A microfluidic chip-based cell multi-modal analysis system and method

Abstract

The application discloses a cell multi-modal analysis system and method based on a micro-fluidic chip, and the system comprises a data acquisition and processing module, a cell classification module and a cell three-dimensional reconstruction module.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cell analysis, and in particular to a cell multi-modal analysis system and method based on a microfluidic chip. BACKGROUND

[0002] In the field of cell analysis, microfluidic chip technology has become an important platform for cell detection due to its high throughput and low sample consumption. Cell multi-modal analysis (such as cell classification, cell three-dimensional reconstruction, and cell deformation analysis) using microfluidic chips is of great significance to the optimization design of microfluidic chips and medical research.

[0003] However, the existing technology requires independent platforms or devices to perform cell classification, cell three-dimensional reconstruction, and cell deformation analysis, resulting in high cost and long analysis period for cell multi-modal analysis. Moreover, the existing technology only uses bright-field images to locate cells, and does not fully apply them to cell classification, cell three-dimensional reconstruction, and cell deformation analysis, further reducing the efficiency and accuracy of cell multi-modal analysis.

[0004] To address the above problems, no effective solutions have been proposed so far. SUMMARY

[0005] The embodiments of the present specification provide a cell multi-modal analysis system and method based on a microfluidic chip to solve the problem that the existing technology cannot accurately, quickly, and low-costly perform cell multi-modal analysis.

[0006] In a first aspect, the embodiments of the present specification provide a cell multi-modal analysis system based on a microfluidic chip, comprising:

[0007] a data acquisition and processing module configured to input a mixed cell sample, a first cell sample of a target category filtered by a cell classification module, and a reconstructed second cell sample output by a cell three-dimensional reconstruction module into a cell flow channel of a microfluidic chip respectively, and acquire a first bright-field image, a second bright-field image, and a third bright-field image respectively;

[0008] a cell classification module configured to receive the first bright-field image, output a cell category probability through a cell classification model, filter the first cell sample of the target category, and feed back to the data acquisition and processing module;

[0009] a cell three-dimensional reconstruction module configured to receive the second bright-field image, output the reconstructed second cell sample through a cell reconstruction model, and feed back to the data acquisition and processing module;

[0010] a cell deformation analysis module configured to receive the third bright-field image, output deformation data through a cell deformation model, and feed back to the data acquisition and processing module.

[0011] In some embodiments, the data acquisition and processing module is further configured to send the first bright field image to a cell classification module, send the second bright field image to a cell three-dimensional reconstruction module, and send the third bright field image to a cell deformation analysis module; and the data acquisition and processing module is further configured to, after receiving the reconstructed second cell sample, set an obstacle in the cell flow channel, input the reconstructed second cell sample into the cell flow channel after the obstacle is set, and collect a bright field image of the reconstructed second cell sample after the obstacle, as the third bright field image.

[0012] In some embodiments, the cell flow channel comprises an excitation area, the excitation area comprises a left excitation area and a right excitation area, respectively receiving excitation light of a first wavelength and excitation light of a second wavelength, the excitation light of the first wavelength being used to excite the first antibody dye in the flowing cell sample and emit light, and the excitation light of the second wavelength being used to excite the second antibody dye in the flowing cell sample and emit light.

[0013] In some embodiments, the cell classification module comprises a first trainer, the first trainer being configured to train a cell classification model, and the cell classification model being trained by the following method:

[0014] receiving a first target bright field image sequence and a first target fluorescence image sequence of a target mixed cell sample collected by the data acquisition and processing module, the target mixed cell sample being provided with a first antibody dye and a second antibody dye;

[0015] determining a cell category corresponding to the excited target antibody dye according to a light-emitting position of each frame of the first target fluorescence image in the excitation area;

[0016] labeling the cell category in each frame of the corresponding first target bright field image to form a first target bright field image sequence with cell category labels;

[0017] training a convolutional neural network model based on the first target bright field image sequence with cell category labels to obtain a cell classification model.

[0018] In some embodiments, the training of the convolutional neural network model based on the first target bright field image sequence with cell category labels to obtain a cell classification model comprises:

[0019] inputting the first target bright field image sequence with cell category labels into the convolutional neural network model to extract image features of each frame of the first target bright field image;

[0020] integrating the image features into target image features, outputting a cell category probability of the target mixed cell sample based on the target image features, and screening a first target cell sample of a target category from the target mixed cell sample according to the cell category probability.

[0021] when the first loss value of the cell category probability and the real cell category probability is less than a first preset loss threshold, the trained convolutional neural network model is taken as the cell classification model.

[0022] In some embodiments, the cell three-dimensional reconstruction module comprises a second trainer configured to train a cell reconstruction model, and the cell reconstruction model is trained by:

[0023] receiving a second target bright field image sequence of a first target cell sample collected by the data acquisition and processing module, the first target cell sample being sample data of a first cell sample;

[0024] estimating a viewing angle parameter of each frame of the second target bright field image sequence, and extracting a contour point of the first target cell sample;

[0025] inputting the viewing angle parameter and the contour point into an initial cell reconstruction model, mapping corresponding volume density and color, and performing volume rendering on the corresponding second target bright field image based on the corresponding volume density and color to obtain a reconstructed second target cell sample;

[0026] when the second loss value of the reconstructed second target cell sample and the first target cell sample is less than a second preset loss threshold, the trained initial cell reconstruction model is taken as the cell reconstruction model.

[0027] In some embodiments, the cell deformation analysis module comprises a third trainer configured to train a cell deformation model, and the cell deformation model is trained by:

[0028] receiving a third target bright field image sequence of a second target cell sample collected by the data acquisition and processing module, the second target cell sample being sample data of a second cell sample;

[0029] extracting a deformation index of each frame of the third target bright field image sequence, and constructing a time series feature vector according to the deformation index;

[0030] inputting the time series feature vector into a Transformer model, capturing a deformation time sequence dependency relationship of the time series feature vector through a self-attention mechanism, mapping the deformation time sequence dependency relationship to simulated deformation data through a fully connected layer, and outputting the simulated deformation data;

[0031] when the third loss value of the simulated deformation data and the real deformation data is less than a third preset loss threshold, the trained Transformer model is taken as the cell deformation model.

[0032] In some embodiments, the data acquisition and processing module is further configured to integrate the cell category probability fed back by the cell classification module, the first cell sample of the target category, the second cell sample after reconstruction fed back by the cell three-dimensional reconstruction module, and the deformation data fed back by the cell deformation analysis module, to generate a report of the cell multi-modal analysis.

[0033] The data acquisition and processing module is further connected to a visualization module configured to receive and display the report of the cell multi-modal analysis.

[0034] In some embodiments, the deformation data comprises at least one of the following: cell elastic modulus, shear stress response, and deformation recovery rate; and the data acquisition and processing module is further configured to adjust the cell flow channel flow rate or the obstacle parameter in the microfluidic chip based on the deformation data after receiving the deformation data.

[0035] In a second aspect, the embodiments of the present specification also provide a cell multi-modal analysis method based on a microfluidic chip, comprising:

[0036] The data acquisition and processing module inputs the mixed cell sample, the first cell sample of the target category screened by the cell classification module, and the second cell sample after reconstruction output by the cell three-dimensional reconstruction module into the cell flow channel of the microfluidic chip, respectively, to acquire a first bright field image, a second bright field image, and a third bright field image, respectively.

[0037] The cell classification module receives the first bright field image, outputs a cell category probability through a cell classification model, screens the first cell sample of the target category, and feeds back to the data acquisition and processing module.

[0038] The cell three-dimensional reconstruction module receives the second bright field image, outputs the second cell sample after reconstruction through a cell reconstruction model, and feeds back to the data acquisition and processing module.

[0039] The cell deformation analysis module receives the third bright field image, outputs deformation data through a cell deformation model, and feeds back to the data acquisition and processing module.

[0040] The embodiment of the present specification provides a cell multi-modal analysis system and method based on a microfluidic chip, which comprises: a data acquisition and processing module, which is used for inputting a mixed cell sample, a first cell sample of a target category screened by a cell classification module, and a second cell sample output by a cell three-dimensional reconstruction module into a cell flow channel of the microfluidic chip respectively, and acquiring a first bright field image, a second bright field image, and a third bright field image respectively; the cell classification module is used for receiving the first bright field image, outputting a cell category probability through a cell classification model, screening the first cell sample of the target category, and feeding back to the data acquisition and processing module; the cell three-dimensional reconstruction module is used for receiving the second bright field image, outputting the second cell sample after reconstruction through a cell reconstruction model, and feeding back to the data acquisition and processing module; and the cell deformation analysis module is used for receiving the third bright field image, outputting deformation data through a cell deformation model, and feeding back to the data acquisition and processing module. The cell classification module, the cell three-dimensional reconstruction module, the cell deformation analysis module, the data acquisition and processing module, and the microfluidic chip are integrated in one system, the cell multi-modal analysis can be accurately, quickly, and low-costly performed by means of the microfluidic chip, the integrated detection of cell classification, three-dimensional reconstruction, and deformation analysis is realized, and the problem of high cost and long analysis period of cell multi-modal analysis caused by multiple platforms is avoided. In actual application, the cell classification module, the cell three-dimensional reconstruction module, and the cell deformation analysis module only need to input corresponding bright field images to obtain output results, and the cost of fluorescent reagents can be effectively saved. Through the steps of classification, reconstruction, and deformation analysis, the real-time three-dimensional morphological changes of cells in a flow state can be captured in real time, and dynamic deformation data can be obtained, so that the optimization design of the microfluidic chip and medical research can be accurately, quickly, and low-costly performed. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, brief descriptions will be given to the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort. In the drawings:

[0042] Figure 1 A module composition schematic diagram of a cell multi-modal analysis system based on a microfluidic chip provided by the embodiment of the present specification is provided;

[0043] Figure 2 A structure composition schematic diagram of a microfluidic chip provided by the embodiment of the present specification is provided;

[0044] Figure 3 A module composition schematic diagram of a cell multi-modal analysis system provided by the embodiment of the present specification during training is provided;

[0045] Figure 4 A flowchart of a cell multi-modal analysis method based on a microfluidic chip provided by an embodiment of the present specification is shown in the figure;

[0046] Figure 5 A schematic diagram of a first target bright field image sequence and a first target fluorescence image sequence collected by an embodiment of the present specification is shown in the figure;

[0047] Figure 6 A schematic diagram of cells flowing through a cell channel with obstacles provided by an embodiment of the present specification is shown in the figure;

[0048] Figure 7 A schematic diagram of the structural composition of an electronic device provided by an embodiment of the present specification is shown in the figure. DETAILED DESCRIPTION

[0049] In order for those skilled in the art to better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be described clearly and completely below in conjunction with the drawings in the embodiments of the present specification. Obviously, the described embodiments are only part of the embodiments of the present specification, not all the embodiments. Based on the embodiments in the present specification, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present specification.

[0050] As described above, in the field of cell analysis technology, the prior art has the following defects:

[0051] (1) Multi-platform fragmentation problem: cell classification (fluorescence microscope), cell three-dimensional reconstruction (confocal microscope), and cell deformation analysis (atomic force microscope) all require independent platforms or devices, resulting in large sample transfer loss and long analysis period;

[0052] (2) Dynamic deformation capture missing: traditional static detection cannot capture real-time three-dimensional morphological changes of cells in a flowing state, and cannot obtain deformation data of cells after collision with obstacles;

[0053] (3) Bright field image is not fully utilized: the existing system only uses bright field images for basic positioning, and does not fully apply them to cell classification, cell three-dimensional reconstruction, and cell deformation analysis;

[0054] In summary, the prior art cannot accurately, quickly, and at low cost, perform cell multi-modal analysis, and thus cannot accurately, quickly, and at low cost, perform microfluidic chip optimization design and medical research, etc.

[0055] Referring to Figure 1 The embodiments of the present specification provide a cell multi-modal analysis system based on a microfluidic chip, which can include:

[0056] a data acquisition and processing module, configured to input the mixed cell sample A, the first cell sample C of the target category screened by the cell classification module, and the reconstructed second cell sample E output by the cell three-dimensional reconstruction module into a cell flow channel of a microfluidic chip respectively, and to acquire the first bright field image B, the second bright field image D, and the third bright field image F respectively;

[0057] a cell classification module, configured to receive the first bright field image B, output a cell category probability through a cell classification model, screen the first cell sample C of the target category, and feed back to the data acquisition and processing module;

[0058] a cell three-dimensional reconstruction module, configured to receive the second bright field image D, output the reconstructed second cell sample E through a cell reconstruction model, and feed back to the data acquisition and processing module;

[0059] a cell deformation analysis module, configured to receive the third bright field image F, output deformation data G through a cell deformation model, and feed back to the data acquisition and processing module.

[0060] Specifically, the data acquisition and processing module can be connected with the microfluidic chip, and can also be connected with the cell classification module, the cell three-dimensional reconstruction module, and the cell deformation analysis module. The data acquisition and processing module can first collect a mixed cell sample A (the mixed cell sample A can be a cell solution containing multiple types of cells (such as tumor cells + immune cells)), then input the mixed cell sample A into the cell flow channel of the microfluidic chip (specifically, input into the cell flow inlet), and then collect a first bright field image B of the mixed cell sample A and send it to the cell classification module. After the cell classification module receives the first bright field image B of the mixed cell sample A, it can output the cell category probability of the mixed cell sample A (such as the probability of belonging to tumor cells, the probability of belonging to immune cells, etc.) through the cell classification model, and then according to the cell category probability of the mixed cell sample A, screen a first cell sample C of a target category from the mixed cell sample A (such as screening a cell sample with the maximum cell category probability or greater than a preset category probability threshold as the first cell sample C of the target category) and feed back or send to the data acquisition and processing module. Then, the data acquisition and processing module can input the first cell sample C of the target category into the cell flow channel of the microfluidic chip, and then collect a second bright field image D of the first cell sample C of the target category and send it to the cell three-dimensional reconstruction module. After the cell three-dimensional reconstruction module receives the second bright field image D of the first cell sample C of the target category, it can output the reconstructed second cell sample E through the cell reconstruction model and feed back or send to the data acquisition and processing module. Then, the data acquisition and processing module can input the reconstructed second cell sample E into the cell flow channel of the microfluidic chip, and then collect a third bright field image F of the reconstructed second cell sample E and send it to the cell deformation analysis module. After the cell deformation analysis module receives the third bright field image F of the reconstructed second cell sample E, it can output the deformation data G of the reconstructed second cell sample E through the cell deformation model and feed back or send to the data acquisition and processing module.

[0061] By integrating the cell classification module, the cell three-dimensional reconstruction module, the cell deformation analysis module, and the data acquisition and processing module with the microfluidic chip in one system, the present application can accurately, quickly, and at low cost perform cell multi-modal analysis, realize integrated detection of cell classification, three-dimensional reconstruction, and deformation analysis, and avoid the problem of high cost and long analysis period caused by multi-platform fragmentation. In actual application, the cell classification module, the cell three-dimensional reconstruction module, and the cell deformation analysis module only need to input the corresponding bright field image to obtain the output result, which can effectively save the cost of fluorescent reagents. By first classifying, then reconstructing, and then analyzing deformation, the real-time three-dimensional morphological changes of cells in a flow state can be captured in real time, and dynamic deformation data can be obtained, so that the optimization design of the microfluidic chip and medical research, etc. can be accurately, quickly, and at low cost.

[0062] In some embodiments, the data acquisition and processing module described above can also be configured to send the first bright-field image B to the cell classification module, send the second bright-field image D to the cell three-dimensional reconstruction module, and send the third bright-field image F to the cell deformation analysis module. The data acquisition and processing module can also be configured to, after receiving the reconstructed second cell sample E, set an obstacle in the cell flow channel, input the reconstructed second cell sample E into the cell flow channel after the obstacle is set, and acquire a bright-field image of the reconstructed second cell sample E after the obstacle, as the third bright-field image F.

[0063] Specifically, by setting an obstacle in the cell flow channel of the microfluidic chip, the reconstructed second cell sample E can be subjected to complex stresses such as collision, extrusion, shearing, and flow around the obstacle when flowing through the obstacle. The bright-field image after the collision (i.e., the bright-field image after the cell sample is subjected to the complex stresses, as the third bright-field image F) can be acquired and input into the cell deformation analysis module, so that the deformation data of the reconstructed second cell sample E after being subjected to stress (the deformation data can also be referred to as cell mechanics response data) can be accurately and dynamically analyzed.

[0064] In some embodiments, the shape of the obstacle can be set according to actual needs, and the shape of the obstacle can include at least one of a circle, a square, or a triangle. By using obstacles of different shapes, different types of forces can be applied to the reconstructed second cell sample E flowing through the cell flow channel, simulating the deformation process of the reconstructed second cell sample E under complex stress.

[0065] In some embodiments, referring to FIG. 1, Figure 2 As shown in FIG. 1, the cell flow channel can include an excitation area, which can include a left excitation area and a right excitation area, respectively receiving excitation light of a first wavelength and excitation light of a second wavelength. The excitation light of the first wavelength is used to excite the first antibody dye in the flowing cell sample and emit light, and the excitation light of the second wavelength is used to excite the second antibody dye in the flowing cell sample and emit light.

[0066] Specifically, the cell sample (such as the mixed cell sample A, etc.) can flow into the cell flow channel inlet of the microfluidic chip, first pass through the left excitation area, then pass through the right excitation area, and finally flow out from the cell flow channel outlet. Among them, the left excitation area and the right excitation area are respectively connected to the corresponding wavelength laser through the optical fiber, and receive the excitation light of the corresponding wavelength emitted from the corresponding wavelength laser, so that the excitation light of the corresponding wavelength excites the first antibody dye or the second antibody dye in the cell sample flowing through the excitation area and emits light. The microfluidic chip can also be connected to a fluorescence camera and a bright field camera, and after the antibody dye emits light, the fluorescence camera and the bright field camera are used to synchronously collect the fluorescence image and the bright field image of the cell sample, and then the data acquisition and processing module collects the fluorescence image and / or the bright field image of the corresponding cell sample. Among them, the first antibody dye can be a fluorescent dye for labeling a tumor marker (such as a fluorescent dye for labeling an EpCam marker), and the second antibody dye can be a fluorescent dye for labeling an immune cell marker (such as a fluorescent dye for labeling a CD45 + marker). The first antibody dye can be excited by the excitation light of the first wavelength of the left excitation area, and the second antibody dye can be excited by the excitation light of the second wavelength of the right excitation area, wherein the first wavelength is different from the second wavelength. Subsequently, the corresponding bright field image can be automatically labeled with a cell category by determining which antibody dye emits light, without the need for manual labeling, thereby improving the accuracy of cell category labeling, so that the cell classification model can be accurately trained. The training of the cell classification model will be described later, and this specification will not be repeated here.

[0067] It should be noted that the antibody dye in the cell sample can also be set according to actual needs, and the number of antibody dyes can also be not less than two, which is not limited in this specification.

[0068] In some embodiments, referring to Figure 3 The above cell classification module can include a first trainer, which is used to train a cell classification model, and the cell classification model is trained in the following manner:

[0069] The first target bright field image sequence B' and the first target fluorescence image sequence B'' of the target mixed cell sample A' collected by the data acquisition and processing module are received, and the target mixed cell sample A' has a first antibody dye and a second antibody dye;

[0070] According to the light-emitting position of each frame of the first target fluorescence image in the excitation area, the cell category corresponding to the excited target antibody dye is determined;

[0071] The cell category label is labeled in the corresponding each frame of the first target bright field image, and the first target bright field image sequence B' with the cell category label is formed;

[0072] training a convolutional neural network model based on the first target bright field image sequence B' with the cell category label to obtain a cell classification model.

[0073] In some embodiments, the training of the convolutional neural network model based on the first target bright field image sequence B' with the cell category label to obtain the cell classification model can include:

[0074] inputting the first target bright field image sequence B' with the cell category label into the convolutional neural network model to extract image features of each frame of the first target bright field image;

[0075] integrating the image features into target image features, outputting a cell category probability of the target mixed cell sample A' based on the target image features, and screening a first target cell sample C' of a target category from the target mixed cell sample A' according to the cell category probability;

[0076] when the first loss value of the cell category probability and the true cell category probability is less than a first preset loss threshold, the trained convolutional neural network model is used as the cell classification model.

[0077] Specifically, the data acquisition and processing module can first acquire the target mixed cell sample A', then input the target mixed cell sample A' into the cell flow channel (specifically, input into the cell flow inlet) of the microfluidic chip, and then acquire the first target bright field image sequence B' (which can include multiple frames of first target bright field images of the target mixed cell sample A' at different angles or different morphologies) and the first target fluorescence image sequence B'' (which can include multiple frames of first target fluorescence images of the target mixed cell sample A' at different angles or different morphologies) of the target mixed cell sample A', and send them to the first trainer of the cell classification module.

[0078] After the first trainer of the cell classification module receives the first target bright field image sequence B' and the first target fluorescence image sequence B" of the target mixed cell sample A', the cell class corresponding to the target antibody dye (such as the first antibody dye or the second antibody dye) excited in the target mixed cell sample A' can be determined based on the light-emitting position of each frame of the first target fluorescence image in the excitation area. For example, if the light-emitting position is on the left side of the excitation area, the target antibody dye excited in the target mixed cell sample A' is the first antibody dye, and the corresponding cell class is tumor cells. If the light-emitting position is on the right side of the excitation area, the target antibody dye excited in the target mixed cell sample A' is the second antibody dye, and the corresponding cell class is immune cells. Then, the spatial coordinate matching can be performed according to the light-emitting position of each frame of the first target fluorescence image in the excitation area. The corresponding position of each frame of the first target bright field image corresponding to the light-emitting position is matched, and the cell class label is automatically marked in each frame of the first target bright field image. For example, the cell class label of tumor cells is label = 1, and the cell class label of immune cells is label = 0. The first target bright field image sequence B' of the target mixed cell sample A' with the cell class label is formed. The convolutional neural network model (CNN) can be trained using the first target bright field image sequence B' of the target mixed cell sample A' with the cell class label. The fluorescence light-emitting position directly corresponds to the cell class, and the corresponding bright field image is automatically marked or labeled without manual labeling, which effectively improves the training accuracy and efficiency of the cell classification model.

[0079] Specifically, before training the convolutional neural network model, the first target bright field image sequence B' with cell category labels can be input into the convolutional neural network model, and the image features of each frame of the first target bright field image can be automatically extracted based on the CNN model, and the image features can be integrated or aggregated into target image features. The target image features can be processed using Softmax and output the cell category probability of the target mixed cell sample A'. The cell classification module can also filter the first target cell sample C' of the target category from the target mixed cell sample A' according to the cell category probability of the target mixed cell sample A' (such as filtering the cell sample with the maximum cell category probability or greater than the preset category probability threshold of the target mixed cell sample A' as the first target cell sample C' of the target category), and feeding back the first target cell sample C' of the target category to the data acquisition and processing module to lay the foundation for subsequent acquisition of input data of the second trainer. Finally, the first loss value of the cell category probability of the target mixed cell sample A' and the true cell category probability can be calculated, and when the first loss value is less than the first preset loss threshold, the training is stopped, and the trained convolutional neural network model is used as the cell classification model. The model parameters can also be adjusted when the first loss value is greater than the first preset loss threshold, and the first target bright field image sequence with cell category labels is input into the convolutional neural network model again, and the above process is repeated to train the convolutional neural network model until the first loss value is less than the first preset loss threshold.

[0080] Specifically, the model structure of the above CNN model can be:

[0081] Input→Conv2D×3→MaxPool→Dense×2→Softmax, and correspondingly, after inputting the first target bright field image sequence B' with cell category labels after image preprocessing into the convolutional neural network model, the image features of each frame of the first target bright field image can be extracted based on each convolution kernel of the convolution layer (Conv2D×3), and then the extracted image features of each frame of the first target bright field image can be down-sampled based on the pooling layer (MaxPool) to reduce the amount of calculation, and then the down-sampled image features can be integrated or aggregated into target image features based on the fully connected layer (Dense×2), and finally the target image features can be processed based on Softmax to output normalized cell category probability.

[0082] In actual application, only the first bright field image B of the mixed cell sample A (without a fluorescence image) needs to be acquired, input into the trained cell classification model, and the cell classification probability of the mixed cell sample A can be automatically, accurately and quickly output. The cell classification module can quickly screen the first cell sample C of the target category and feed back to the data acquisition and processing module, so that the data acquisition and processing module starts a new round of acquisition.

[0083] In some embodiments, referring to Figure 3 As shown in the above cell three-dimensional reconstruction module can include a second trainer, the second trainer is used to train the cell reconstruction model, the cell reconstruction model is trained by the following way:

[0084] Receiving the second target bright field image sequence of the first target cell sample collected by the data acquisition and processing module, the first target cell sample is the sample data of the first cell sample;

[0085] Estimating the viewing angle parameters of each frame of the second target bright field image sequence, and extracting the contour points of the first target cell sample;

[0086] Inputting the viewing angle parameters and the contour points into the initial cell reconstruction model, mapping into corresponding volume density and color, performing volume rendering on the corresponding second target bright field image based on the corresponding volume density and color, and obtaining the reconstructed second target cell sample;

[0087] When the second loss value of the reconstructed second target cell sample and the first target cell sample is less than the second preset loss threshold, the trained initial cell reconstruction model is used as the cell reconstruction model.

[0088] Specifically, after receiving the first target cell sample C' of the target category (i.e. the full name of the above first target cell sample), the data acquisition and processing module can input the first target cell sample C' of the target category into the cell flow channel of the microfluidic chip, and then collect the second target bright field image sequence D' of the first target cell sample C' of the target category (which can include multiple frames of second target bright field images of the first target cell sample C' of the target category with different angles or different morphologies, such as 20 frames of bright field images taken when the first target cell sample C' rotates in flow) and send it to the second trainer of the cell three-dimensional reconstruction module.

[0089] After the second trainer of the cell three-dimensional reconstruction module receives the second target bright field image sequence D' of the first target cell sample C' of the target category, the second target bright field image sequence D' can be preprocessed, such as background removal, cell segmentation, image frame alignment, and the like, to improve the training accuracy of the model. Then, the viewing angle parameters (such as i.e., the angle and the orientation) of each frame of the second target bright field image sequence D' can be estimated based on the flow direction of the first target cell sample C' in the cell flow channel, and the contour points (i.e., sampling points, such as X = {x, y}, i.e., the horizontal and vertical spatial coordinates) of the first target cell sample C' can be extracted. Then, the viewing angle parameters and the contour points of each frame of the second target bright field image sequence D' can be input into the initial cell reconstruction model (the initial cell reconstruction model can include a NeRF model), and the corresponding viewing angle parameters and contour points can be mapped to the corresponding volume density and color by the model. Then, the corresponding second target bright field image can be volume-rendered based on the corresponding volume density and color, and the volume-rendered second target bright field image sequence D' can be obtained. Then, the first target cell sample C' can be reconstructed based on the volume-rendered second target bright field image sequence D', and the reconstructed second target cell sample E' can be obtained and fed back to the data acquisition and processing module to lay a foundation for subsequent input data of the third trainer. Finally, the second loss value of the reconstructed second target cell sample E' and the first target cell sample C' can be calculated, and when the second loss value is less than a second preset loss threshold, the training is stopped, and the trained initial cell reconstruction model is taken as the cell reconstruction model. When the second loss value is greater than the second preset loss threshold, the model parameters can be adjusted, and the viewing angle parameters and the contour points of each frame of the second target bright field image sequence D' can be input into the initial cell reconstruction model again, and the above process can be repeated to train the initial cell reconstruction model until the second loss value is less than the second preset loss threshold.

[0090] The mapping can be performed according to the following formula:

[0091] wherein X is the contour point, d is the viewing angle parameter, is the volume density, and c is the color.

[0092] In this way, a high-precision cell reconstruction model can be trained. In actual application, only the second bright field image D of the first cell sample C of the target category needs to be input into the trained cell reconstruction model, and the three-dimensional reconstruction result (i.e., the reconstructed second cell sample E) of the first cell sample C can be automatically, accurately, and quickly output, and then fed back to the data acquisition and processing module. The influence of the microfluidic chip structure on the three-dimensional morphology of the cell can be analyzed, and the data acquisition and processing module can start a new round of acquisition.

[0093] In some embodiments, reference can be made to Figure 3As shown, the cell deformation analysis module includes a third trainer configured to train a cell deformation model, wherein the cell deformation model is trained by:

[0094] receiving a third target bright-field image sequence of a second target cell sample collected by the data acquisition and processing module, the second target cell sample being sample data of the second cell sample;

[0095] extracting deformation indicators of each frame of the third target bright-field image sequence, and constructing a time series feature vector according to the deformation indicators;

[0096] inputting the time series feature vector into the Transformer model, capturing deformation time sequence dependency of the time series feature vector through self-attention mechanism, mapping the deformation time sequence dependency to simulated deformation data through a fully connected layer, and outputting the simulated deformation data;

[0097] when the third loss value of the simulated deformation data and the real deformation data is less than the third preset loss threshold, the trained Transformer model is used as the cell deformation model.

[0098] Specifically, after receiving the reconstructed second target cell sample E' (i.e., the full name of the second target cell sample), the data acquisition and processing module can first set an obstacle in the cell flow channel, then input the reconstructed second target cell sample E' into the cell flow channel after the obstacle is set, collect a target bright-field image sequence after the reconstructed second target cell sample E' collides with the obstacle as a third target bright-field image sequence F' (which can include multiple frames of third target bright-field images passing through the obstacle), and send the third target bright-field image sequence F' of the second target cell sample to the third trainer of the cell deformation analysis module.

[0099] After receiving the reconstructed third target bright-field image sequence F' of the second target cell sample E', the third trainer of the cell deformation analysis module can first perform image preprocessing on each frame of the third target bright-field image, then extract deformation indicators (deformation indicators are features extracted from images that describe their geometric or morphological changes, such as aspect ratio, convex hull area ratio, boundary curvature variation, circularity, etc.) of each frame of the third target bright-field image after image preprocessing, then construct a time sequence feature vector of each frame of the third target bright-field image based on the deformation indicators of each frame of the third target bright-field image (i.e. arrange the deformation indicators of each frame of the third target bright-field image in chronological order to form a feature matrix), then input the time sequence feature vector of each frame of the third target bright-field image into the Transformer model, capture the deformation time sequence dependency of the time sequence feature vector of each frame of the third target bright-field image based on the self-attention mechanism (such as the deformation of the 5th frame of the third target bright-field image being influenced by the previous 4 frames of images, etc.), and finally map the deformation time sequence dependency of each frame of the third target bright-field image to simulated deformation data (mechanical response data after stress, such as cell elastic modulus, shear stress response, and deformation recovery rate, etc.) of the reconstructed second target cell sample E' through a fully connected layer, and feedback to the data acquisition and processing module.

[0100] In this way, a high-precision cell deformation analysis model can be trained. In actual application, only the second bright-field image F of the reconstructed second cell sample E is needed to be input into the trained cell deformation analysis model, and the deformation data of the second cell sample E can be automatically, accurately and quickly output, and then fed back to the data acquisition and processing module for optimization design of the microfluidic chip, etc.

[0101] By first classifying, then reconstructing, and then analyzing deformation, real-time three-dimensional morphological changes of cells in a flow state can be captured in real time, and dynamic mechanical response characteristics can be obtained, so that optimization design of microfluidic chips and medical research, etc. can be accurately, quickly and cost-effectively performed. By using the corresponding bright-field image sequence, the stress conditions of the corresponding cell sample in the microenvironment can be predicted, and non-contact and high-throughput cell mechanics measurement can be achieved.

[0102] It should be noted that the first preset loss threshold, the second preset loss threshold, and the third preset loss threshold in the above embodiments can be set according to actual needs, and the present specification does not make specific limitations thereon. In addition to stopping training when the first loss value is less than the first preset loss threshold, the second loss value is less than the second preset loss threshold, and the third loss value is less than the third preset loss threshold, the training can also be stopped when the number of training times of the corresponding model reaches the corresponding preset iteration number, and the present specification does not make specific limitations thereon.

[0103] In some embodiments, the data acquisition and processing module can also be used to integrate the cell category probability fed back by the cell classification module and the target category first cell sample C, the reconstructed second cell sample D fed back by the cell three-dimensional reconstruction module, and the deformation data F fed back by the cell deformation analysis module, to generate a report of cell multi-modal analysis.

[0104] The data acquisition and processing module can also be connected to a visualization module, which can be used to receive and display the report of cell multi-modal analysis.

[0105] By generating and displaying the report of cell multi-modal analysis, the dimensions of cell classification, cell reconstruction, and cell deformation can be associated, which can reveal rules that cannot be discovered by traditional single modal, ensure the interpretability of the output results of the modules, and help researchers to carry out related research work.

[0106] In some embodiments, the deformation data can include at least one of the following: cell elastic modulus, shear stress response, and deformation recovery rate; and the data acquisition and processing module, after receiving the deformation data, can also be used to adjust the cell flow channel flow rate or obstacle parameter in the microfluidic chip based on the deformation data.

[0107] Specifically, the obstacle parameter can include obstacle shape, density, and the like. The obstacle parameter can be adjusted according to the deformation data, for example, when the cell elastic modulus is less than a set threshold, the flow rate can be reduced to reduce deformation; when the deformation recovery rate is slow, the obstacle density can be reduced to reduce the frequency of continuous collision; and when the shear stress is greater than the cell tolerance limit, the sharp obstacle can be replaced by a circular one to homogenize the stress, and the like.

[0108] By dynamically adjusting the cell flow channel flow rate or obstacle parameter of the microfluidic chip based on the deformation data, a leap from “static experiment” to “intelligent response” is achieved. The optimized design of the cell flow channel flow rate or obstacle parameter can ultimately improve the efficiency and accuracy of cell sorting, diagnosis, or drug screening, and the like.

[0109] In some embodiments, the reconstructed second cell sample output by the three-dimensional reconstruction module can include the three-dimensional structure (such as volume, shape) of the cell sample, and the subsequent deformation analysis module can also receive the three-dimensional structure of the cell sample to quantify the deformation difference of the cell sample before and after being stressed, and the like. The cell classification module, the cell three-dimensional reconstruction module, and the cell deformation analysis module share the same microfluidic chip platform, and the function switching can be realized by replacing the chip or adjusting the excitation zone configuration.

[0110] Referring to Figure 4 The embodiments of the present specification provide a cell multi-modal analysis method based on a microfluidic chip, based on the above-mentioned cell multi-modal analysis system, the method can include:

[0111] S401: The data acquisition and processing module inputs the mixed cell sample, the first cell sample of the target category screened by the cell classification module, and the reconstructed second cell sample output by the cell three-dimensional reconstruction module into the cell flow channel of the microfluidic chip respectively, and acquires the first bright field image, the second bright field image, and the third bright field image respectively;

[0112] S402: The cell classification module receives the first bright field image, outputs the cell category probability through the cell classification model, screens the first cell sample of the target category, and feeds back to the data acquisition and processing module;

[0113] S403: The cell three-dimensional reconstruction module receives the second bright field image, outputs the reconstructed second cell sample through the cell reconstruction model, and feeds back to the data acquisition and processing module;

[0114] S404: The cell deformation analysis module receives the third bright field image, outputs the deformation data through the cell deformation model, and feeds back to the data acquisition and processing module.

[0115] The embodiment description of S401-S404 can refer to the description of the preceding cell multi-modal analysis system embodiment, and the description will not be repeated here.

[0116] The present application can realize high-throughput and automatic multi-dimensional (cell classification, three-dimensional reconstruction, and deformation analysis) characterization of cells by combining a microfluidic chip with multi-modal cell machine learning models (cell classification model, cell reconstruction model, and cell deformation analysis model).

[0117] Each embodiment in the present specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the difference from other embodiments. For specific details, refer to the description of the related processing embodiments described above, which will not be repeated here.

[0118] The above describes the present application, however, it is worth noting that the specific embodiments are only for better illustrating the present application, and the description of the specific embodiments in the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0119] Referring to Figure 5 as shown, Figure 5 The schematic diagram of the first target bright field image sequence and the first target fluorescence image sequence collected is shown.Figure 5 The black rectangle represents the acquired first target fluorescence image, and the gray rectangle represents the acquired first target bright-field image. The first target fluorescence image and the first target bright-field image are acquired simultaneously and have a spatial correspondence. Based on the emission position of the acquired first target fluorescence image in the excitation region, the cell category corresponding to the excited antibody dye can be determined. Cell category labels are then automatically added to the first target bright-field image at the corresponding position. The labeled sequence of first target bright-field images with cell category labels is then used to train a convolutional neural network model (i.e., the initial cell classification model).

[0120] See Figure 6 As shown, Figure 6 The diagram illustrates cell flow through a cell channel with obstructions. By placing obstructions in the cell channel of a microfluidic chip, the reconstructed second cell sample experiences combined stresses such as collision, compression, shearing, and flow around the obstruction. A bright-field image is then acquired after the collision (i.e., a bright-field image after the aforementioned combined stresses is acquired, serving as a third bright-field image), and input into the cell deformation analysis module. This allows for accurate and dynamic analysis of the deformation data generated by the reconstructed second cell sample under stress, which can be used for microfluidic chip optimization design and medical research.

[0121] This specification also provides an electronic device based on the aforementioned microfluidic chip-based cell multimodal analysis method, including a processor and a memory for storing processor-executable programs / instructions. Specifically, the processor can execute the following steps according to the program / instructions: A data acquisition and processing module inputs a mixed cell sample, a first cell sample of the target category selected by a cell classification module, and a reconstructed second cell sample output by a cell 3D reconstruction module into the cell channel of the microfluidic chip, respectively, acquiring a first bright-field image, a second bright-field image, and a third bright-field image; a cell classification module receives the first bright-field image, outputs cell category probabilities through a cell classification model, selects the first cell sample of the target category, and feeds it back to the data acquisition and processing module; a cell 3D reconstruction module receives the second bright-field image, outputs the reconstructed second cell sample through a cell reconstruction model, and feeds it back to the data acquisition and processing module; a cell deformation analysis module receives the third bright-field image, outputs deformation data through a cell deformation model, and feeds it back to the data acquisition and processing module.

[0122] To execute the above instructions more accurately, please refer to... Figure 7 As shown in the embodiments of this specification, another specific electronic device is also provided, wherein the electronic device includes a network communication port 701, a processor 702, and a memory 703. The above structures are connected by internal cables so that the various structures can perform specific data interaction.

[0123] The processor 702 can be specifically configured for the data acquisition and processing module to input the mixed cell sample, the first cell sample of the target category screened by the cell classification module, and the reconstructed second cell sample output by the cell three-dimensional reconstruction module into the cell flow channel of the microfluidic chip respectively, and to acquire the first bright field image, the second bright field image, and the third bright field image respectively; the cell classification module receives the first bright field image, outputs the cell category probability through the cell classification model, screens the first cell sample of the target category, and feeds back to the data acquisition and processing module; the cell three-dimensional reconstruction module receives the second bright field image, outputs the reconstructed second cell sample through the cell reconstruction model, and feeds back to the data acquisition and processing module; and the cell deformation analysis module receives the third bright field image, outputs the deformation data through the cell deformation model, and feeds back to the data acquisition and processing module.

[0124] The memory 703 can be specifically configured to store corresponding instruction programs.

[0125] In this embodiment, the network communication port 701 can be a virtual port that is bound with different communication protocols, so as to send or receive different data. For example, the network communication port can be a port responsible for web data communication, can also be a port responsible for FTP data communication, and can also be a port responsible for mail data communication. In addition, the network communication port can also be an entity communication interface or a communication chip. For example, it can be a wireless mobile network communication chip such as GSM, CDMA, etc.; it can also be a Wifi chip; and it can also be a Bluetooth chip.

[0126] In this embodiment, the processor 702 can be implemented in any appropriate manner. For example, the processor can take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, and the like. The present specification is not limited thereto.

[0127] In this embodiment, the memory 703 can include multiple levels. In a digital system, as long as it can save binary data, it can be a memory; in an integrated circuit, a circuit without a physical form and with a storage function is also called a memory, such as RAM, FIFO, etc.; and in a system, a storage device with a physical form is also called a memory, such as a memory stick, a TF card, etc.

[0128] The embodiment of the present specification further provides a computer storage medium based on the above-mentioned cell multi-modal analysis method based on a microfluidic chip, the computer storage medium stores computer programs / instructions, and when the computer programs / instructions are executed, the following functions and effects are realized: the data acquisition and processing module inputs the mixed cell sample, the first cell sample of the target category screened by the cell classification module, and the reconstructed second cell sample output by the cell three-dimensional reconstruction module into the cell flow channel of the microfluidic chip respectively, and acquires the first bright field image, the second bright field image, and the third bright field image respectively; the cell classification module receives the first bright field image, outputs the cell category probability through the cell classification model, screens the first cell sample of the target category, and feeds back to the data acquisition and processing module; the cell three-dimensional reconstruction module receives the second bright field image, outputs the reconstructed second cell sample through the cell reconstruction model, and feeds back to the data acquisition and processing module; and the cell deformation analysis module receives the third bright field image, outputs the deformation data through the cell deformation model, and feeds back to the data acquisition and processing module.

[0129] In the embodiment, the storage medium includes but is not limited to a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk drive (HDD), or a memory card. The storage medium can be used to store computer program instructions. The network communication unit can be an interface set according to a standard specified by a communication protocol, and is used for network connection communication.

[0130] In the embodiment, the functions and effects realized by the program instructions stored in the computer storage medium can be explained in comparison with other embodiments, and will not be described herein.

[0131] Although the description has been provided with method operation steps as described in the embodiments or flowcharts, more or less operation steps can be included based on conventional or non-inventive means. The order of steps listed in the embodiments is only one of the many ways of performing the steps, and does not represent the only way of performing the steps. In actual device or client product execution, the method order shown in the embodiments or the drawings can be executed in sequence or in parallel (for example, in a parallel processor or multi-threaded environment, or even in a distributed data processing environment). The term "comprise", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, product or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, product or equipment. Without more limitations, it does not exclude the presence of other same or equivalent elements in the process, method, product or equipment including the elements. The terms "first", "second" and the like are used to indicate names, and do not represent any particular order.

[0132] Those skilled in the art also know that, in addition to implementing the controller in the form of pure computer readable program code, the same function can be achieved by logically programming the method steps into the form of logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers. Therefore, such a controller can be considered as a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0133] The description can be described in the general context of computer-executable instructions, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, and the like that perform particular tasks or implement particular abstract data types. The description can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are connected through a communication network. In a distributed computing environment, program modules can be located in both local and remote computer storage media including storage devices.

[0134] Those skilled in the art can clearly understand the present specification can be implemented by means of software and necessary general hardware platforms through the above description of the embodiments. Based on such understanding, the technical solutions of the present specification can essentially be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments of the present specification.

[0135] The various embodiments in the present specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. The present specification can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, small computers, large computers, distributed computing environments including any of the above systems or devices, etc.

[0136] Although the present specification is described through the embodiments, those skilled in the art know that there are many variations of the present specification without departing from the spirit of the present specification, and it is intended that the appended claims include these variations without departing from the spirit of the present specification.

Claims

1. A microfluidic chip-based cell multi-modal analysis system, characterized in that, The method comprises the following steps: a data acquisition and processing module is configured to input a mixed cell sample, a first cell sample of a target category screened by a cell classification module, and a reconstructed second cell sample output by a cell three-dimensional reconstruction module into a cell flow channel of a microfluidic chip respectively, and to acquire a first bright field image, a second bright field image, and a third bright field image respectively; the cell classification module is configured to receive the first bright field image, output a cell category probability through a cell classification model, screen the first cell sample of the target category, and feed back to the data acquisition and processing module; the cell three-dimensional reconstruction module is configured to receive the second bright field image, output the reconstructed second cell sample through a cell reconstruction model, and feed back to the data acquisition and processing module; the cell deformation analysis module is configured to receive the third bright field image, output deformation data through a cell deformation model, and feed back to the data acquisition and processing module; the data acquisition and processing module is further configured to send the first bright field image to the cell classification module, send the second bright field image to the cell three-dimensional reconstruction module, and send the third bright field image to the cell deformation analysis module; and the data acquisition and processing module is further configured to, after receiving the reconstructed second cell sample, set an obstacle in the cell flow channel, input the reconstructed second cell sample into the cell flow channel after the obstacle is set, acquire a bright field image of the reconstructed second cell sample after colliding with the obstacle as the third bright field image; the cell flow channel comprises an excitation area, the excitation area comprises a left excitation area and a right excitation area, and the left excitation area and the right excitation area receive excitation light of a first wavelength and excitation light of a second wavelength respectively, the excitation light of the first wavelength is used to excite a first antibody dye in a cell sample flowing through and emit light, and the excitation light of the second wavelength is used to excite a second antibody dye in the cell sample flowing through and emit light.

2. The cellular multi-modality analysis system of claim 1, wherein, The cell classification module comprises a first trainer, and the first trainer is configured to train a cell classification model, and the cell classification model is trained in the following manner: receiving a first target bright field image sequence and a first target fluorescence image sequence of a target mixed cell sample collected by the data acquisition and processing module, the target mixed cell sample being provided with a first antibody dye and a second antibody dye; determining a cell category corresponding to the excited target antibody dye according to a light-emitting position of each frame of the first target fluorescence image in the excitation area; labeling the cell category in each frame of the first target bright field image to form a first target bright field image sequence with a cell category label; training a convolutional neural network model based on the first target bright field image sequence with the cell category label to obtain the cell classification model.

3. The cellular multi-modal analysis system of claim 2, wherein, The training of the convolutional neural network model based on the first target bright field image sequence with the cell category label to obtain the cell classification model comprises: inputting the first target bright field image sequence with the cell category label into the convolutional neural network model to extract image features of each frame of the first target bright field image; integrate the image features into target image features, output a cell category probability of the target mixed cell sample based on the target image features, and screen a first target cell sample of a target category from the target mixed cell sample according to the cell category probability; when a first loss value of the cell category probability and a real cell category probability is less than a first preset loss threshold, the trained convolutional neural network model is used as the cell classification model.

4. The cellular multi-modality analysis system of claim 1, wherein, The cell three-dimensional reconstruction module comprises a second trainer configured to train a cell reconstruction model, and the cell reconstruction model is trained in the following manner: receive a second target bright field image sequence of the first target cell sample collected by the data acquisition and processing module, the first target cell sample being sample data of the first cell sample; estimate a viewing angle parameter of each frame of the second target bright field image sequence, and extract contour points of the first target cell sample; input the viewing angle parameter and the contour points into the initial cell reconstruction model, map the viewing angle parameter and the contour points into corresponding volume density and color, perform volume rendering on the corresponding second target bright field image based on the corresponding volume density and color, and obtain a reconstructed second target cell sample; when a second loss value of the reconstructed second target cell sample and the first target cell sample is less than a second preset loss threshold, the trained initial cell reconstruction model is used as the cell reconstruction model.

5. The cellular multi-modality analysis system of claim 1, wherein, The cell deformation analysis module comprises a third trainer configured to train a cell deformation model, and the cell deformation model is trained in the following manner: receive a third target bright field image sequence of the second target cell sample collected by the data acquisition and processing module, the second target cell sample being sample data of the second cell sample; extract a deformation index of each frame of the third target bright field image sequence, and construct a time series feature vector according to the deformation index; input the time series feature vector into a Transformer model, capture a deformation time sequence dependency of the time series feature vector through a self-attention mechanism, map the deformation time sequence dependency into simulated deformation data through a fully connected layer, and output the simulated deformation data; when a third loss value of the simulated deformation data and real deformation data is less than a third preset loss threshold, the trained Transformer model is used as the cell deformation model.

6. The cellular multi-modality analysis system of claim 1, wherein, The data acquisition and processing module is further configured to integrate the cell category probability fed back by the cell classification module and the first cell sample of the target category, the reconstructed second cell sample fed back by the cell three-dimensional reconstruction module, and deformation data fed back by the cell deformation analysis module, and generate a report of cell multi-modal analysis. The data acquisition and processing module is further connected to a visualization module, and the visualization module is configured to receive and display the report of cell multi-modal analysis.

7. The cellular multi-modal analysis system of claim 6, wherein, The deformation data comprises at least one of the following: cell elastic modulus, shear stress response, and deformation recovery rate; and the data acquisition and processing module is further configured to adjust a cell flow channel flow rate or an obstacle parameter in the microfluidic chip based on the deformation data after receiving the deformation data.

8. A method for cell multi-modal analysis based on a microfluidic chip, characterized in that, The method is applied to the cell multi-modal analysis system according to any one of claims 1 to 7, and the method comprises: The data acquisition and processing module inputs the mixed cell sample, the first cell sample of the target category screened by the cell classification module, and the reconstructed second cell sample output by the cell three-dimensional reconstruction module into the cell flow channel of the microfluidic chip respectively, and acquires the first bright field image, the second bright field image, and the third bright field image respectively; The cell classification module receives the first bright field image, outputs the cell category probability through the cell classification model, screens the first cell sample of the target category, and feeds back to the data acquisition and processing module; The cell three-dimensional reconstruction module receives the second bright field image, outputs the reconstructed second cell sample through the cell reconstruction model, and feeds back to the data acquisition and processing module; The cell deformation analysis module receives the third bright field image, outputs the deformation data through the cell deformation model, and feeds back to the data acquisition and processing module; The data acquisition and processing module is further configured to send the first bright field image to the cell classification module, send the second bright field image to the cell three-dimensional reconstruction module, and send the third bright field image to the cell deformation analysis module; and the data acquisition and processing module is further configured to, after receiving the reconstructed second cell sample, set an obstacle in the cell flow channel, input the reconstructed second cell sample into the cell flow channel after the obstacle is set, acquire the bright field image of the reconstructed second cell sample after colliding with the obstacle as the third bright field image. The cell flow channel comprises an excitation area, the excitation area comprises a left excitation area and a right excitation area, and the left excitation area and the right excitation area receive excitation light of a first wavelength and excitation light of a second wavelength respectively, the excitation light of the first wavelength is used to excite the first antibody dye in the cell sample flowing through and emit light, and the excitation light of the second wavelength is used to excite the second antibody dye in the cell sample flowing through and emit light.

Images