.
λ
i
(
j
)
=
max
(
⌊
L
p
(
j
)
L
0
(
j
)
⌋
,
for
p
∈
Λ
i
(
j
)
,
G
i
(
j
)
<
G
0
(
j
)
and
N
i
(
j
)
<
N
0
(
j
)
)
is a score obtained from a length of longest downstream branch of the i-th branch under the j-th signal intensity threshold t j , where Λ i (j) is a set formed by all downstream branches of the i-th branch under the j-th signal intensity threshold t j , p is any one branch belonging to the Λ i (j) set.
11. The non-transitory computer readable storage medium of claim 8 , wherein said tracing comprises fast automatically structural tracing.
12. The non-transitory computer readable storage medium of claim 8 , wherein the step of determining a structural importance score threshold is based on said summed structural importance score of every voxel.
13. The non-transitory computer readable storage medium of claim 8 , wherein intervals among said first signal intensity threshold, said second signal intensity threshold, . . . , said n-th signal intensity threshold are determined from a quality of said single three-dimensional raw neuroimage.
14. The non-transitory computer readable storage medium of claim 8 , wherein n includes 50.