Method and apparatus for detecting cell reprogramming
Disclosed herein are methods for identifying cells undergoing reprogramming and reprogrammed cells from a fluorescence microscopic image of one or more cells. According to some embodiments, the method includes an image processing step, a cell detection step, and, optionally, a clustering step.
1. A method for identifying cells undergoing reprogramming and reprogrammed cells from a fluorescence microscopic image of one or more cells, comprising the steps of,
(a) applying grayscale conversion and unsharp masking to the fluorescence microscopic image to obtain an enhanced image;
(b) converting the enhanced image to a binary image;
(c) identifying a cell-like ellipse boundary for cells in the binary image, wherein each cell-like ellipse boundary represents an isolated cell or a group of cells; and
(d) labeling each isolated cell or group of cells with a random color to obtain a plurality of colored components, wherein each color component represents one or more cells undergoing reprogramming or one or more reprogrammed cells;
wherein in the step (a), the unsharp masking is performed by applying an unsharp mask according to equation 1:
f s ( x,y )=| f ( x,y )− f ( x,y )* G σ | (equation 1),
where the fs(x,y) is the enhanced image, x and y are the image coordinates, * represents the image convolution, and Gσ is the Gaussian filter with the standard deviation σ and the filter size of 2σ+1.
2. The method of claim 1 , wherein the step (b) is performed according to equation 2:
g
(
x
,
y
)
=
{
1
f
s
(
x
,
y
)
≥
T
0
otherwise
,
(
equation
2
)
where T is the threshold for the fluorescence microscopy image.
3. The method of claim 1 , further comprising the steps of,
(e) evaluating the smoothness of each pixel of the colored components in relative to a local area of the pixel by,
(e-1) determining the magnitude of image gradients M(x,y) of each pixel according to equation 3:
M ( x,y )=√{square root over ( G x 2 +G y 2 )}⋅ (equation 3),
where G x =∂f/∂x and G y =∂f/∂y, Gx and Gy respectively represent the image gradients in x and y directions, and
(e-2) determining the smoothness σ(x,y) of each pixel according to equation 4:
σ
(
x
,
y
)
=
1
N
-
1
∑
x
∑
y
(
M
(
x
,
y
)
-
μ
)
2
,
(
equation
4
)
where N is the total number of pixels and μ is the mean pixel value in the local area; and
(f) determining the presence of a vague region R that satisfies the criterion of equation 5:
R ={( x,y ):σ( x,y )≥ T} (equation 5).
4. The method of claim 3 , wherein the method comprises performing steps (a) to (f) to a series of fluorescence microscopic images of said one or more cells to identify the beginning of the cell reprogramming.
5. A non-transitory, tangible computer-readable storage medium, encoded with computer-readable instructions for executing a method of claim 1 .
6. The non-transitory, tangible computer-readable storage medium of claim 5 , wherein the step (b) is performed according to equation 2:
g
(
x
,
y
)
=
{
1
f
s
(
x
,
y
)
≥
T
0
otherwise
,
(
equation
2
)
where T is the threshold for the fluorescence microscopy image.
7. The non-transitory, tangible computer-readable storage medium of claim 5 , wherein the method further comprises the steps of,
(e) evaluating the smoothness of each pixel of the colored components in relative to a local area of the pixel by,
(e-1) determining the magnitude of image gradients M(x,y) of each pixel according to equation 3:
M ( x,y )=√{square root over ( G x 2 +G y 2 )}⋅ (equation 3),
where G x =∂f/∂x and G y =∂f/∂y, Gx and Gy respectively represent the image gradients in x and y directions, and
(e-2) determining the smoothness σ(x,y) of each pixel according to equation 4:
σ
(
x
,
y
)
=
1
N
-
1
∑
x
∑
y
(
M
(
x
,
y
)
-
μ
)
2
,
(
equation
4
)
where N is the total number of pixels and μ is the mean pixel value in the local area; and
(f) determining the presence of a vague region R that satisfies the criterion of equation 5:
R ={( x,y ):σ( x,y )≥ T} (equation 5).
8. A system for identifying cells undergoing reprogramming and reprogrammed cells from a fluorescence microscopic image of one or more cells, comprising,
an apparatus configured to obtain a fluorescence microscopic image of one or more cells; and
a control unit, comprising a processor and a memory for storing a plurality of instructions which, when executed by the processor, causing the processor to perform a method of claim 1 .
9. The system of claim 8 , wherein the step (b) is performed according to equation 2:
g
(
x
,
y
)
=
{
1
f
s
(
x
,
y
)
≥
T
0
otherwise
,
(
equation
2
)
where T is the threshold for the fluorescence microscopy image.
10. The system of claim 8 , wherein the method further comprises the steps of,
(e) evaluating the smoothness of each pixel of the colored components in relative to a local area of the pixel by,
(e-1) determining the magnitude of image gradients M(x,y) of each pixel according to equation 3:
M ( x,y )=√{square root over ( G x 2 +G y 2 )}⋅ (equation 3),
where G x =∂f/∂x and G y =∂f/∂y, Gx and Gy respectively represent the image gradients in x and y directions, and
(e-2) determining the smoothness σ(x,y) of each pixel according to equation 4:
σ
(
x
,
y
)
=
1
N
-
1
∑
x
∑
y
(
M
(
x
,
y
)
-
μ
)
2
,
(
equation
4
)
where N is the total number of pixels and μ is the mean pixel value in the local area; and
(f) determining the presence of a vague region R that satisfies the criterion of equation 5:
R ={( x,y ):σ( x,y )≥ T} (equation 5).
11. A method for identifying and tracing cells undergoing reprogramming and reprogrammed cells from a fluorescence microscopic image of one or more cells, comprising the steps of,
(a) for every pixel of the fluorescence microscopic image, capturing an image of region of interest (ROI) of the pixel;
(b) applying a trained convolutional neural network (CNN) model to the ROI to calculate the respective probabilities of the pixel belonging to any of a plurality of classes, wherein each of the plurality classes indicate a cell clustering pattern of the ROI; and
(c) obtaining a plurality of probability maps that respectively indicate the probabilities of the plurality of classes at every pixel of the fluorescence microscopic image;
wherein in the step (b), the trained CNN model is established by a training method comprising the steps of,
(1) selecting a region of a training image as a template image;
(2) manually classifying the template image as belonging to one of the plurality of classes;
(3) producing a training set comprising a plurality of template images from a plurality of training images by repeating steps (1) and (2); and
(4) using the plurality of template images of the training set as inputs to train an CNN architecture to obtain the trained CNN model;
wherein in the step (3), the plurality of template images are divided into a first set of template images and a second set of template images, and in the step (4), the method comprising the steps of,
using the first set of template images to calculate a plurality of parameters of the CNN model;
using the second set of template images to calculate a plurality of error vectors of the plurality of parameters; and
using the error vectors to re-calculate the parameters.
12. The method of claim 11 , further comprising the step of,
(d) converting the fluorescence microscopic image into a gray-level image according to the plurality of probability maps.
13. The method of claim 12 , further comprising the step of,
(e) determining the conditions of reprogramming gray-level image.
14. The method of claim 11 , wherein the pixel size of the ROI is at least 256 by 256 pixels.
15. A non-transitory, tangible computer-readable storage medium, encoded with computer-readable instructions for executing a method of claim 11 .
16. A system for identifying and tracing cells undergoing reprogramming and reprogrammed cells from a fluorescence microscopic image of one or more cells, comprising,
an apparatus configured to obtain a fluorescence microscopic image of one or more cells; and
a control unit, comprising a processor and a memory for storing a plurality of instructions which, when executed by the processor, causing the processor to perform a method of claim 11 .