CN110838126A - Cell image segmentation method, device, computer equipment and storage medium - Google Patents

Abstract

The application relates to a cell image segmentation method, a cell image segmentation device, a computer device and a storage medium. The method comprises the following steps: reading a cell gray level image; performing histogram equalization operation on the cell gray level image to obtain an equalized image; performing morphological operation on the equalized image to obtain a morphological image; the morphological operations comprise a top hat operation and a gradient operation; detecting the cell edge in the morphological image through an edge detection algorithm to obtain a cell edge image; carrying out binarization processing on the cell edge image to obtain a binarized cell image; and carrying out cell segmentation according to the binary cell image to obtain a cell segmentation image. By adopting the method, the accuracy of cell image segmentation can be improved.

Description

Cell image segmentation method, device, computer equipment and storage medium

technical field

The present application relates to the field of image processing, and in particular, to a cell image segmentation method, device, computer equipment and storage medium.

Background technique

Screening of stable cell lines is of great significance in the field of biotechnology. The traditional stable cell line screening method is to detect the total amount of target protein secreted by cells into the medium, and screen high-yielding cell lines according to the protein expression. However, the above method has disadvantages such as complicated operation and long time-consuming. With the development of artificial intelligence technology, people begin to take cell line images with the help of high-power microscopes, and use deep learning and other methods to screen cell line images to overcome the above shortcomings. Image segmentation methods for cells.

The traditional cell image segmentation is based on the threshold method. However, the threshold method is greatly affected by the threshold setting, which easily leads to inaccurate image segmentation and cannot handle cell adhesion.

Therefore, traditional cell image segmentation methods have the problem of low image segmentation accuracy.

SUMMARY OF THE INVENTION

Based on this, it is necessary to provide a high-accuracy cell image segmentation method, a cell image segmentation device, a computer device, and a computer-readable storage medium for the above-mentioned technical problems.

A cell image segmentation method, comprising:

Read cell grayscale images;

performing a histogram equalization operation on the cell grayscale image to obtain an equalized image;

performing a morphological operation on the equalized image to obtain a morphological image; the morphological operation includes a top-hat operation and a gradient operation;

Detect the cell edge in the morphological image by an edge detection algorithm to obtain a cell edge image;

performing binarization processing on the cell edge image to obtain a binarized cell image;

Perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

In one embodiment, the cell segmentation is performed according to the binarized cell image to obtain a cell segmentation image, including:

When there are adherent cells in the binarized cell image, the number of cells in the adherent cells is calculated by the chain code method;

According to the number of cells in the adhesion cells, obtain the optimal minimum point in the H-minima transformation method;

Perform distance transformation on the binarized cell image according to the optimal minimum point to obtain a distance transformed image;

The distance transformed image is segmented using a watershed algorithm to obtain the cell segmented image.

In one embodiment, when there are adherent cells in the binarized cell image, the number of cells in the adherent cells is calculated by a chain code method, including:

Marking the chain code start position of the binarized cell image to obtain the chain code start mark;

Marking the position where the chain code direction changes in the binarized cell image to obtain a chain code change mark;

Marking the chain code end position of the binarized cell image to obtain the chain code end mark;

According to the chain code start marker, the chain code change marker and the chain code end marker, the number of cells in the adherent cells is calculated.

In one embodiment, the optimal minimum point in the H-minima transformation method is obtained according to the number of cells in the adherent cells, including:

Get the H-minima transform threshold;

According to the H-minima transformation threshold, the number of minimum points is obtained;

The optimal minimum point is obtained by comparing the number of minimum points with the number of cells in the adherent cells.

In one embodiment, performing binarization processing on the cell edge image to obtain a binarized cell image includes:

Perform thresholding on the cell edge image to obtain a thresholded image;

Through the opening operation, the non-cellular area in the thresholded image background is removed to obtain the opening operation image;

Through the closing operation, the inner edge of the cell in the opening operation image is removed to obtain the closing operation image;

When the inner edge of the cell still exists in the closed operation image, the closed operation image is filled with holes to obtain the binarized cell image.

In one embodiment, the cell segmentation is performed according to the binarized cell image to obtain a cell segmentation image, including:

Obtain cell contour information by searching the cell contour in the binarized cell image;

According to the cell outline information, the smallest rectangle containing the cell outline is obtained to obtain the cell segmentation image.

In one embodiment, after the step of performing cell segmentation according to the binarized cell image to obtain the cell segmentation image, the step includes:

From the cell segmentation image, collect cell information; the cell information includes cell size and cell internal structure;

According to the cell information, training sample data is obtained; the training sample data is used for cell feature identification and cell screening.

A cell image segmentation device, comprising:

Input module for reading cell grayscale images;

an equalization module, configured to perform a histogram equalization operation on the grayscale image of the cells to obtain an equalized image;

a morphology module, configured to perform a morphological operation on the equalized image to obtain a morphological image; the morphological operation includes a top-hat operation and a gradient operation;

an edge detection module for detecting the cell edge in the morphological image through an edge detection algorithm to obtain a cell edge image;

The binarization module is used for binarizing the cell edge image to obtain a binarized cell image;

A segmentation module, configured to perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

Read cell grayscale images;

performing a histogram equalization operation on the cell grayscale image to obtain an equalized image;

performing a morphological operation on the equalized image to obtain a morphological image; the morphological operation includes a top-hat operation and a gradient operation;

Detect the cell edge in the morphological image by an edge detection algorithm to obtain a cell edge image;

performing binarization processing on the cell edge image to obtain a binarized cell image;

Perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

A computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

Read cell grayscale images;

performing a histogram equalization operation on the cell grayscale image to obtain an equalized image;

performing a morphological operation on the equalized image to obtain a morphological image; the morphological operation includes a top-hat operation and a gradient operation;

Detect the cell edge in the morphological image by an edge detection algorithm to obtain a cell edge image;

performing binarization processing on the cell edge image to obtain a binarized cell image;

Perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

The above cell image segmentation method, device, computer equipment and computer-readable storage medium perform histogram equalization operation on the read cell grayscale image, which can enhance the image contrast and clarify the cell edge, so as to accurately detect the cell edge. Morphological operations on the equalized image can further clarify the edge of the cells and make the edge of the detected cells more accurate. The cell edge in the morphological image is detected by the edge detection algorithm, and the cell edge image is obtained. Since the cell edge image contains not only the outer contour of the cell, but also the inner contour of the cell, the cell edge image is further binarized to remove the cells. Internal outline. Perform cell segmentation according to the obtained binarized cell image. Since there is no interference from the inner contour of the cell, a cell segmentation image with high accuracy can be obtained, so as to achieve accurate segmentation of single-cell images. In the subsequent deep learning process, it can provide Massive high-quality training samples improve the accuracy of cell feature recognition and cell screening.

Description of drawings

1 is a schematic flowchart of a cell image segmentation method according to an embodiment;

Fig. 2 is an application environment diagram of a cell image segmentation method according to an embodiment;

3A/B/C/D/E are image processing result diagrams of each step of a cell image segmentation method according to an embodiment;

4 is an image segmentation result diagram of a cell image segmentation method according to an embodiment;

5A/B is a schematic diagram of a chain code method of a cell image segmentation method according to an embodiment;

Fig. 6 is the distance transformation result diagram of a kind of cell image segmentation method of one embodiment;

7 is a schematic diagram of a watershed algorithm for a cell image segmentation method according to an embodiment;

8 is a structural block diagram of a cell image segmentation device according to an embodiment;

FIG. 9 is an internal structure diagram of a computer device according to an embodiment.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

In one embodiment, as shown in FIG. 1, a cell image segmentation method is provided. The cell image segmentation method provided in this embodiment can be applied to the application environment shown in FIG. 2 . In this application environment, a user terminal 202 and a cell image segmentation server 204 are included. Wherein, the user terminal 202 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices capable of acquiring high-magnification cell images, and the cell image segmentation server 204 can be, but is not limited to, various types of image processing Functional personal computers, laptops, smartphones, tablets and portable wearables. The above-mentioned cell image segmentation method, which is applied to the cell image segmentation server 204 in FIG. 2 as an example, may include the following steps:

Step S102, reading the cell grayscale image.

The cell grayscale image may be an image with only grayscale values captured by a microscope on the cell. For example, the grayscale images of cells can be grayscale images of CHO (Chinese Hamster Ovary, Chinese Hamster Ovary) cell images at different stages.

In the specific implementation, CHO cell images of different periods are captured during the observation period, and the cell images are stored in the cell image segmentation server 204. When image segmentation processing is required, the grayscale of the CHO cell image is read from the cell image segmentation server 204. Degree Chart.

For example, 5 time points can be selected during the cell line culture process, and images of CHO cells can be photographed with a high-power microscope to obtain 5 grayscale images of cells, and these images can be stored in the cell image segmentation server 204. When cell image segmentation needs to be performed At the time, a grayscale image of cells is read from the cell image segmentation server 204 .

Step S104, performing a histogram equalization operation on the cell grayscale image to obtain an equalized image.

The histogram equalization operation may be an operation of equalizing the contrasts of different elements in the image through a histogram algorithm.

In a specific implementation, a histogram equalization operation can be performed on the cell grayscale image, the number of each grayscale value in a pixel is counted, the probability of each grayscale value occurring is calculated, and the grayscale value is mapped according to the probability. Through the histogram equalization operation, the image contrast can be enhanced and the edge of the cell can be sharpened, so that the edge of the cell can be detected accurately.

In practical applications, the above-mentioned histogram equalization operation can be achieved by the following formula:

Among them, the round() function represents the rounding operation, which rounds the number of decimal places of the argument in parentheses, M and N are the length and width of pixels in the cell grayscale image respectively, L represents the grayscale level, v is the pixel value in the grayscale image of the cell, cdf _min is the minimum value of the cumulative distribution function, and the cumulative distribution function cdf is calculated by the following formula

Among them, p _x (j) represents the probability of the occurrence of the pixel with gray level j, that is, the histogram of the image with the pixel value j, normalized to [0,1].

For example, for a cell grayscale image of 400×300 pixels, M=400, N=300, 8-bit depth is used, L is 2^8=256, and v is the specific grayscale value of each pixel, which can be 0~ Any value between 255. For each pixel in the cell grayscale image, an improved histogram equalization operation is performed according to the formula to obtain an equalized image.

Step S106, performing morphological operations on the equalized image to obtain a morphological image.

The morphological operation is an operation of connecting adjacent elements or separating adjacent elements into independent elements, and the morphological operation may specifically include a top-hat operation and a gradient operation.

In the specific implementation, the morphological top-hat operation can be used to highlight the cell outline in the large background of the cell image, and then the morphological gradient operation can be used to find the cell edge. The cell edge can be accurately extracted in the subsequent edge detection process.

In practical applications, src can be used to represent an equalized image, and element can be used to represent a structural element. The formula for performing a top-hat operation on an equalized image is:

dst=tophat(src, element)=src-open(src, element)=src-dilate(erode(src, element));

Among them, open(src, element) represents the morphological opening operation, src is first eroded and then dilated, and dilate(src, element) represents the dilation operation.

Then, the gradient operation is performed on the image dst, and the formula is

dst'=morph _grad (dst,element)=dilate(dst,element)-erode(dst,element);

Among them, erode(dst, element) represents the erosion operation on dst, and the obtained image dst' is a morphological image.

Step S108 , detect the cell edge in the morphological image through an edge detection algorithm to obtain a cell edge image.

Among them, the edge detection algorithm is an algorithm that can identify the optimal cell contour in the morphological image, and can be the Canny method.

In the specific implementation, the Canny method first denoises the morphological image through a Gaussian smoothing filter, then calculates the gradient intensity and direction of each pixel in the image, and eliminates spurs through the Non-Maximum Suppression method. In response, non-edge pixels were excluded, candidate edges were retained, and finally cell edges were determined by double-threshold. The above Canny method is not easily disturbed by noise, and can detect strong edges and weak edges respectively by using double thresholds. When weak edges and strong edges are connected, the output image contains weak edges, and the obtained edge detection results have high accuracy.

In practical applications, the Canny method can perform the following steps:

(1) Use the Gaussian filter to smooth the morphological image and filter out the noise. The Gaussian filter kernel generation formula with the size of (2k+1)*(2k+1) is:

(2) Calculate the gradient intensity and direction of each pixel in the morphological image;

(3) Through the non-maximum suppression method, the spurious response caused by edge detection is eliminated;

(4) Determine real and potential edges in morphological images by double-threshold detection;

(5) Complete edge detection by suppressing isolated weak edges to obtain cell edge images.

Step S110, performing binarization processing on the cell edge image to obtain a binarized cell image.

Among them, the binarization processing can be performed on the cell edge image, so that the output image only contains two colors of black and white, and can include threshold processing, opening operation, closing operation and hole filling.

In the specific implementation, since the interior of the CHO cell is relatively complex, the obtained cell edge image not only includes the outer contour of the cell, but also includes the inner contour of the cell. In order to perform cell segmentation according to the extracellular contour, first threshold the cell edge image to obtain a thresholded image, for example, set the threshold value to 100, and set the gray value of the pixel point whose gray value is higher than 100 in the cell edge image. is 255, and for pixels whose gray value is lower than or equal to 100, the gray value is 0. Perform the opening operation on the thresholded image, remove the non-cellular area in the background, and obtain the opening operation image. Then, perform the closing operation on the opening operation image, and remove the inner edge of the cells in the opening operation image to obtain the closing operation image. When the inner edge of the cell still exists in the closed operation image, the hole can be filled in the closed operation image, and the formula is:

X _k = (X _k-1 ⊕B)∩A ^c k=1,2,3...;

Among them, X ₀ is an image that is completely black and has a white pixel at the hole, B represents the structuring element, A ^c represents the complement of the closed operation image, and ⊕ represents that the B structuring element performs the expansion operation on X _k-1 .

Step S112: Perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

The cell segmentation is to segment the entire single cell according to the cell contour in the binarized cell image, and the obtained cell segmentation image can be the smallest rectangle containing the outer contour of the entire single cell.

In the specific implementation, you can call the findContours function in the opencv library to obtain the cell contour, call the boundingRect function to obtain the minimum outer rectangle of the contour, determine the cell area, and store the cell segmentation image.

In practical applications, the findContours function uses the principle of traversing the pixel value of each pixel to find the contour, and its formula is

contours, hierarchy=cv2.findContours(image,

cv2.RETR_TREE,;

cv2.CHAIN_APPROX_SIMPLE)

Among them, cv2 is the abbreviation of opencv, image is the input binarized cell image, cv2.RETR_TREE is a mode for retrieving contours - building a hierarchical tree structure contour, cv2.CHAIN_APPROX_SIMPLE represents an approximation method for contours - Compress the elements in the horizontal, vertical, and diagonal directions, and only retain the coordinates of the end point in this direction. For example, a rectangular contour only needs 4 points to save the contour information, and contours stores the point vector of each contour, contours[i] Represents the i-th contour, hierarchy represents the topological information of the contour, and the corresponding topological information of the contours[i] contour is hierarchy[i][0]～hierarchy[i][3], which respectively represent the next contour, the previous contour, Parent contour, the index of the embedded contour, if there is no corresponding item, the corresponding hierarchy[i] is set to a negative number. Then iterate over the contours[i].

The formula for the boundingRect function is

x,y,w,h=cv2.boundingRect(contours[i]);

Where x and y are the pixel coordinates of the upper left fixed point of the minimum circumscribed rectangle of the contour, w and h are the width and height of the rectangle, respectively, intercept this part, and traverse the number of pixels with a pixel value of 255 in the area as the area of the contour, Then, according to the area of the cell, the outline image of the appropriate cell is screened, and the minimum circumscribed rectangle image of the cell is saved.

It should be understood that although the various steps in the flowchart of FIG. 1 are shown in sequence according to the arrows, these steps are not necessarily executed in the sequence shown by the arrows. Unless explicitly stated herein, the execution of these steps is not strictly limited to the order, and the steps may be executed in other orders. Moreover, at least a part of the steps in FIG. 1 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed and completed at the same time, but may be executed at different times. The execution of these sub-steps or stages The sequence is also not necessarily sequential, but may be performed alternately or alternately with other steps or sub-steps of other steps or at least a portion of a phase.

The above cell image segmentation method performs histogram equalization operation on the read cell grayscale image, which can enhance the image contrast and clarify the cell edge, so as to accurately detect the cell edge. Morphological operations on the equalized image can further clarify the edge of the cells and make the edge of the detected cells more accurate. The cell edge in the morphological image is detected by the edge detection algorithm, and the cell edge image is obtained. Since the cell edge image contains not only the outer contour of the cell, but also the inner contour of the cell, the cell edge image is further binarized to remove the cells. Internal outline. The cell segmentation is performed according to the obtained binarized cell image. Since there is no interference from the inner contour of the cell, a cell segmentation image with high accuracy can be obtained.

Fig. 3 is an image processing result diagram of each step of a cell image segmentation method according to an embodiment, wherein Fig. 3A is a grayscale image of cells read in step S102; Fig. 3B is a histogram equalization process after step S104 , the obtained equalized image; Fig. 3C is the cell edge image obtained after the edge detection in step S108; Figs. 3D and 3E are the closed operation image and the binarized cell image obtained during the binarization process in step S110, respectively.

FIG. 4 is an image segmentation result diagram of a cell image segmentation method according to an embodiment. For each single cell in FIG. 3A , the minimum outer rectangle containing the outline of the single cell is obtained by segmentation, and the cell area can be calculated according to the result, and the segmentation result is saved. In the cell image segmentation server, it can be used for cell feature recognition and cell screening in deep learning.

In another embodiment, the above step S112 may specifically include: when there are adherent cells in the binarized cell image, calculating the number of cells in the adherent cells by a chain code method; according to the number of cells in the adherent cells, obtaining The optimal minimum point in the H-minima transformation method; according to the optimal minimum point, the binarized cell image is subjected to distance transformation to obtain a distance transformed image; the distance transformed image is segmented using the watershed algorithm to obtain cell segmentation image.

Among them, the adherent cells are two or more single cells that adhere to each other in the binarized cell image.

Among them, the chain code method is based on the Freeman chain code, and the cell adhesion image is marked. FIG. 5B shows a schematic diagram of the chain code when two single cells are adhered.

Among them, the optimal minimum point is the minimum point when the number of minimum points is equal to the number of adhesion cells in the H-minima transformation method.

In the specific implementation, when there are adherent cells, the Freeman chain code of the cell adhesion part in the binarized cell image is obtained. According to Figure 5A, when the 8-connected chain code is used, the eight directions of the pixel point start from 0 in the positive right direction. Counting, increasing counterclockwise, the Freeman chain code in Figure 5B is 455677567011231233.

To improve the Freeman chain code, the steps are as follows:

(1) select starting position, add mark 'B', as shown in Figure 5B;

(2) Traverse in a counterclockwise direction. When the directions of two adjacent chain codes are different, for example, the direction of the previous chain code is upper left, right left, and lower left, and the direction of the latter chain code is upper right, right, and lower right. , or the direction of the previous chain code is upper left, upper right, and upper right, and the direction of the next chain code is lower left, lower, and lower right, then insert 'C' in the two chain codes, as shown in the gray dots in Figure 5B shown;

(3) Traverse to the end of the chain code and add the mark 'E', as shown in Figure 5B.

Based on the improved Freeman chain code, the chain code output in Figure 5B is B4556C77C56C70112C3C12C33E. According to the experiment, it can be concluded that between 'B' and 'E', for each additional adherent cell, 4 'C's need to be marked, so that N _C is the number of 'C' in the improved Freeman chain code, and the number of cells in the adherent cells can be calculated as

According to the obtained number of adhesion cells, determine the h threshold value in the H-minima transformation. The h threshold value can be randomly selected. Each h threshold value corresponds to an image of a minimum point. When the number of minimum points in the image is equal to the number of adhesion cells When the number of h thresholds is determined, the h threshold value can be determined as the currently selected h threshold value, and the optimal minimum value point is obtained according to the determined h threshold value, and then a minimum value pixel point set is generated.

Next, the binarized cell image is converted to a distance transformed image by distance transform. The Euclidean distance transformation is used to calculate the minimum distance from the non-zero pixel point to the minimum value point in the binarized cell image, as the transformed value of the point. The distance formula is as follows:

Where (x, y) is the coordinate of the non-zero pixel value, and (i, j) is the coordinate of the minimum value point. The distance transformed image obtained by distance transformation is a grayscale image, as shown in Figure 6, in which the black part is two cells, and the white area is the edge of the "basin", which can be divided into two cells.

Finally, the distance-transformed image is segmented by the watershed algorithm, and the masker label of the distance-transformed image is obtained by calling the connectedComponents function in opencv, where the foreground (cell part) label is 1, the background is 0, and the water starts from the place where the label is 1. Let the water diffuse to find the final boundary, store the boundary as data, and segment the adherent cells according to the data to obtain the cell segmentation image shown in Figure 7.

The above cell image segmentation method, in view of the existence of adherent cells in the binarized cell image, accurately obtains the number of cells in the adherent cells through the chain code method, and determines the optimal minimum value in the H-minima transformation method according to the number of cells. value point, according to the optimal minimum value point, the binarized cell image is subjected to distance transformation to obtain a distance transformed image. The distance transformed image realizes the separation of adhesion cells. The watershed algorithm is used to segment the distance transformed image. Single-cell images are accurately segmented in the image.

In another embodiment, the above step S112 may further include: marking the starting position of the chain code in the binarized cell image to obtain the starting mark of the chain code; Mark the position of the chain code to obtain the chain code change mark; mark the end position of the chain code of the binarized cell image to obtain the chain code end mark; calculate the adhesion cell according to the chain code start mark, chain code change mark and chain code end mark the number of cells in .

由于步骤S112的上述处理过程在前述实施例中已有详细说明，在此不再赘述。In the specific implementation, as shown in Figure 5, on the basis of the traditional Freeman chain code method, first determine the starting position of the chain code, introduce the mark 'B' at the starting position of the chain code, and then traverse counterclockwise. When the direction of the two chain codes is different, the position where the direction of the chain code changes can be determined and marked as 'C'. Finally, when the end of the chain code is traversed, the position of the end of the chain code is determined and the mark 'E' is added. Count the number N _C of 'C' between 'B' and 'E', since each additional adherent cell increases the number of labeled 'C' by 4, the number of cells in the adherent cell can be calculated as

Since the above-mentioned processing process of step S112 has been described in detail in the foregoing embodiments, it will not be repeated here.

The above method marks the starting position of the chain code, the position where the direction of the chain code changes, and the end position of the chain code of the binarized cell image, respectively, and calculates the number of cells in the adherent cells according to the marks, so as to facilitate the use of distance transformation in subsequent steps. The method separates adherent cells and uses the watershed algorithm to accurately segment single-cell images from adherent cell images.

In another embodiment, the above step S112 may further include: obtaining an H-minima transformation threshold; obtaining the number of minimum points according to the H-minima transformation threshold; The number of cells is compared to obtain the optimal minimum point.

Among them, the H-minima transformation threshold is the h threshold in the H-minima transformation method.

In the specific implementation, the h threshold can be randomly selected, and each h threshold corresponds to an image of a minimum value point. When the number of minimum value points in the image is equal to the number of adhesion cells, the h threshold value can be determined as the currently selected h threshold value, According to the determined h threshold, the optimal minimum value point is obtained, and then a minimum value pixel point set is generated. Since the above-mentioned processing process of step S112 has been described in detail in the foregoing embodiments, it will not be repeated here.

The above method obtains the optimal minimum point according to the H-minima transform threshold, which is convenient to use the distance transform method to separate the adherent cells in the subsequent steps, and use the watershed algorithm to accurately segment the single-cell image from the adherent cell image.

In another embodiment, the above step S110 may specifically include: performing threshold processing on the cell edge image to obtain a thresholded image; removing non-cellular areas in the background of the thresholded image through an opening operation to obtain an opening operation image; operation, remove the inner edge of the cell in the open operation image, and obtain the closed operation image; when there are still internal cell edges in the closed operation image, the closed operation image is filled with holes to obtain a binarized cell image.

In the specific implementation, a threshold is set to binarize the cell edge image. For example, the threshold is set to 100. For the pixel points whose gray value is higher than 100 in the cell edge image, the gray value is set to 255. Pixels whose values are lower than or equal to 100 are assigned a grayscale value of 0. Then perform the opening operation to remove the non-cellular area in the background, and perform the closing operation to remove the inner edge of the cells in the opening operation image, and the obtained image is the closed operation image shown in Figure 3D. When there are still cells in the closed operation image When there is an inner edge, for example, the black dotted area inside the cell in Figure 3D, the closed operation image is filled with holes, resulting in a binarized cell image as shown in 3E. Since the above-mentioned processing process of step S110 has been described in detail in the foregoing embodiments, it will not be repeated here.

The above method distinguishes the cells in the cell edge image from the background through threshold processing, which is convenient for subsequent segmentation of the cell image. The open operation can remove the non-cellular area in the background area and avoid wrong segmentation of the background. The closed operation can remove the interior of the cell. Edge, when there is still an inner edge of the cell in the closed operation image, fill the hole in the closed operation image, which is convenient to obtain an accurate cell segmentation image for the outer edge of the cell.

In another embodiment, the above step S112 may further include: obtaining cell contour information by searching the cell contour in the binarized cell image; obtaining the smallest rectangle containing the cell contour according to the cell contour information to obtain the cell segmentation image.

In the specific implementation, for the binarized cell image shown in Figure 3E, the findContours function in the opencv library can be called to search for the contour by traversing the pixel value of each pixel to obtain the cell contour information, and then use the boundingRect function to obtain Outline the smallest outer rectangle, determine the cell area, and store the cell segmentation image. Since the above-mentioned processing process of step S112 has been described in detail in the foregoing embodiments, it will not be repeated here.

The above method searches for the cell contour in the binarized cell image, and the obtained cell contour information is relatively accurate. According to the cell contour information, the smallest rectangle containing the cell contour is obtained, and the obtained cell segmentation image has high accuracy.

In another embodiment, after the above step S112, the method may further include: collecting cell information from the cell segmentation image; the cell information includes cell size and cell internal structure; obtaining training sample data according to the cell information; the training sample data is used for Cell Characterization and Cell Screening.

Among them, the cell information is information such as the size and internal structure of a single cell.

In the specific implementation, information such as the size of each single cell and protein expression parameters can be collected from the cell segmentation image shown in Figure 4, and the collected information can be classified and sorted to obtain training sample data. When deep learning is required, etc. When performing cell feature identification or cell screening, training is performed based on the training sample data, and a cell feature identification model or cell screening model can be established according to different functions to be realized, and cell feature identification or cell screening can be performed based on the established model.

The above method collects cell information from cell segmentation images. Since the cell segmentation images have high accuracy, the collected cell information also has high accuracy. This information is used as a training sample, and based on the training sample, deep learning is used to perform cell features. Identification and cell screening, resulting in high accuracy and low computational time.

In one embodiment, as shown in FIG. 8, a cell image segmentation device 800 is provided, including: an input module 802, an equalization module 804, a morphology module 806, an edge detection module 808, a binarization module 810, and a segmentation module Module 812, where:

an input module 802, configured to read a grayscale image of cells;

an equalization module 804, configured to perform a histogram equalization operation on the cell grayscale image to obtain an equalized image;

Morphology module 806, configured to perform morphological operations on the equalized image to obtain a morphological image;

an edge detection module 808, configured to detect the cell edge in the morphological image through an edge detection algorithm to obtain a cell edge image;

The binarization module 810 is configured to perform binarization processing on the cell edge image to obtain a binarized cell image;

The segmentation module 812 is configured to perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

In one embodiment, the segmentation module 812 includes: when there are adherent cells in the binarized cell image, calculating the number of cells in the adherent cells by a chain code method; obtaining H-minima according to the number of cells in the adherent cells The optimal minimum point in the transformation method; according to the optimal minimum point, the binarized cell image is subjected to distance transformation to obtain a distance transformed image; the distance transformed image is segmented using the watershed algorithm to obtain a cell segmentation image.

In one embodiment, the segmentation module 812 further includes: marking the starting position of the chain code in the binarized cell image to obtain the starting mark of the chain code; Mark to get the chain code change mark; mark the chain code end position of the binarized cell image to get the chain code end mark; calculate the chain code start mark, chain code change mark and chain code end mark according to the chain code start mark number of cells.

In one embodiment, the segmentation module 812 further includes: obtaining an H-minima transform threshold; obtaining the number of minimum points according to the H-minima transform threshold; by comparing the number of minimum points with the number of cells in the adherent cells The numbers are compared to obtain the optimal minimum point.

In one embodiment, the binarization module 810 includes: performing thresholding on the cell edge image to obtain a thresholded image; removing non-cellular areas in the background of the thresholded image through an opening operation to obtain an opening operation image; performing a closing operation , remove the inner edge of the cell in the open operation image, and get the closed operation image; when there is still an inner edge of the cell in the closed operation image, fill the hole in the closed operation image to obtain the binarized cell image.

In one embodiment, the segmentation module 812 further includes: obtaining cell contour information by searching the cell contour in the binarized cell image; obtaining the smallest rectangle containing the cell contour according to the cell contour information to obtain the cell segmentation image.

In one embodiment, the segmentation module 812 further includes: collecting cell information from the cell segmentation image; the cell information includes cell size and cell internal structure; obtaining training sample data according to the cell information; the training sample data is used for cell feature recognition and cell screening.

For the specific definition of the cell image segmentation device, please refer to the definition of the cell image segmentation method above, which will not be repeated here. Each module in the above cell image segmentation device can be implemented in whole or in part by software, hardware and combinations thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

The cell image segmentation device provided above can be used to execute the cell image segmentation method provided in any of the foregoing embodiments, and has corresponding functions and beneficial effects.

In one embodiment, a computer device is provided, and the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. 9 . The computer equipment includes a processor, memory, a network interface, a display screen, and an input device connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The nonvolatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program implements an indoor positioning method of an air sensor when executed by a processor. The display screen of the computer equipment may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment may be a touch layer covered on the display screen, or a button, a trackball or a touchpad set on the shell of the computer equipment , or an external keyboard, trackpad, or mouse.

Those skilled in the art can understand that the structure shown in FIG. 9 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied. Include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, including a memory and a processor, a computer program is stored in the memory, and the processor implements the following steps when executing the computer program:

Read cell grayscale images;

Perform histogram equalization operation on the grayscale image of cells to obtain an equalized image;

Perform morphological operations on the equalized image to obtain a morphological image;

The cell edge in the morphological image is detected by edge detection algorithm, and the cell edge image is obtained;

Perform binarization processing on the cell edge image to obtain a binarized cell image;

Perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

In one embodiment, the processor further implements the following steps when executing the computer program: when there are adherent cells in the binarized cell image, calculating the number of cells in the adherent cells by a chain code method; according to the number of cells in the adherent cells , obtain the optimal minimum point in the H-minima transformation method; according to the optimal minimum point, perform distance transformation on the binarized cell image to obtain the distance transformed image; use the watershed algorithm to segment the distance transformed image to obtain Cell segmentation images.

In one embodiment, the processor further implements the following steps when executing the computer program: marking the starting position of the chain code in the binarized cell image to obtain the starting mark of the chain code; generating a chain code direction in the binarized cell image Mark the changed position to get the chain code change mark; mark the chain code end position of the binarized cell image to get the chain code end mark; calculate the chain code start mark, chain code change mark and chain code end mark according to the chain code start mark, chain code change mark and chain code end mark The number of cells in the adherent cells.

In one embodiment, when the processor executes the computer program, the following steps are further implemented: obtaining the H-minima transformation threshold; obtaining the number of minimum points according to the H-minima transformation threshold; Compare the number of cells in and get the optimal minimum point.

In one embodiment, the processor also implements the following steps when executing the computer program: performing threshold processing on the cell edge image to obtain a thresholded image; removing non-cellular areas in the background of the thresholded image through the opening operation to obtain the opening operation image; Through the closing operation, the inner edge of the cell in the open operation image is removed, and the closed operation image is obtained; when the closed operation image still has the inner edge of the cell, the closed operation image is filled with holes to obtain the binarized cell image.

In one embodiment, when the processor executes the computer program, the processor further implements the following steps: obtaining cell contour information by searching the cell contour in the binarized cell image; obtaining the smallest rectangle containing the cell contour according to the cell contour information to obtain cell segmentation image.

In one embodiment, the processor further implements the following steps when executing the computer program: collecting cell information from the cell segmentation image; cell information including cell size and cell internal structure; obtaining training sample data according to the cell information; for cell characterization and cell screening.

In one embodiment, a computer-readable storage medium is provided on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

Read cell grayscale images;

Perform histogram equalization operation on the grayscale image of cells to obtain an equalized image;

Perform morphological operations on the equalized image to obtain a morphological image;

The cell edge in the morphological image is detected by edge detection algorithm, and the cell edge image is obtained;

Perform binarization processing on the cell edge image to obtain a binarized cell image;

Perform cell segmentation according to the binarized cell image to obtain a cell segmentation image.

In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: when there are adherent cells in the binarized cell image, the number of cells in the adherent cells is calculated by a chain code method; number to obtain the optimal minimum point in the H-minima transformation method; according to the optimal minimum point, perform distance transformation on the binarized cell image to obtain the distance transformed image; use the watershed algorithm to segment the distance transformed image, Obtain cell segmentation images.

In one embodiment, the computer program further implements the following steps when executed by the processor: marking the starting position of the chain code in the binarized cell image to obtain the starting mark of the chain code; marking the chain code direction in the binarized cell image Mark the changed position to get the chain code change mark; mark the chain code end position of the binarized cell image to get the chain code end mark; according to the chain code start mark, chain code change mark and chain code end mark, Count the number of cells in the adherent cells.

In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: obtaining the H-minima transformation threshold; obtaining the number of minimum points according to the H-minima transformation threshold; The number of cells in the cell is compared to obtain the optimal minimum point.

In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: performing threshold processing on the cell edge image to obtain a thresholded image; removing non-cellular areas in the background of the thresholded image through the opening operation to obtain the opening operation image ; Through the closing operation, remove the inner edge of the cell in the open operation image to obtain the closed operation image; when the closed operation image still exists the cell inner edge, fill the hole in the closed operation image to obtain the binarized cell image.

In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: by searching the cell contour in the binarized cell image to obtain cell contour information; according to the cell contour information, obtaining the smallest rectangle containing the cell contour to obtain the cell contour Split the image.

In one embodiment, the computer program further implements the following steps when executed by the processor: collecting cell information from the cell segmentation image; cell information including cell size and cell internal structure; obtaining training sample data according to the cell information; training sample data For cell characterization and cell screening.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage In the medium, when the computer program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other medium used in the various embodiments provided in this application may include non-volatile and/or volatile memory. Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Road (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features It is considered to be the range described in this specification.

The above-mentioned embodiments only represent several embodiments of the present application, and the descriptions thereof are specific and detailed, but should not be construed as a limitation on the scope of the invention patent. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the scope of protection of the patent of the present application shall be subject to the appended claims.

Claims (10)

1. A method of cell image segmentation, comprising:

reading a cell gray level image;

performing histogram equalization operation on the cell gray level image to obtain an equalized image;

performing morphological operation on the equalized image to obtain a morphological image; the morphological operations comprise a top hat operation and a gradient operation;

detecting the cell edge in the morphological image through an edge detection algorithm to obtain a cell edge image;

carrying out binarization processing on the cell edge image to obtain a binarized cell image;

and carrying out cell segmentation according to the binary cell image to obtain a cell segmentation image.

2. The method according to claim 1, wherein said performing cell segmentation based on said binarized cell image to obtain a cell segmentation image comprises:

when the adhesion cells exist in the binarized cell image, calculating the number of the cells in the adhesion cells by a chain code method;

obtaining an optimal minimum value point in an H-minima transformation method according to the number of the cells in the adhesion cells;

according to the optimal minimum value point, performing distance transformation on the binary cell image to obtain a distance transformation image;

and segmenting the distance transformation image by using a watershed algorithm to obtain the cell segmentation image.

3. The method according to claim 2, wherein the calculating the number of cells in the adherent cells by a chain code method when adherent cells are present in the binarized cell image comprises:

marking the chain code initial position of the binary cell image to obtain a chain code initial mark;

marking the position of the chain code with changed direction in the binary cell image to obtain a chain code change mark;

marking the chain code ending position of the binary cell image to obtain a chain code ending mark;

and calculating the cell number in the adherent cells according to the chain code starting marker, the chain code change marker and the chain code ending marker.

4. The method as claimed in claim 2, wherein the obtaining the optimal minimum point in the H-minima transformation method according to the cell number in the adherent cells comprises:

acquiring an H-minima transformation threshold;

obtaining the number of minimum value points according to the H-minima conversion threshold;

and comparing the number of the minimum value points with the number of the cells in the adherent cells to obtain the optimal minimum value points.

5. The method according to claim 1, wherein the binarizing the cell edge image to obtain a binarized cell image comprises:

carrying out threshold processing on the cell edge image to obtain a thresholded image;

removing a non-cell area in the thresholding image background through an opening operation to obtain an opening operation image;

removing the inner edge of the cell in the opening operation image through closing operation to obtain a closing operation image;

and when the internal edges of the cells still exist in the closed operation image, filling holes in the closed operation image to obtain the binary cell image.

6. The method according to claim 1, wherein said performing cell segmentation based on said binarized cell image to obtain a cell segmentation image comprises:

searching a cell contour in the binary cell image to obtain cell contour information;

and acquiring a minimum rectangle containing the cell outline according to the cell outline information to obtain the cell segmentation image.

7. The method according to claim 1, wherein the step of performing cell segmentation based on the binarized cell image to obtain a cell segmentation image is followed by:

collecting cell information from the cell segmentation image; the cellular information includes cell size and cell internal structure;

obtaining training sample data according to the cell information; the training sample data is used for cell feature recognition and cell screening.

8. A cell image segmentation apparatus, comprising:

the input module is used for reading the cell gray level image;

the equalization module is used for carrying out histogram equalization operation on the cell gray level image to obtain an equalized image;

the morphological module is used for performing morphological operation on the equalized image to obtain a morphological image; the morphological operations comprise a top hat operation and a gradient operation;

the edge detection module is used for detecting the cell edges in the morphological image through an edge detection algorithm to obtain a cell edge image;

the binarization module is used for carrying out binarization processing on the cell edge image to obtain a binarization cell image;

and the segmentation module is used for carrying out cell segmentation according to the binary cell image to obtain a cell segmentation image.

9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the cell image segmentation method of any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the cell image segmentation method according to any one of claims 1 to 7.

Images