Method and device for determining activity of electrotransfected cells
Acquisition and processing of electrotransfected cell suspension images through optical microscopy, and train the cell detection model using the SSD model framework to automatically predict the number and activity of living cells, solving the high cost, low accuracy and operation of existing electrotransfected cell activity assays. Complex problems, efficient and accurate cell activity assays are achieved, and are suitable for the integration of electrotransfection equipment.
Patent Information
- Application Number
- CN202411035358.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-22
AI Technical Summary
The existing electrotransfected cell activity determination methods have high cost, low detection accuracy, complex operation and cannot be integrated into electrotransfection equipment. In particular, the accuracy of the trypan blue staining method depends on manual observation, which is prone to subjective bias. and weak accuracy.
The electrotransfected cell suspension image was obtained through optical microscope, Gaussian smoothing processing and discrete wavelet transformation were performed to generate coarse-scale images and fine-scale images. The cell detection model was trained using the SSD model framework, and the number and activity of living cells were automatically predicted, and self-calibration was adopted. and unsupervised training methods to reduce manual annotation work and improve model training efficiency and accuracy.
It realizes the electrotransfected cell activity assay with low cost, high detection accuracy and simple operation, and can be integrated into the electrotransfection equipment, improves the determination accuracy and efficiency, reduces the equipment requirements, and is suitable for closed-loop systems.
Smart Images

Figure CN119006388A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of information processing technology, and in particular to a method and device for measuring the activity of electrotransfected cells. Background Art
[0002] Electrofection technology is a technique that reversibly perforates the cell membrane through a high-intensity electric field, instantly increasing the permeability of the cell membrane and allowing it to absorb exogenous factors in the surrounding medium, thereby changing the properties of the cell. It is highly efficient and accurate. The purpose of electrofection is to enable the transfected parent cells to divide and produce new cells containing exogenous factors. Therefore, measuring cell activity after electrofection to ensure that the cells can divide and proliferate normally is an indispensable part of the electrofection process.
[0003] Existing methods for measuring cell activity by electrofection include metabolic assays, membrane permeability assays, and trypan blue staining cell counting. Among them, methods such as metabolic assays and membrane permeability assays require special measuring equipment, such as flow cytometers or fluorescence microscopes, which are not only costly but also cannot be integrated into electrofection equipment. Cell counting by trypan blue staining has the advantages of being simple and fast. Cell samples after trypan blue staining need to be observed under a microscope to estimate the proportion of live cells. However, trypan blue staining cell counting has the following defects:
[0004] First, after adding trypan blue to the cell suspension, if the observation time exceeds 3-5 minutes, its accuracy will be greatly reduced, which requires precise control of the observation time. Secondly, there are a lot of subjective judgments in the manual observation process. For example, the conclusions when cells are stained in small amounts are often biased by subjectivity, which reduces the confidence in the conclusions. Third, due to the differences in observation equipment, the accuracy of manual observation based on ordinary optical microscopes is weaker than that of manual observation under fluorescence microscopes, but fluorescence microscopes have the disadvantages of high prices and difficulty in integration compared to optical microscopes.
[0005] Therefore, there is an urgent need for an electro-transfection cell activity assay method that is low in cost, high in detection accuracy, simple to operate, and has low requirements on equipment. Summary of the invention
[0006] The invention provides a method and a device for measuring the activity of electrotransfected cells.
[0007] A method for measuring the activity of electrotransfected cells, comprising:
[0008] Obtain images of electrotransfected cell suspensions using an optical microscope;
[0009] The electro-transfected cell suspension image is subjected to Gaussian smoothing and discrete wavelet transform, respectively, to obtain a coarse-scale image I c and fine-scale image I f ; The coarse-scale image Ic Including color information of the image; the fine-scale image I f Including edge and texture information of the image;
[0010] According to the coarse-scale image I c and fine-scale image I f Train the initial cell detection model to obtain a trained cell detection model;
[0011] An image of an electro-transfected cell suspension to be detected is obtained, and the image of the electro-transfected cell suspension to be detected is input into the trained cell detection model, and the number of living cells and their corresponding cell activity are predicted by the model.
[0012] According to the electro-transfection cell activity assay method as described above, the initial cell detection model adopts the SSD model framework.
[0013] According to the electro-transfection cell activity determination method as described above, the coarse-scale image I c and fine-scale image I f Train cell detection to obtain a trained cell detection model including:
[0014] The coarse-scale image I c and fine-scale image I f Perform fusion to obtain a fused image;
[0015] The coarse-scale image I c Perform thresholding to obtain a binary mask M c ;
[0016] According to the binary mask M c And the fused image is used to train an initial cell detection model to obtain a trained cell detection model.
[0017] According to the electrotransfection cell activity determination method as described above, the binary mask M c and the fused image, training the initial cell detection model, and obtaining the trained cell detection model includes: training the initial cell detection model using the following loss function:
[0018]
[0019] Among them, M cj represents the primary mask M c The jth detection box in M pi represents the i-th detection box predicted by the SSD model, represents the traditional SSD-based target detection loss function, α ij is the weight coefficient.
[0020] According to the electrotransfection cell activity assay method as described above, the weight coefficient α ij The value of is calculated by the following formula:
[0021]
[0022] Among them, l ij Represents the detection box M pi and the detection box M cj The center distance, b i Represents the detection box M pi The side length, b j Indicates M cj The side length of .
[0023] According to the above-mentioned electrotransfection cell activity determination method, the cell activity is judged by the proportion of live cells, and the proportion of live cells is calculated by the following formula:
[0024]
[0025] N live Represents the number of live cell detection frames obtained in the microscope image, N dead represents the number of dead cell detection frames obtained in the microscope image; p live Represents the proportion of living cells.
[0026] An electrotransfection cell activity assay device, comprising:
[0027] An acquisition unit, used for acquiring an image of the electro-transfected cell suspension through an optical microscope;
[0028] A processing unit is used to perform Gaussian smoothing and discrete wavelet transform on the electro-transfected cell suspension image to obtain a coarse-scale image I c and fine-scale image I f ; The coarse-scale image I c Including color information of the image; the fine-scale image I f Including edge and texture information of the image;
[0029] A training unit is used to train the coarse-scale image I c and fine-scale image I f Train the initial cell detection model to obtain a trained cell detection model;
[0030] The detection unit is used to obtain the electro-transfected cell suspension image to be detected, and input the electro-transfected cell suspension image to be detected into the trained cell detection model, and predict the number of living cells and their corresponding cell activity through the model.
[0031] The electrotransfected cell activity determination method and device provided by the present invention respectively performs Gaussian smoothing and discrete wavelet transform on the electrotransfected cell suspension image to obtain a coarse-scale image I c and fine-scale image I f , and using the coarse-scale image I c and fine-scale image I f Train the cell detection model and use the trained cell detection model to predict the number of living cells and their corresponding cell activity. This method is not only low-cost, but also has high detection accuracy and easy operation, and can also be integrated into the electrotransfection equipment.
[0032] The method provided by the present invention adopts a coarse-scale image self-calibration method (that is, by c Perform thresholding to obtain a binary mask M c , and using the binary mask M c And fused images to train cell detection models), avoiding too much manual data annotation work, improving the efficiency of model training, and also ensuring the accuracy of model training. It relies entirely on the characteristics of the data itself for training, and the extracted features are clear and have reference value. The device provided by the present invention inputs an image of the sample under a microscope, which is compatible with microscopes and microscope cameras of various brands on the market, and outputs the proportion of live cells in the sample and a set of high-dimensional features generated incidentally, which can be used to represent the activity of cells after electroporation. In addition to low cost and simple operation, the system involved in the present invention is also easy to integrate and is suitable for integration into an electrotransfection instrument to form a closed-loop system. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0034] Figure 1 A flow chart of the method for determining the activity of electrotransfected cells provided in this aspect;
[0035] Figure 2 This is a flowchart for processing images of electro-transfected cell suspensions;
[0036] Figure 3 This is a flow chart of the cell detection model training of the present invention. DETAILED DESCRIPTION
[0037] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0038] Figure 1 The flow chart of the method for measuring the activity of electrotransfected cells provided in this aspect is as follows: Figure 1 As shown, the method comprises the following steps:
[0039] Step 1: Obtain images of electrotransfected cell suspension using an optical microscope.
[0040] Specifically, the image of the electro-transfected cell suspension can generally be obtained by a microscope or a microscope combined with a microscope camera, and the magnification of the microscope is 10 times. The magnification of 10 times can ensure that there are enough cells in a single field of view, so that the accuracy of the one-time count has statistical significance and can be used to represent the counting ratio of the sample. A lower magnification will also lead to excessive cell density, unclear details, increased recognition difficulty, and thus affect the detection accuracy of living cells. Therefore, the image of the electro-transfected cell suspension provided by the present invention is an image taken with a microscope magnification of 10 times.
[0041] Step 2: The electro-transfected cell suspension image is subjected to Gaussian smoothing and discrete wavelet transform respectively to obtain a coarse-scale image I c and fine-scale image I f ; The coarse-scale image I c Including color information of the image; the fine-scale image I f Including edge and texture information of the image.
[0042] Specifically, Figure 2 This is a flowchart for processing images of electro-transfected cell suspensions. Figure 2 As shown in the figure, the purpose of setting up the multi-scale preprocessing link is to process the low-magnification cell suspension image data into two different scale images: coarse / fine. Color information is an important feature to distinguish living cells from dead cells. The coarse scale information mainly focuses on the spatial distribution of color information. Coarse scale image I c The generation method is Gaussian smoothing, and its main function is to blur image details and retain the distribution of color features. Fine-scale image I f It aims to preserve image details, especially information such as cell edges that are essential for cell detection. The generation method is discrete wavelet transform, and the calculation method is shown in formula (1).
[0043]
[0044] I in (x,y) is the pixel value of the input image at [x,y], [M,N] is the size of the input image, is the wavelet coefficient. out The image after discrete wavelet transformation is reconstructed into 4 channels of image data I from data of different frequency bands. f .
[0045] Step 3: Based on the coarse-scale image I c and fine-scale image I f Train the initial cell detection model to obtain a trained cell detection model.
[0046] Specifically, the present invention adopts the SSD model as the basic framework of the initial cell detection model. The input of the model is the coarse-scale image I c and fine-scale image I f After the fusion image, the output of the model is two sets of detection boxes containing live cells and dead cells {(x i ,y i ,w i ,h i )},i∈N live and {(x i ,y i ,w i ,h i )}, i ∈N dead , where x i ,y i is the position coordinate of the upper left corner of the detection box, w i ,h i N is the length and width of the detection box. live and N dead are the numbers of live cell detection frames and dead cell detection frames obtained by SSD from microscope images, respectively.
[0047] Figure 3 This is a flow chart of the cell detection model training of the present invention, such as Figure 3 As shown, unlike the traditional fully supervised training method, the present invention adopts an unsupervised method to train the initial cell detection model. First, the thresholding method is used to train the initial cell detection model on the coarse scale image I c Generate the primary binary mask M c , using this as a label to construct a training data set (I c ,M c ) and drives the SSD model SSD(·,θ) to train to obtain the model parameters θ. In the process of supervised training, considering the primary mask M cThe generation principle of M has strong inaccuracy. The loss function used in the training of traditional target detection algorithms overemphasizes the overlap of detection boxes and labels, and introduces a large penalty term for non-overlapping areas. c Under the influence of , this operation will cause the model training process to oscillate or even fail to converge. Therefore, the present invention proposes to use the loss function shown in formula (2)
[0048]
[0049] Among them, M cj represents the primary mask M c The jth detection box in M pi Represents the i-th detection box predicted by the SSD model. Represents the traditional SSD-based target detection loss function. Weight coefficient α ij The value of is calculated by formula (3).
[0050]
[0051] l ij Represents the detection box M pi and M cj The center distance, b i and b j Represents the detection box M pi and M cj The output of the initial cell detection model will be used as a label to drive the model to self-train again until convergence, and finally a trained cell detection model will be obtained.
[0052] Step 4: Obtain an image of the electro-transfected cell suspension to be detected, and input the image of the electro-transfected cell suspension to be detected into the trained cell detection model, and use the model to predict the number of living cells and their corresponding cell activity.
[0053] Specifically, the present invention calculates the proportion of live cells by formula (4) to evaluate the cell activity after electroporation. At the same time, the high-dimensional quantitative features extracted by the cell detection model are output. The features can be used to assist in the evaluation of the electroporation effect.
[0054]
[0055] The electrotransfected cell activity assay method and device provided by the present invention collect images of the cell suspension after trypan blue staining under an optical microscope, and automatically identify and count live cells through an artificial intelligence-based target detection algorithm to evaluate the cell activity after electrotransfection.
[0056] The present invention also provides a device for measuring the activity of electrotransfected cells, comprising:
[0057] An acquisition unit, used for acquiring an image of the electro-transfected cell suspension through an optical microscope;
[0058] A processing unit is used to perform Gaussian smoothing and discrete wavelet transform on the electro-transfected cell suspension image to obtain a coarse-scale image I c and fine-scale image I f ; The coarse-scale image I c Including color information of the image; the fine-scale image I f Including edge and texture information of the image;
[0059] A training unit is used to train the coarse-scale image I c and fine-scale image I f Train the initial cell detection model to obtain a trained cell detection model;
[0060] The detection unit is used to obtain the electro-transfected cell suspension image to be detected, and input the electro-transfected cell suspension image to be detected into the trained cell detection model, and predict the number of living cells and their corresponding cell activity through the model.
[0061] The device provided by the present invention has an input of an image of a sample under a microscope, which is compatible with microscopes and microscope cameras of various brands on the market, and an output of the live cell ratio of the sample and a set of high-dimensional features generated incidentally, which can be used to represent the activity of cells after electroporation. In addition to low cost and simple operation, the system involved in the present invention is also easy to integrate and is suitable for integration into an electroporation instrument to form a closed-loop system.
[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for measuring the activity of electrotransfected cells, It is characterized in that include: Obtain images of electrotransfected cell suspensions using an optical microscope; The electro-transfected cell suspension image is subjected to Gaussian smoothing and discrete wavelet transform, respectively, to obtain a coarse-scale image I c and fine-scale image I f ; The coarse-scale image I c Including color information of the image; the fine-scale image I f Including edge and texture information of the image; According to the coarse-scale image I c and fine-scale image I f Train the initial cell detection model to obtain a trained cell detection model; An image of an electro-transfected cell suspension to be detected is obtained, and the image of the electro-transfected cell suspension to be detected is input into the trained cell detection model, and the number of living cells and their corresponding cell activity are predicted by the model.
2. The method for measuring the activity of electrotransfected cells according to claim 1, It is characterized in that The initial cell detection model adopts the SSD model framework.
3. The method for measuring the activity of electrotransfected cells according to claim 2, It is characterized in that The method according to the coarse-scale image I c and fine-scale image I f Train cell detection to obtain a trained cell detection model including: The coarse-scale image I c and fine-scale image I f Perform fusion to obtain a fused image; The coarse-scale image I c Perform thresholding to obtain a binary mask M c ; According to the binary mask M c And the fused image is used to train an initial cell detection model to obtain a trained cell detection model.
4. The method for measuring the activity of electrotransfected cells according to claim 3, It is characterized in that The binary mask M c And the fused image, training the initial cell detection model, and obtaining the trained cell detection model includes: using the following loss function to train the initial cell detection model: Among them, M cj represents the primary mask M c The jth detection box in M pi represents the i-th detection box predicted by the SSD model, represents the traditional SSD-based target detection loss function, α ij is the weight coefficient.
5. The method for measuring the activity of electrotransfected cells according to claim 4, It is characterized in that The weight coefficient α ij The value of is calculated by the following formula: Among them, l ij Represents the detection box M pi and the detection box M cj The center distance, b i Represents the detection box M pi The side length, b j Indicates M cj The side length of .
6. The method for measuring the activity of electrotransfected cells according to claim 1, It is characterized in that The cell activity is judged by the ratio of live cells, which is calculated by the following formula: N live Represents the number of live cell detection frames obtained in the microscope image, N dead represents the number of dead cell detection frames obtained in the microscope image; p live Represents the proportion of living cells.
7. An electrotransfection cell activity assay device, It is characterized in that include: An acquisition unit, used for acquiring an image of the electro-transfected cell suspension through an optical microscope; A processing unit is used to perform Gaussian smoothing and discrete wavelet transform on the electro-transfected cell suspension image to obtain a coarse-scale image I c and fine-scale image I f ; The coarse-scale image I c Including color information of the image; the fine-scale image I f Including edge and texture information of the image; A training unit is used to train the coarse-scale image I c and fine-scale image I f Train the initial cell detection model to obtain a trained cell detection model; The detection unit is used to obtain the electro-transfected cell suspension image to be detected, and input the electro-transfected cell suspension image to be detected into the trained cell detection model, and predict the number of living cells and their corresponding cell activity through the model.
