CN110910367A

CN110910367A - Bioreactor cell culture quality evaluation method

Info

Publication number: CN110910367A
Application number: CN201911136338.9A
Authority: CN
Inventors: 宫平; 郭红壮; 马辰昊; 葛辉琼; 张宁; 阚宝慧; 朱海焕; 吴昊; 李旭; 吉翔宇; 谭国桢; 郭旭
Original assignee: Changchun University of Science and Technology
Current assignee: Changchun University of Science and Technology
Priority date: 2019-11-19
Filing date: 2019-11-19
Publication date: 2020-03-24

Abstract

The invention discloses a method for evaluating cell culture quality of a bioreactor, which relates to the technical field of bioreactors and comprises image enhancement, image segmentation, smoothing treatment, connected region extraction, standard cell area judgment and cell distribution statistics, and automatic evaluation of cell culture quality can be completed through cell number statistics. The method can realize automatic evaluation of the growth state of the cultured cells, and provides a reliable guarantee means for promoting the application of the cell culture technology in a large scale in a cell factory.

Description

Bioreactor cell culture quality evaluation method

Technical Field

The invention relates to the technical field of bioreactors, in particular to a method for evaluating the cell culture quality of a bioreactor.

Background

The bioreactor is a device for obtaining a desired product through biological reaction or self metabolism by realizing in vitro culture by simulating in vivo growth environment of enzymes or organisms (such as cells, microorganisms and the like). The bioreactor plays an important role in vaccine production, monoclonal antibody preparation, medicine production, tumor prevention and treatment, wine brewing, biological fermentation, organic pollutant degradation and the like.

In the process of cell culture using a bioreactor, it is necessary to evaluate the quality of the cell culture. However, the existing evaluation method mainly depends on experience to judge the cell culture state, and no objective evaluation method exists, so that the objective and accurate evaluation result cannot be guaranteed.

Disclosure of Invention

The embodiment of the invention provides a method for evaluating the cell culture quality of a bioreactor, which can solve the problems in the prior art.

The invention provides a bioreactor cell culture quality evaluation method, which comprises the following steps:

enhancing useful information in the image by utilizing an image enhancement technology, and weakening or eliminating useless information;

segmenting the cells in the enhanced image from the background or other cells by adopting an image segmentation technology;

smoothing the segmented image, keeping the original important features in the segmented image, removing noise information and keeping the enhancement of image features;

and extracting a cell communication region from the smoothed image, and counting cells in the image by using a standard cell area judgment method and a cell distribution statistical method to realize the evaluation of the cell culture quality.

The method for evaluating the cell culture quality of the bioreactor comprises the steps of image enhancement, image segmentation, smoothing treatment, connected region extraction, standard cell area judgment and cell distribution statistics, and can finish automatic evaluation of the cell culture quality through the statistics of the number of cells. The method can realize automatic evaluation of the growth state of the cultured cells, and provides a reliable guarantee means for promoting the application of the cell culture technology in a large scale in a cell factory.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a schematic flow diagram of an enhancement algorithm;

FIG. 2 is a comparison graph of the effect of an image enhancement experiment;

FIG. 3 is a graph of the clustering effect for different numbers of clusters;

FIG. 4 is a graph of the clustering effect of different initial clustering centers;

FIG. 5 shows the results of segmentation of a microscopic image of a cell by different methods;

FIG. 6 is a schematic view of connected region labeling;

FIG. 7 is a flow chart of a method for standard cell determination based on statistics;

FIG. 8 shows the results of cell microscopic image counting experiments with Matlab software interface;

FIG. 9 is a graph showing evaluation of the counting accuracy of Vero cell images and SP20 cell images;

FIG. 10 is a graph showing the error fractions of a Vero cell image and an SP20 cell image;

FIG. 11 is a micrograph of a continuous acquisition SP20 cell culture process.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The invention provides a method for evaluating the cell culture quality of a bioreactor, which mainly comprises the following steps: firstly, the useful information in the image is highlighted by using an image enhancement technology, the useless information is weakened or even eliminated, then the cells in the image are separated from the background or other cells by using the image segmentation technology, the original important features in the segmented image are retained and the noise information is further removed by using the image smoothing processing, the image feature enhancement is kept, finally, the cell connected region is extracted from the smoothed image, and the cells are counted by using a standard cell judgment method and cell area distribution, so that the evaluation of the cell culture quality is realized.

Cell image enhancement technology based on Retinex

The traditional SSR and MSR algorithms strengthen reflected light components and filter incident light components, clear images can be obtained by the calculation method, but the contrast is reduced, and meanwhile, the defects of complex algorithm, large calculation data quantity, halo generation and the like exist. Aiming at the problems that the color brightness change of cells cultured in a bioreactor of a cell factory is small, and the processing effect is not obvious, the original method is improved, the invention provides a rapid enhancement algorithm, the complexity of the algorithm is reduced after the improvement, the real-time performance of the algorithm is improved, the good color constancy is kept, and the method has good application in cell image processing. The improved algorithm firstly directly makes a quotient of an original image and an incident component, namely an ambient brightness function, and obtains a reflection component, namely:

R(x,y)＝F(x,y)/I(x,y) (1)

the ambient brightness function belongs to a low-frequency component, which is a representation of the dynamic range of the image, and the reflection component R (x, y) is a high-frequency component, which belongs to the self-characteristics of the object and does not change with the change of the light intensity. The ambient brightness function I (x, y) is calculated by convolution of a gaussian function and an initial function, and then:

I(x,y)＝G(x,y)*F(x,y) (2)

as shown in FIG. 1, for an original image F, a low-pass filter is used to process the original image F to obtain an input image F₁Using a Gaussian function G and an input image F₁Convolution estimation of incident component I₁Then to the incident component I₁Correcting to obtain a judgment image I₂By inputting an image F₁And the incident component I₁Calculating the reflection component R by the quotient₁Optimizing the reflection component R₁Obtaining a modified image R₂Finally, calculating and judging the image I₂And modifying the image R₂Is obtained to obtain an enhanced image F₂. The contrast can be enhanced by correcting I (x, y), the object reflection component R (x, y) can be changed and optimized, and the image can be adjusted globally and locally at the same time, so that the aim of enhancing the local detail information of the image is fulfilled.

The improved algorithm reduces the complexity of the algorithm, adopts a rough calculation method for the ambient brightness function of the initial image, and accurately calculates the incident component and the reflection component by finding out a proper global contrast enhancement processing function and a proper local enhancement processing function, thereby finally realizing the image enhancement.

The optical imaging system is used for collecting microscopic images of irregular Vero cells, spherical SP20 cells and dense SP20 cells, the corresponding microscopic images are an original image A, an original image B and an original image C respectively, and the enhancement algorithm provided by the invention is applied to enhance the color image. The simulation result is shown in fig. 2, and the contrast and entropy value comparison between the improved algorithm and the conventional Retinex algorithm is shown in table 1.

TABLE 1 improved algorithm vs. conventional algorithm

Compared with the traditional SSR and MSR algorithms, the improved algorithm can improve the contrast, brightness and saturation of the initial image and better maintain the color information of the image according to the experimental results and the analysis of the table 1. The method can simultaneously enhance the information of the whole image, highlight the information of local areas, obviously distinguish the colors of cells and the background, solve the problem that the edges are easy to generate halos, and has small calculation amount of algorithm and high processing speed.

Cell microscopic image segmentation process

The image segmentation technology is one of the important steps of cell image processing, and its main purpose is to divide the region of the image that has strong correlation with the target, and its division strategy can be according to brightness, color, texture, etc. Region-based image segmentation methods, edge detection-based image segmentation methods, threshold-based image segmentation methods, and some new theory-based image segmentation methods are common image segmentation methods in the field of image processing. The method specifically comprises a region growing algorithm, a watershed algorithm, an edge detection algorithm, an Otus global threshold segmentation algorithm, a gray level statistical algorithm and the like. However, the cell image of the cell factory bioreactor in microscopic photoelectric monitoring is different from the general conventional image, the target of the cell is small, the image resolution is high, and meanwhile, the cell image is influenced by the cell morphology, the culture laminated plate, the culture medium, the illumination unevenness and the like, and the traditional method is difficult to obtain an ideal segmentation result.

The invention adopts a k-means clustering segmentation algorithm to segment the enhanced cell image, and the algorithm comprises the following contents:

(1) determination of the value of the number of clusters k

The clustering result and the operation speed are directly determined by the k value, the k value is inversely proportional to the clustering speed, and if a fine segmentation effect is required, the corresponding segmentation speed is reduced. On the contrary, if the segmentation speed is increased, the effect becomes fuzzy, the values of k are respectively 2, 3, 4, 5 and 6 by using the microscopic cell image, the clustering effect is shown in fig. 3, and the clustering time is shown in table 2.

TABLE 2 clustering times for different numbers of clusters

The clustering result shows that when k is 5, the segmentation result has the best effect and the clustering speed is high.

(2) Determination of a distance function

The distance function of the algorithm adopts Euclidean measurement, and the pixel value of a pixel point is compared with the pixel value of a cluster center to determine which cluster the pixel point belongs to.

(3) Determination of initial cluster center

The selection of the initial clustering center determines the segmentation effect, the steps of a Kaufman method, a nearest-farthest distance method and a gray level averaging method are respectively given to the selection of the initial clustering center, and verification experiments are carried out.

1) Kaufman method

The method selects a point with the shortest distance from a data center as an initial point, and then selects other points from other pixel points according to a heuristic principle. The realization process is as follows:

is provided with Z₁Is the initial classification center point, which is the point that is the shortest from the center of the data set.

Let x_iTo remove all points except the original classification center point, the following operations are performed:

calculating x_iT of_jiThe value:

T_ji＝max(D_j-d_ji,0) (3)

D_j＝min[dis(x_j,Z_s)](4)

d_ij＝dis(x_j,x_i)＝||x_j-x_i|| (5)

wherein Z is_sFor the determined cluster center, x_iCalculating

If x_iFurthest from the centre point and not used as centre point, i.e.

At a maximum, this point x_iAs the next classification center point.

And when all the k classification center points are determined, finishing the algorithm, and otherwise, continuously repeating the operation.

All points of the whole image are assigned to the closest classification center point, i.e. the point is clustered to the center point.

2) Method of nearest maximum distance

First, let all original class distances be the largest, then, the nearest and farthest distances are obtained from all data in the dataset to determine the original classification center. The realization process is as follows:

using a certain point in the data set as a classification center Z₁。

For x_i,Z_sIs processed as follows, wherein x_iRepresenting points of idleness in the data set, Z_sIndicating the points at which the classification has been determined.

Calculating d_si＝dis[Z_s,x_i]＝||Z_s-x_i||。

If d is_si<D_iThen d is_si＝D_i。

D_iRepresents a minimum distance, thisThe minimum distance is from the point to the center of the other categories.

Next cluster center selection distance D_iMaximum x_i。

And when the k clustering centers finish processing the algorithm, otherwise, continuously repeating the operation.

And after the k clustering centers are all selected, distributing each data in the data set to the clustering center belonging to the data set.

3) Method for averaging gray scale

According to the image histogram, the gray value of the image is distributed between 0 and 255, the gray level is divided into k classes in the interval, initially, the mean value of each class is used as a clustering center, and the initial clustering center is determined by the gray value of the whole image.

The Kaufman method has the advantages that when the image is small, although the initial time is longest, the iteration times are the smallest, and the result is the closest. The gray-scale average method has high initial speed but high iteration times. The maximum-minimum distance method is relatively fast in initial speed. The initial speed of the Kaufman method is very low when the image is large, the advantages of the maximum and minimum distance method are obvious, the clustering result effect of the maximum and minimum distance method is almost the same, but the initialization speed of the maximum and minimum distance method is high when the image is large, and the initial clustering center can be calculated. In order to verify the clustering effect of the three algorithms, the three methods are respectively adopted for experiments, the clustering effect is shown in fig. 4, and the clustering time is shown in table 3.

TABLE 3 clustering time determination algorithm for different initial clustering centers

From subjective effects and processing time, the three algorithms are basically similar, the clustering time of the gray-scale average value method is shortest, and finally an initial clustering center in the k-means clustering segmentation algorithm is selected by the method.

The clustered and segmented images still have some prominent noise points and some gaps caused by boundary over-segmentation, further smoothing processing is needed, and opening and closing operations combining morphological corrosion and expansion are selected in the processing. After treatment, the adhesion of the narrow part of the cell boundary is cut off, irregular prominent points are removed, then the gap of the boundary is completed, and fine holes are filled, finally, a median filtering method is adopted to continuously remove the minimal noise points, and the redundant connected domain is eliminated. Obtain clear, well-defined and ideal cell microscopic segmentation image.

Results and analysis of the experiments

Firstly, a cell microscopic image is processed by utilizing a Retinex model rapid enhancement algorithm based on contrast correction, and then the image is segmented. Fig. 5 shows the simulation results, and table 4 shows the algorithm run times.

TABLE 4 run time comparison of the image segmentation algorithm of the present invention with the conventional algorithm

In the experiment, the first method adopts an image segmentation method based on region growing and watershed algorithm, and the second method adopts a segmentation method of fuzzy c-means clustering.

The experimental result shows that the image segmentation effect by using the algorithm of the invention is better, the cell gap is more obvious, the processed sample image has smooth boundary and no small noise point, and the subsequent cell accuracy calculation is not affected basically by the phenomenon of few cell adhesion areas, wherein the number of the adhesion cells is far less than the number of the standard cells.

Microscopic image cell count

The microscopic image cell counting is the key of quality evaluation and analysis in the cell culture process, and a plurality of cell communication areas exist in the cell microscopic image segmentation result, so the cell counting evaluation is completed by adopting an improved rapid cell communication area judgment method, a standard cell judgment method and area-based cell number statistics.

The traditional method for marking the image connected region is mainly a four-neighborhood and eight-neighborhood scanning search method, wherein the latter method is more common, and the algorithm has the defects that other eight pixel points around each pixel point neighborhood need to be traversed, so that the finally determined connected region is marked with a unique number, and repeated search of the pixel points or the neighborhoods in the traversing process causes a large amount of calculation.

Aiming at the problem, the invention provides a connected region labeling algorithm with higher operation speed, compared with the traditional eight-neighborhood scanning method, the new algorithm can reduce the operation times, improve the operation speed and avoid the label missing problem of other rapid labeling algorithms. And converting the smoothed target image into a binary image, marking each connected region, counting the size of each connected region, and storing the size of each connected region into a program array.

The improved method for marking the connected region comprises the following steps:

(1) as shown in fig. 6(a), the target image is scanned from top to bottom and from left to right, and the cell boundary is the first unnumbered pixel point with binary value of 1.

(2) Recording the pixel point as a starting point, scanning eight neighborhoods of the pixel point, wherein the scanning sequence is up, right down, left up, finding the next unnumbered pixel point with binary value of 1, and recording the pixel point as the starting point to continue scanning. And stopping scanning until no unnumbered pixel point with binary value of 1 can be found in the eight-neighborhood.

(3) In the scanning process of step (2), marking all the points with the scanning sequence as the upper one, as shown in fig. 6(a), when the marked point is located at the lowest part of the connected region, repeating step (2) with the marked point as the starting point, as shown in fig. 6(b), until no pixel point with the unnumbered number and the binary value of 1 can be found, and at this time, finding all the pixel points in the connected region.

(4) And (3) after the connected region marking is finished, skipping the connected region and continuously executing the step (1) to find a new unnumbered pixel point with the binary value of 1 as the starting point of the next connected region. And (5) repeating the step (2) and the step (3) until all connected regions of the whole image are marked.

Standard cell judgment and area distribution cell number statistics based on mathematical statistics

At present, the judgment of the standard cell area is carried out by visually selecting the standard area or selecting the middle value or the average value of the areas of all connected areas in a descending order based on the existing software. The manual selection of the standard cell area cannot realize automatic evaluation, and the selection of the intermediate value or the average value is greatly influenced by over-segmentation of cells, noise or cell adhesion. Therefore, the invention provides a standard cell area judgment method based on mathematical statistics, which has the advantage that the standard cell area can be well judged as long as the number of independent cells is greater than the number of adhesion areas for a segmentation image with relatively good effect.

According to the magnification of the microscopic optical system and the size of the cultured cells, the standard cell area is basically and completely distributed in the interval of 200-3200 pixels, and the connected area smaller than 200 pixels is an over-segmentation area or a noise area, which is not counted. In actual culture, no overlarge cell individual exists, and the connected region larger than 3200 pixels is an adherent cell or a dead cell shedding region, so that statistics is not carried out. The proposed decision method is shown in fig. 7.

The method comprises the following implementation steps:

(1) dividing the connected region into several regions according to the number of pixels, wherein the number of pixels of the single connected region is between 200 and 3200, and every [3000N ] pixels is set as an order of magnitude, wherein [3000N ] represents the maximum integer less than or equal to 3000N. The division into N intervals, the initial N was set to 30, and the standard cells were substantially all in the N intervals.

(2) Distributing each connected region into each interval according to the pixel number by using a mathematical statistics method, and enabling the ith interval to be n_iAnd i is more than or equal to 1 and less than or equal to N. The interval with the maximum number of connected areas is the jth interval, and the number is n_j。

(3) Judging the following formula:

n_j-1+n_j+n_j+1>n_i-1+n_i+n_i+1，j≠i (3)

(4) if the formula (3) is satisfied, the average value of all the pixel numbers of the connected regions in the j-1, j, j +1 th interval is obtained and is used as the standard cell area. If equation (3) is not satisfied, the number of intervals N is decremented by 1 and assigned to N. Repeating steps (1) to (4) until a standard cell area is selected.

The method can automatically calculate the optimal standard cell area aiming at different cells. And judging all connected regions after obtaining the area of the standard cells, counting the number of the cells if the area of a single connected region is less than 0.5 time of the area of the standard cells, wherein the area of the single connected region is between 0.5 and 1.5 times of the area of the standard cells, counting 1 cell, and taking the multiple of the area of the connected region relative to the area of the standard cells as the number of the cells, and rounding up.

Results and analysis of the experiments

The treated micro-segmented images of the cells were subjected to connected region calibration, standard cell determination and cell number statistics using area distribution, respectively, and the counting results are shown in fig. 8.

The running time of the cell microscopic image counting algorithm and the accuracy of cell counting are shown in table 5.

TABLE 5 comparison table of cell microscopic image counting algorithm running time and cell counting accuracy

Experimental results show that the cell microscopic image counting method can realize counting of cells with different shapes, and the processing time is less than 0.5 second. To verify the accuracy of the method, 20 Vero cell images and 20 SP20 cell images were expert labeled. And (3) binarizing the sample image by using the cell area as a positive sample and the background area as a negative sample, wherein the pixel value of the positive sample is 255 and the pixel value of the negative sample is 0.

The method is tested by adopting two indexes of true positive rate TPR and false positive rate FPR. The true positive rate TPR represents the ratio of the number of correctly detected cells to the total number detected, and specifically:

wherein TP indicates the number of correctly detected cells and FN indicates the number of missed cells.

The false positive rate FPR represents the ratio of the number of false detection cells and missed detection cells to the total detection number, and specifically comprises the following steps:

wherein FP represents the number of misdetected cells.

The true positive rate TPR and the false positive rate FPR data obtained by cell counting of 20 Vero cell images and 20 SP20 cell images by the method of the present invention are shown in FIG. 9.

In the 20 Vero cell image counts TPR and FPR data, the mean TPR was 95.87% and the mean FPR was 6.73%. Of the 20 SP20 cell image counts TPR and FPR data, the mean TPR was 95.89% and the mean FPR was 5.67%.

The error fraction ME is adopted to objectively evaluate the cell growth area ratio in the algorithm, the error fraction ME can reflect the proportion of the negative sample pixel wrongly distributed to the positive sample region or the proportion of the positive sample pixel wrongly distributed to the negative sample region, the ratio of the difference between the segmentation result and the expert marking result can be measured, the closer the value is to 0, the more accurate the algorithm is, the more can be expressed as:

wherein, B_GMarking the expert with a positive sample, F_GMarking experts with negative examples, B_OComputing positive samples for the algorithm, F_ONegative examples are calculated for the algorithm.

The error fractions ME were calculated by comparing the growth areas calculated for 20 Vero cell images and 20 SP20 cell images with expert markers, and the data are shown in fig. 10.

The average ME in the 20 Vero cell image miscalculation ME data was 5.24%, and the average ME in the 20 SP20 cell image miscalculation ME data was 4.86%.

As can be seen from three evaluation modes of TPR, FPR and ME, the cell microscopic image counting method has high counting accuracy which is more than 95%, and the percentage evaluation of the growth area is reasonable in the aspect of fraction error.

The quality of monitoring images of SP20 cells in the continuous culture process was evaluated by the method of the present invention, and from the beginning of the passage, one monitoring image was selected every 12 hours, as shown in FIG. 11, and the results of the calculation of the number of cells and the growth area by the method of the present invention are shown in Table 6.

TABLE 6 trend chart of cell number and growth area during continuous culture

The experimental result accords with the cell growth trend, and the superiority of the method is verified. The method of the invention makes a certain breakthrough in the aspects of rapidity and accuracy of cell counting, and weakens the statistical error caused by the influence of the quality of the segmented image. The method has wide application range, can be applied to sparse cells, has good effect on dense cell counting, can be suitable for cell images of different types and shapes, and realizes quality evaluation of the continuous cell culture process.

While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims

1. The method for evaluating the cell culture quality of the bioreactor is characterized by comprising the following steps of:

2. The method of claim 1, wherein the enhancing the image by the image enhancement technique specifically comprises:

for the original image F, processing the original image F by adopting a low-pass filter to obtain an input image F₁Using a Gaussian function G and an input image F₁Convolution estimation of incident component I₁Then to the incident component I₁Correcting to obtain a judgment image I₂By inputting an image F₁And the incident component I₁Calculating the reflection component R by the quotient₁Optimizing the reflection component R₁Obtaining a modified image R₂Finally, calculating and judging the image I₂And modifying the image R₂Is obtained to obtain an enhanced image F₂。

3. The method for evaluating the cell culture quality of a bioreactor according to claim 1, wherein the enhanced image is segmented by a k-means clustering segmentation algorithm, and the k-means clustering segmentation algorithm is specifically as follows:

the clustering number is 5, the Euclidean distance is used as a distance function, the mean value of each class is used as an initial clustering center initially, and further smoothing processing is carried out by opening and closing operation combining morphological corrosion and expansion after segmentation.

4. The method for evaluating the quality of cell culture in a bioreactor according to claim 1, wherein the method for extracting the cell connected region in the image comprises:

step 1, scanning a target image from top to bottom and from left to right, and taking a scanned first unnumbered pixel point with binary value of 1 as a cell boundary;

step 2, recording cell boundary pixel points as initial points, carrying out eight-neighborhood scanning on the cell boundary pixel points, wherein the scanning sequence is up, right down, left up, finding the next unnumbered pixel point with the binary value of 1, recording the pixel point as the initial point and continuing scanning until the unnumbered pixel point with the binary value of 1 cannot be found in the eight-neighborhood, and stopping scanning;

step 3, marking all points with the scanning sequence as the upper points in the scanning process of the step 2, and repeating the step 2 by taking the marked points as the starting points when the marked points are positioned at the lowest part of the communicated area respectively until no unmarked pixel points with the binary value of 1 can be found;

and 4, after the marking of the connected region is finished, skipping the connected region, continuing to execute the step 1, searching a new unnumbered pixel point with the binary value of 1 as the starting point of the next connected region, and repeating the step 2 and the step 3 until all the connected regions of the whole image are marked.

5. The method for evaluating the cell culture quality of a bioreactor according to claim 1, wherein the standard cell area determination method comprises:

step 1, dividing a connected region into a plurality of regions according to the number of pixels, wherein the number of the pixels of the independent connected region is between 200 and 3200, and every [3000/N ] pixel is set as an order of magnitude, wherein [3000/N ] represents the maximum integer less than or equal to 3000/N, and N represents the number of the regions;

step 2, distributing each connected region to each interval according to the pixel number by using a mathematical statistics method, and enabling the number of the connected regions in the ith interval to be n_iI is more than or equal to 1 and less than or equal to N, the interval with the largest number of connected areas is the jth interval, and the number of the connected areas is N_j；

Step 3, judging the following formula:

n_j-1+n_j+n_j+1>n_i-1+n_i+n_i+1，j≠i

step 4, if the above formula is satisfied, calculating the average value of all the pixel numbers of the connected regions in the j-1, j, j +1 th interval, and taking the average value as the standard cell area; if not, the interval number N is reduced by 1, and the steps 1-4 are repeated until the standard cell area is selected.

6. The method for evaluating the cell culture quality of a bioreactor according to claim 5, wherein the statistical method of cell distribution comprises:

and judging all communication areas after obtaining the area of the standard cell, counting the number of the cells if the area of a single communication area is smaller than 0.5 time of the area of the standard cell, wherein the area of the single communication area is between 0.5 and 1.5 times of the area of the standard cell, counting 1 cell, the area of the single communication area is larger than 1.5 times of the area of the standard cell, and taking the multiple of the area of the communication area relative to the area of the standard cell as the number of the cells and rounding up.