CN116393188B

CN116393188B - Microfluidic chip and method suitable for capturing circulating tumor cells

Info

Publication number: CN116393188B
Application number: CN202310671592.9A
Authority: CN
Inventors: 张开山; 郑红平; 郭志敏; 刘艳省; 赵丹; 吴乐中; 李超; 田华; 饶浪晴
Original assignee: HANGZHOU WATSON BIOTECH Inc
Current assignee: HANGZHOU WATSON BIOTECH Inc
Priority date: 2023-06-08
Filing date: 2023-06-08
Publication date: 2024-02-27
Anticipated expiration: 2043-06-08
Also published as: CN116393188A

Abstract

A microfluidic chip suitable for capturing circulating tumor cells and a method thereof acquire a cell motion trail graph acquired by a microscope; then, combining the technology based on deep learning and artificial intelligence, the electric field gradient is adaptively adjusted based on the motion track global map of different types of cells so as to improve the separation accuracy and efficiency. Therefore, the gradient strength of the electric field can be automatically adjusted according to the motion track characteristics of the cells, so that the capturing efficiency and accuracy of the cells are improved, and the separation effect and quality of the cells are further improved.

Description

Microfluidic chip and method suitable for capturing circulating tumor cells

Technical Field

The application relates to the technical field of intelligent chips, and in particular relates to a microfluidic chip and a method thereof suitable for capturing circulating tumor cells.

Background

Circulating tumor cells (Circulating Tumor Cell, CTCs) refer to tumor cells that shed from a primary tumor and enter the circulatory system. The detection and separation of CTC has important roles in early diagnosis, prognosis judgment, treatment effect evaluation and the like of tumors.

Currently, many researchers have begun exploring microfluidic chips suitable for the capture of circulating tumor cells. These microfluidic chips mainly utilize physical properties in micro channels, such as laminar flow separation, biomolecular affinity, surface tension, etc., to separate CTCs in a blood sample. For example, a multifunctional microfluidic chip that integrates multiple separation methods, such as chemical affinity separation, electroretention separation, etc., is capable of capturing different types of CTCs simultaneously. Among them, electroretention separation is a technique that uses an electric field gradient to orient cells. By arranging a series of electrodes with different frequencies and different voltages in the microfluidic chip, complex electric field gradients can be formed in the micro-channel, so that the directional capturing and separating of cells are realized. However, in the process of performing directional cell capturing by actually utilizing an electric field gradient, the problems of insufficient manual control precision, insufficient stability, insufficient flexibility and the like exist, the capturing efficiency and accuracy of cells are affected, and the separation effect and quality of the cells are reduced.

Thus, a solution is desired.

Disclosure of Invention

The present application has been made in order to solve the above technical problems. The embodiment of the application provides a microfluidic chip and a method thereof suitable for capturing circulating tumor cells, wherein the microfluidic chip acquires a cell motion trail graph acquired by a microscope; then, combining the technology based on deep learning and artificial intelligence, the electric field gradient is adaptively adjusted based on the motion track global map of different types of cells so as to improve the separation accuracy and efficiency. Therefore, the gradient strength of the electric field can be automatically adjusted according to the motion track characteristics of the cells, so that the capturing efficiency and accuracy of the cells are improved, and the separation effect and quality of the cells are further improved.

In a first aspect, there is provided a microfluidic chip adapted for capturing circulating tumor cells, comprising:

the image acquisition module is used for acquiring a cell motion trail graph acquired by a microscope;

the space strengthening module is used for obtaining a space strengthening cell motion trail feature matrix through a convolutional neural network model using a space attention mechanism;

the characteristic matrix dividing module is used for dividing the characteristic matrix of the motion trail of the space-enhanced cell to obtain a plurality of motion trail submatrices;

The context coding module is used for respectively expanding the plurality of motion track sub-matrixes into a plurality of motion track sub-feature vectors and then obtaining a plurality of context motion track sub-vectors through a context coder based on a converter;

the difference measurement module is used for calculating Euclidean distances between every two contextual motion track sub-vectors of the contextual motion track sub-vectors to obtain a classification feature vector consisting of the Euclidean distances;

the manifold curved surface optimization module is used for optimizing the manifold curved surface of the Gaussian probability density of the classification feature vector to obtain an optimized classification feature vector; and

and the control result generation module is used for enabling the optimized classification feature vector to pass through a classifier to obtain a classification result, and the classification result is used for indicating whether the electric field gradient strength is enhanced or not.

In the above microfluidic chip for capturing circulating tumor cells, the space enhancement module is configured to: each layer of the convolutional neural network model using the spatial attention mechanism performs the following steps on input data in the forward transfer process of the layer: convolving the input data to generate a convolved feature map; pooling the convolution feature map to generate a pooled feature map; non-linearly activating the pooled feature map to generate an activated feature map; calculating the mean value of each position of the activation feature map along the channel dimension to generate a spatial feature matrix; calculating a Softmax-like function value of each position in the space feature matrix to obtain a space score matrix; calculating the position-wise dot multiplication of the spatial feature matrix and the spatial score matrix to obtain a feature matrix; the characteristic matrix of the last layer output of the convolutional neural network model using the spatial attention mechanism is the characteristic matrix of the spatial enhanced cell motion trail.

In the above microfluidic chip for capturing circulating tumor cells, the context encoding module comprises: the query vector construction unit is used for carrying out one-dimensional arrangement on the plurality of motion track sub-feature vectors to obtain a motion track global feature vector; the self-attention unit is used for calculating the product between the global motion track feature vector and the transpose vector of each motion track sub-feature vector in the plurality of motion track sub-feature vectors to obtain a plurality of self-attention association matrixes; the normalization unit is used for respectively performing normalization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices; the attention calculating unit is used for obtaining a plurality of probability values through a Softmax classification function by each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and a weighting unit, configured to weight each motion track sub-feature vector in the plurality of motion track sub-feature vectors with each probability value in the plurality of probability values as a weight, so as to obtain the plurality of context motion track sub-vectors.

In the above microfluidic chip for capturing circulating tumor cells, the difference measurement module includes: a euclidean distance calculating unit for calculating a plurality of euclidean distances between every two contextual movement track sub-vectors of the plurality of contextual movement track sub-vectors in the following distance formula; wherein, the distance formula is:

wherein,representing each of the plurality of contextual motion trail subvectorsTwo context motion trail subvectors, < ->Representing the contextual motion trail subvector +.>Characteristic values of the respective positions in>Representing cosine distances between every two contextual motion trail subvectors of the plurality of contextual motion trail subvectors; and an arrangement unit configured to arrange the plurality of euclidean distances to obtain the classification feature vector.

In the above microfluidic chip for capturing circulating tumor cells, the manifold curved surface optimizing module is configured to: optimizing the manifold curved surface of the Gaussian probability density of the classification feature vector by using the following optimization formula to obtain an optimized classification feature vector; wherein, the optimization formula is:

wherein,and->Is the mean and standard deviation of the feature value sets of each position in the classification feature vector, +. >Is the +.o of the classification feature vector>Characteristic value of individual position, and->Is the +.f of the optimized classification feature vector>Characteristic values of the individual positions.

In the above microfluidic chip for capturing circulating tumor cells, the control result generating module includes: the full-connection coding unit is used for carrying out full-connection coding on the optimized classification feature vector by using a plurality of full-connection layers of the classifier so as to obtain a coding classification feature vector; and the classification unit is used for passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.

In a second aspect, there is provided a method suitable for capturing circulating tumor cells comprising:

acquiring a cell motion trail graph acquired by a microscope;

the cell motion trail graph is subjected to a convolutional neural network model by using a spatial attention mechanism to obtain a spatial reinforced cell motion trail feature matrix;

performing feature matrix division on the space-enhanced cell motion trail feature matrix to obtain a plurality of motion trail submatrices;

the plurality of motion track sub-matrixes are respectively unfolded into a plurality of motion track sub-feature vectors, and then a context encoder based on a converter is used for obtaining a plurality of context motion track sub-vectors;

Calculating Euclidean distances between every two contextual motion trail sub-vectors of the contextual motion trail sub-vectors to obtain a classification feature vector consisting of a plurality of Euclidean distances;

carrying out manifold curved surface optimization of Gaussian probability density on the classification feature vector to obtain an optimized classification feature vector; and

and the optimized classification feature vector passes through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the electric field gradient strength is enhanced or not.

In the above method for capturing circulating tumor cells, the step of obtaining a space-enhanced cell motion trajectory feature matrix from the cell motion trajectory graph by using a convolutional neural network model of a space attention mechanism comprises: each layer of the convolutional neural network model using the spatial attention mechanism performs the following steps on input data in the forward transfer process of the layer: convolving the input data to generate a convolved feature map; pooling the convolution feature map to generate a pooled feature map; non-linearly activating the pooled feature map to generate an activated feature map; calculating the mean value of each position of the activation feature map along the channel dimension to generate a spatial feature matrix; calculating a Softmax-like function value of each position in the space feature matrix to obtain a space score matrix; calculating the position-wise dot multiplication of the spatial feature matrix and the spatial score matrix to obtain a feature matrix; the characteristic matrix of the last layer output of the convolutional neural network model using the spatial attention mechanism is the characteristic matrix of the spatial enhanced cell motion trail.

In the above method for capturing a circulating tumor cell, the developing the plurality of motion trajectory sub-matrices into a plurality of motion trajectory sub-feature vectors, and then obtaining a plurality of context motion trajectory sub-vectors by a context encoder based on a converter includes: one-dimensional arrangement is carried out on the plurality of motion trail sub-feature vectors so as to obtain a motion trail global feature vector; calculating the product between the global motion track feature vector and the transpose vector of each motion track sub-feature vector in the plurality of motion track sub-feature vectors to obtain a plurality of self-attention association matrixes; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and weighting each motion track sub-feature vector in the motion track sub-feature vectors by taking each probability value in the probability values as a weight so as to obtain the context motion track sub-vectors.

In the above method for capturing circulating tumor cells, calculating euclidean distances between every two contextual motion trajectory sub-vectors of the plurality of contextual motion trajectory sub-vectors to obtain a classification feature direction consisting of a plurality of euclidean distances comprises: calculating a plurality of Euclidean distances between every two contextual movement track sub-vectors of the contextual movement track sub-vectors according to the following distance formula; wherein, the distance formula is:

wherein,representing each two contextual motion trail sub-vectors of the plurality of contextual motion trail sub-vectors,/->Representing the contextual motion trail subvector +.>Characteristic values of the respective positions in>Representing cosine distances between every two contextual motion trail subvectors of the plurality of contextual motion trail subvectors; and arranging the Euclidean distances to obtain the classification feature vector.

Compared with the prior art, the microfluidic chip and the method thereof suitable for capturing the circulating tumor cells acquire a cell motion trail graph acquired by a microscope; then, combining the technology based on deep learning and artificial intelligence, the electric field gradient is adaptively adjusted based on the motion track global map of different types of cells so as to improve the separation accuracy and efficiency. Therefore, the gradient strength of the electric field can be automatically adjusted according to the motion track characteristics of the cells, so that the capturing efficiency and accuracy of the cells are improved, and the separation effect and quality of the cells are further improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments or the description of the prior art will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is an application scenario diagram of a microfluidic chip suitable for capturing circulating tumor cells according to an embodiment of the present application.

Fig. 2 is a block diagram of a microfluidic chip suitable for capturing circulating tumor cells according to an embodiment of the present application.

Fig. 3 is a block diagram of the context encoding module in a microfluidic chip adapted for capturing circulating tumor cells according to an embodiment of the present application.

Fig. 4 is a block diagram of the variance measurement module in a microfluidic chip adapted for capturing circulating tumor cells according to an embodiment of the present application.

Fig. 5 is a block diagram of the control result generation module in the microfluidic chip adapted for capturing circulating tumor cells according to an embodiment of the present application.

FIG. 6 is a flow chart of a method suitable for capturing circulating tumor cells according to an embodiment of the present application.

Fig. 7 is a schematic diagram of a system architecture suitable for a method of circulating tumor cell capture according to an embodiment of the present application.

Detailed Description

The following description of the technical solutions in the embodiments of the present application will be made with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.

Unless defined otherwise, all technical and scientific terms used in the examples of this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to limit the scope of the present application.

In the description of the embodiments of the present application, unless otherwise indicated and defined, the term "connected" should be construed broadly, and for example, may be an electrical connection, may be a communication between two elements, may be a direct connection, or may be an indirect connection via an intermediary, and it will be understood by those skilled in the art that the specific meaning of the term may be understood according to the specific circumstances.

It should be noted that, the term "first\second\third" in the embodiments of the present application is merely to distinguish similar objects, and does not represent a specific order for the objects, it is to be understood that "first\second\third" may interchange a specific order or sequence where allowed. It is to be understood that the "first\second\third" distinguishing objects may be interchanged where appropriate such that the embodiments of the present application described herein may be implemented in sequences other than those illustrated or described herein.

Aiming at the technical problems, the technical concept of the application is to combine the technology based on deep learning and artificial intelligence, and the electric field gradient is adaptively adjusted based on the motion trail global map of different types of cells so as to improve the accuracy and efficiency of separation.

Specifically, in the technical scheme of the present application, first, a cell motion trajectory image acquired by a microscope is acquired. Here, the cell motion trajectory graph can intuitively reflect the motion state and behavior of the cell in the electric field gradient. The real-time monitoring and feedback of the cells can be realized by collecting the cell motion trail graph through a microscope.

And then, the cell motion trail graph is used for obtaining a space enhanced cell motion trail feature matrix through a convolution neural network model using a space attention mechanism. Here, the spatial attention mechanism may adaptively select important spatial regions to extract cell motion trajectory features related to the electric field gradient. Specifically, the spatial attention mechanism can be implemented by using a convolutional neural network model, by performing a convolutional operation on an input cell motion trajectory graph, and then calculating a weight of each position by using the spatial attention mechanism, representing the importance of the position to the electric field gradient intensity. And finally, multiplying the weight by the corresponding feature map to obtain a space enhanced cell motion trail feature matrix. Thus, the motion rule of the cells in the electric field gradient can be effectively captured.

In order to extract local information of a cell motion track and thereby better capture the change of cells in an electric field gradient, in the technical scheme of the application, the space-enhanced cell motion track feature matrix is subjected to feature matrix division so as to respectively pay attention to the cell motion tracks of different areas, and therefore a plurality of motion track submatrices are obtained. In this way, the judgment ability of the electric field gradient strength is improved.

And then, respectively expanding the plurality of motion trail submatrices into motion trail submatrices, and then obtaining a plurality of context motion trail submatrices through a context encoder based on a converter. Here, the converter-based context encoder is a neural network model that utilizes a self-attention mechanism (self-attention) to capture context information in an input sequence. Specifically, by the context encoder based on the converter, the motion trajectory sub-feature vectors can be context encoded, that is, the relationship between each motion trajectory sub-feature vector and the overall motion trajectory sub-feature vector is considered by using a self-attention mechanism, so that richer and more accurate motion trajectory sub-vectors are obtained.

Further, euclidean distances between every two contextual motion trail sub-vectors of the contextual motion trail sub-vectors are calculated to obtain a classification feature vector composed of the multiple Euclidean distances. Here, the euclidean distance (Euclidean Distance) can measure the absolute distance between two points in a high dimensional space. In the present solution, each contextual movement trace sub-vector may be regarded as a point, which represents the movement characteristics of the cells in the local area. That is, the Euclidean distance between every two context motion trajectory subvectors is calculated, which can reflect the motion differences of cells in different regions. If the cell is subjected to an electric field gradient, its trajectory will change, resulting in an increase in Euclidean distance.

And then, the classification feature vector is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the electric field gradient strength is enhanced. Wherein, the electric field gradient strength refers to the degree of variation of the electric field strength in space, which reflects the non-uniformity of the electric field. The higher the gradient strength of the electric field, the more uneven the electric field, and the greater the force on the cell. In the technical scheme of the application, the classifier is used for judging whether the electric field gradient strength needs to be enhanced according to the characteristics of the cell motion trail. If the cell motion profile indicates that the cell is not sufficiently subjected to an electric field force to overcome the fluid resistance and other interference factors, resulting in the cell deviating from the target location or escaping from the microchannel, the classifier outputs a positive class indicating that the electric field gradient strength needs to be enhanced. If the cell motion profile indicates that the cells have been effectively captured and separated, the classifier outputs a negative class indicating that no enhancement of the electric field gradient strength is required. Thus, through the output of the classifier, the real-time adjustment of the gradient strength of the electric field can be realized, thereby improving the efficiency and accuracy of capturing and separating cells.

In the technical scheme of the application, when the cell motion trajectory graph is obtained by using a convolutional neural network model of a spatial attention mechanism, due to a local spatial distribution strengthening mechanism of the spatial attention mechanism, significant feature distribution imbalance exists among the plurality of motion trajectory submatrices obtained by dividing the spatial attention mechanism by the spatial strengthening cell motion trajectory feature matrix, and although context-dependent image feature semantic coding is performed by a context encoder based on a converter, namely, the correlation of feature distribution among the plurality of context motion trajectory submatrices is enhanced, the plurality of context motion trajectory submatrices still have relatively obvious explicit differences of feature distribution, so that when the Euclidean distance between every two context motion trajectory submatrices of the plurality of context motion trajectory submatrices is calculated, the spatial distance fluctuation caused by the explicit differences brings about poor convergence of the plurality of Euclidean distances in the class probability density space of the classifier on the class probability expression of the classifier, and influences the accuracy of the classifier obtained by the classifier.

Thus, the applicant of the present application classifies the feature vectorThe reference line meshing of the manifold curved surface with Gaussian probability density is specifically expressed as:

wherein the method comprises the steps ofAnd->Is a feature value set +.>Mean and standard deviation of (2), and->Is the +.f of the classification feature vector after optimization>Characteristic values of the individual positions.

The standard line meshing of the manifold curved surface with Gaussian probability density takes the statistical characteristics, namely the mean value and standard deviation, of the high-dimensional feature set of the classification feature vector as standard anchor points of probability density measurement, and the low-dimensional constraint expression of the neighborhood network of the local probability density extremum is obtained through line meshing along the local linear embedding direction of the manifold curved surface, so that the local distribution based on the neighborhood distribution is constrained based on the reference-based relative spatial position relation of the local distribution of the high-dimensional feature through reconstructing the probability density expression of the manifold curved surface, and therefore, the spatial convergence of the class probability density of the high-dimensional feature of the classification feature vector, namely, the consistency of the probability density expression of the classification feature vector in the probability density space is improved, and the accuracy of the classification result obtained by the classification feature vector through a classifier is improved.

The application has the following technical effects: 1. an intelligent electric field gradient strength control scheme is provided. 2. According to the scheme, the gradient strength of the electric field can be automatically adjusted according to the motion track characteristics of the cells, the capturing efficiency and accuracy of the cells are improved, and the separation effect and quality of the cells are improved.

Fig. 1 is an application scenario diagram of a microfluidic chip suitable for capturing circulating tumor cells according to an embodiment of the present application. As shown in fig. 1, in this application scenario, first, a cell motion trajectory graph (e.g., C as illustrated in fig. 1) acquired by a microscope (e.g., M as illustrated in fig. 1) is acquired; the acquired cell motion profile is then input to a server (e.g., S as illustrated in fig. 1) deployed with a microfluidic chip control algorithm suitable for cyclic tumor cell capture, where the server is capable of processing the cell motion profile based on the microfluidic chip control algorithm suitable for cyclic tumor cell capture to generate a classification result indicating whether the electric field gradient is enhanced.

Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.

In one embodiment of the present application, fig. 2 is a block diagram of a microfluidic chip suitable for capturing circulating tumor cells according to an embodiment of the present application. As shown in fig. 2, a microfluidic chip 100 suitable for capturing circulating tumor cells according to an embodiment of the present application includes: an image acquisition module 110 for acquiring a cell motion trajectory graph acquired by a microscope; the space enhancement module 120 is configured to obtain a space enhanced cell motion trajectory feature matrix from the cell motion trajectory graph by using a convolutional neural network model of a spatial attention mechanism; the feature matrix dividing module 130 is configured to perform feature matrix division on the feature matrix of the motion trail of the space-enhanced cell to obtain a plurality of motion trail submatrices; the context coding module 140 is configured to spread the motion trajectory sub-matrices into a plurality of motion trajectory sub-feature vectors, and obtain a plurality of context motion trajectory sub-vectors through a context encoder based on a converter; the difference measurement module 150 is configured to calculate euclidean distances between every two contextual motion trajectory sub-vectors of the plurality of contextual motion trajectory sub-vectors to obtain a classification feature vector composed of a plurality of euclidean distances; the manifold curved surface optimizing module 160 is configured to perform manifold curved surface optimization of gaussian probability density on the classification feature vector to obtain an optimized classification feature vector; and a control result generating module 170, configured to pass the optimized classification feature vector through a classifier to obtain a classification result, where the classification result is used to indicate whether to enhance the electric field gradient strength.

Specifically, in the embodiment of the present application, the image acquisition module 110 is configured to acquire a cell motion trajectory chart acquired by a microscope. Aiming at the technical problems, the technical concept of the application is to combine the technology based on deep learning and artificial intelligence, and the electric field gradient is adaptively adjusted based on the motion trail global map of different types of cells so as to improve the accuracy and efficiency of separation.

Specifically, in the embodiment of the present application, the spatial enhancement module 120 is configured to obtain the spatial enhanced cell motion trajectory feature matrix from the cell motion trajectory graph by using a convolutional neural network model of a spatial attention mechanism. And then, the cell motion trail graph is used for obtaining a space enhanced cell motion trail feature matrix through a convolution neural network model using a space attention mechanism.

Here, the spatial attention mechanism may adaptively select important spatial regions to extract cell motion trajectory features related to the electric field gradient. Specifically, the spatial attention mechanism can be implemented by using a convolutional neural network model, by performing a convolutional operation on an input cell motion trajectory graph, and then calculating a weight of each position by using the spatial attention mechanism, representing the importance of the position to the electric field gradient intensity. And finally, multiplying the weight by the corresponding feature map to obtain a space enhanced cell motion trail feature matrix. Thus, the motion rule of the cells in the electric field gradient can be effectively captured.

Wherein, the space enhancement module 120 is configured to: each layer of the convolutional neural network model using the spatial attention mechanism performs the following steps on input data in the forward transfer process of the layer: convolving the input data to generate a convolved feature map; pooling the convolution feature map to generate a pooled feature map; non-linearly activating the pooled feature map to generate an activated feature map; calculating the mean value of each position of the activation feature map along the channel dimension to generate a spatial feature matrix; calculating a Softmax-like function value of each position in the space feature matrix to obtain a space score matrix; calculating the position-wise dot multiplication of the spatial feature matrix and the spatial score matrix to obtain a feature matrix; the characteristic matrix of the last layer output of the convolutional neural network model using the spatial attention mechanism is the characteristic matrix of the spatial enhanced cell motion trail.

The attention mechanism is a data processing method in machine learning, and is widely applied to various machine learning tasks such as natural language processing, image recognition, voice recognition and the like. On one hand, the attention mechanism is that the network is hoped to automatically learn out the places needing attention in the picture or text sequence; on the other hand, the attention mechanism generates a mask by the operation of the neural network, the weights of the values on the mask. In general, the spatial attention mechanism calculates the average value of different channels of the same pixel point, and then obtains spatial features through some convolution and up-sampling operations, and the pixels of each layer of the spatial features are given different weights.

Specifically, in the embodiment of the present application, the feature matrix dividing module 130 is configured to perform feature matrix division on the feature matrix of the motion trail of the spatially enhanced cell to obtain a plurality of motion trail submatrices. In order to extract local information of a cell motion track and thereby better capture the change of cells in an electric field gradient, in the technical scheme of the application, the space-enhanced cell motion track feature matrix is subjected to feature matrix division so as to respectively pay attention to the cell motion tracks of different areas, and therefore a plurality of motion track submatrices are obtained. In this way, the judgment ability of the electric field gradient strength is improved.

Specifically, in the embodiment of the present application, the context encoding module 140 is configured to obtain a plurality of context motion trajectory sub-vectors by using a context encoder based on a converter after expanding the plurality of motion trajectory sub-matrices into a plurality of motion trajectory sub-feature vectors respectively. And then, respectively expanding the plurality of motion trail submatrices into motion trail submatrices, and then obtaining a plurality of context motion trail submatrices through a context encoder based on a converter. Here, the converter-based context encoder is a neural network model that utilizes a self-attention mechanism (self-attention) to capture context information in an input sequence. Specifically, by the context encoder based on the converter, the motion trajectory sub-feature vectors can be context encoded, that is, the relationship between each motion trajectory sub-feature vector and the overall motion trajectory sub-feature vector is considered by using a self-attention mechanism, so that richer and more accurate motion trajectory sub-vectors are obtained.

Fig. 3 is a block diagram of the context encoding module in the microfluidic chip adapted for capturing circulating tumor cells according to an embodiment of the present application, and as shown in fig. 3, the context encoding module 140 includes: a query vector construction unit 141, configured to perform one-dimensional arrangement on the plurality of motion trail sub-feature vectors to obtain a motion trail global feature vector; a self-attention unit 142, configured to calculate a product between the global feature vector of the motion trajectory and a transpose vector of each motion trajectory sub-feature vector of the plurality of motion trajectory sub-feature vectors to obtain a plurality of self-attention correlation matrices; a normalization unit 143, configured to perform normalization processing on each of the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices; the attention calculating unit 144 is configured to obtain a plurality of probability values by using a Softmax classification function for each normalized self-attention correlation matrix in the plurality of normalized self-attention correlation matrices; and a weighting unit 145, configured to weight each motion trajectory sub-feature vector in the plurality of motion trajectory sub-feature vectors with each probability value in the plurality of probability values as a weight, so as to obtain the plurality of context motion trajectory sub-vectors.

The context encoder aims to mine for hidden patterns between contexts in the word sequence, optionally the encoder comprises: CNN (Convolutional Neural Network ), recurrent NN (RecursiveNeural Network, recurrent neural network), language Model (Language Model), and the like. The CNN-based method has a better extraction effect on local features, but has a poor effect on Long-Term Dependency (Long-Term Dependency) problems in sentences, so Bi-LSTM (Long Short-Term Memory) based encoders are widely used. The repetitive NN processes sentences as a tree structure rather than a sequence, has stronger representation capability in theory, but has the weaknesses of high sample marking difficulty, deep gradient disappearance, difficulty in parallel calculation and the like, so that the repetitive NN is less in practical application. The transducer has a network structure with wide application, has the characteristics of CNN and RNN, has a better extraction effect on global characteristics, and has a certain advantage in parallel calculation compared with RNN (RecurrentNeural Network ).

Specifically, in the embodiment of the present application, the difference measurement module 150 is configured to calculate euclidean distances between every two contextual motion trajectory sub-vectors of the plurality of contextual motion trajectory sub-vectors to obtain a classification feature vector composed of a plurality of euclidean distances. Further, euclidean distances between every two contextual motion trail sub-vectors of the contextual motion trail sub-vectors are calculated to obtain a classification feature vector composed of the multiple Euclidean distances. Here, the euclidean distance (Euclidean Distance) can measure the absolute distance between two points in a high dimensional space. In the present solution, each contextual movement trace sub-vector may be regarded as a point, which represents the movement characteristics of the cells in the local area. That is, the Euclidean distance between every two context motion trajectory subvectors is calculated, which can reflect the motion differences of cells in different regions. If the cell is subjected to an electric field gradient, its trajectory will change, resulting in an increase in Euclidean distance.

Fig. 4 is a block diagram of the variance measurement module in the microfluidic chip suitable for capturing circulating tumor cells according to an embodiment of the present application, as shown in fig. 4, the variance measurement module 150 includes: a euclidean distance calculating unit 151 for calculating a plurality of euclidean distances between every two of the plurality of contextual movement trace sub-vectors in accordance with the following distance formula; wherein, the distance formula is:

wherein,representing each two contextual motion trail sub-vectors of the plurality of contextual motion trail sub-vectors,/->Representing the contextual motion trail subvector +.>Characteristic values of the respective positions in>Representing cosine distances between every two contextual motion trail subvectors of the plurality of contextual motion trail subvectors; and an arrangement unit 152 configured to arrange the plurality of euclidean distances to obtain the classification feature vector.

Specifically, in the embodiment of the present application, the manifold surface optimization module 160 is configured to perform manifold surface optimization of the gaussian probability density on the classification feature vector to obtain an optimized classification feature vector. In the technical scheme of the application, when the cell motion trajectory graph is obtained by using a convolutional neural network model of a spatial attention mechanism, due to a local spatial distribution strengthening mechanism of the spatial attention mechanism, significant feature distribution imbalance exists among the plurality of motion trajectory submatrices obtained by dividing the spatial attention mechanism by the spatial strengthening cell motion trajectory feature matrix, and although context-dependent image feature semantic coding is performed by a context encoder based on a converter, namely, the correlation of feature distribution among the plurality of context motion trajectory submatrices is enhanced, the plurality of context motion trajectory submatrices still have relatively obvious explicit differences of feature distribution, so that when the Euclidean distance between every two context motion trajectory submatrices of the plurality of context motion trajectory submatrices is calculated, the spatial distance fluctuation caused by the explicit differences brings about poor convergence of the plurality of Euclidean distances in the class probability density space of the classifier on the class probability expression of the classifier, and influences the accuracy of the classifier obtained by the classifier.

Thus, the applicant of the present application classifies the feature vectorThe reference line meshing of the manifold curved surface with Gaussian probability density is specifically expressed as: optimizing the manifold curved surface of the Gaussian probability density of the classification feature vector by using the following optimization formula to obtain an optimized classification feature vector; wherein, the optimization formula is:

wherein,and->Is characteristic of each position in the classified feature vectorMean and standard deviation of the value sets, +.>Is the +.o of the classification feature vector>Characteristic value of individual position, and->Is the +.f of the optimized classification feature vector>Characteristic values of the individual positions.

Specifically, in the embodiment of the present application, the control result generating module 170 is configured to pass the optimized classification feature vector through a classifier to obtain a classification result, where the classification result is used to indicate whether to enhance the electric field gradient strength. And then, the classification feature vector is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the electric field gradient strength is enhanced. Wherein, the electric field gradient strength refers to the degree of variation of the electric field strength in space, which reflects the non-uniformity of the electric field. The higher the gradient strength of the electric field, the more uneven the electric field, and the greater the force on the cell. In the technical scheme of the application, the classifier is used for judging whether the electric field gradient strength needs to be enhanced according to the characteristics of the cell motion trail.

If the cell motion profile indicates that the cell is not sufficiently subjected to an electric field force to overcome the fluid resistance and other interference factors, resulting in the cell deviating from the target location or escaping from the microchannel, the classifier outputs a positive class indicating that the electric field gradient strength needs to be enhanced. If the cell motion profile indicates that the cells have been effectively captured and separated, the classifier outputs a negative class indicating that no enhancement of the electric field gradient strength is required. Thus, through the output of the classifier, the real-time adjustment of the gradient strength of the electric field can be realized, thereby improving the efficiency and accuracy of capturing and separating cells.

Fig. 5 is a block diagram of the control result generation module in the microfluidic chip suitable for capturing the circulating tumor cells according to an embodiment of the present application, as shown in fig. 5, the control result generation module 170 includes: a full-connection encoding unit 171, configured to perform full-connection encoding on the optimized classification feature vector by using a plurality of full-connection layers of the classifier to obtain an encoded classification feature vector; and a classification unit 172, configured to pass the encoded classification feature vector through a Softmax classification function of the classifier to obtain the classification result.

In summary, a microfluidic chip 100 suitable for capturing circulating tumor cells, which acquires a cell motion profile acquired by a microscope, is illustrated in accordance with an embodiment of the present application; then, combining the technology based on deep learning and artificial intelligence, the electric field gradient is adaptively adjusted based on the motion track global map of different types of cells so as to improve the separation accuracy and efficiency. Therefore, the gradient strength of the electric field can be automatically adjusted according to the motion track characteristics of the cells, so that the capturing efficiency and accuracy of the cells are improved, and the separation effect and quality of the cells are further improved.

In one embodiment of the present application, fig. 6 is a flow chart of a method suitable for circulating tumor cell capture according to an embodiment of the present application. As shown in fig. 6, a method for circulating tumor cell capture according to an embodiment of the present application comprises: 210, acquiring a cell motion trail graph acquired by a microscope; 220, obtaining a space-enhanced cell motion trail feature matrix from the cell motion trail graph through a convolutional neural network model using a space attention mechanism; 230, performing feature matrix division on the space-enhanced cell motion trail feature matrix to obtain a plurality of motion trail submatrices; 240, respectively expanding the plurality of motion trail submatrices into a plurality of motion trail submatrices, and then obtaining a plurality of context motion trail submatrices through a context encoder based on a converter; 250, calculating Euclidean distances between every two contextual motion trail sub-vectors of the contextual motion trail sub-vectors to obtain a classification feature vector consisting of a plurality of Euclidean distances; 260, performing manifold curved surface optimization of Gaussian probability density on the classification feature vector to obtain an optimized classification feature vector; and, passing the optimized classification feature vector through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the electric field gradient strength is enhanced or not.

Fig. 7 is a schematic diagram of a system architecture suitable for a method of circulating tumor cell capture according to an embodiment of the present application. As shown in fig. 7, in the system architecture of the method for capturing circulating tumor cells, firstly, a cell motion trace map collected by a microscope is acquired; then, the cell motion trail graph is subjected to a convolutional neural network model by using a spatial attention mechanism to obtain a spatial reinforced cell motion trail feature matrix; then, performing feature matrix division on the space-enhanced cell motion trail feature matrix to obtain a plurality of motion trail submatrices; then, the plurality of motion trail submatrices are respectively unfolded into a plurality of motion trail submatrices, and then a context encoder based on a converter is used for obtaining a plurality of context motion trail submatrices; then, calculating Euclidean distances between every two contextual motion trail sub-vectors of the contextual motion trail sub-vectors to obtain a classification feature vector consisting of a plurality of Euclidean distances; then, manifold curved surface optimization of Gaussian probability density is carried out on the classification feature vector so as to obtain an optimized classification feature vector; and finally, the optimized classification feature vector is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the electric field gradient strength is enhanced or not

In a specific example, in the above method for capturing circulating tumor cells, the step of obtaining the spatial enhanced cell motion trajectory feature matrix by using a convolutional neural network model of spatial attention mechanism includes: each layer of the convolutional neural network model using the spatial attention mechanism performs the following steps on input data in the forward transfer process of the layer: convolving the input data to generate a convolved feature map; pooling the convolution feature map to generate a pooled feature map; non-linearly activating the pooled feature map to generate an activated feature map; calculating the mean value of each position of the activation feature map along the channel dimension to generate a spatial feature matrix; calculating a Softmax-like function value of each position in the space feature matrix to obtain a space score matrix; calculating the position-wise dot multiplication of the spatial feature matrix and the spatial score matrix to obtain a feature matrix; the characteristic matrix of the last layer output of the convolutional neural network model using the spatial attention mechanism is the characteristic matrix of the spatial enhanced cell motion trail.

In a specific example, in the above method for capturing a circulating tumor cell, the developing the plurality of motion trajectory submatrices into a plurality of motion trajectory feature vectors respectively, and then obtaining a plurality of context motion trajectory feature vectors by a context encoder based on a converter includes: one-dimensional arrangement is carried out on the plurality of motion trail sub-feature vectors so as to obtain a motion trail global feature vector; calculating the product between the global motion track feature vector and the transpose vector of each motion track sub-feature vector in the plurality of motion track sub-feature vectors to obtain a plurality of self-attention association matrixes; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and weighting each motion track sub-feature vector in the motion track sub-feature vectors by taking each probability value in the probability values as a weight so as to obtain the context motion track sub-vectors.

In a specific example, in the above method for capturing a circulating tumor cell, calculating euclidean distances between every two contextual motion trajectory sub-vectors of the plurality of contextual motion trajectory sub-vectors to obtain a classification feature direction composed of a plurality of euclidean distances includes: calculating a plurality of Euclidean distances between every two contextual movement track sub-vectors of the contextual movement track sub-vectors according to the following distance formula; wherein, the distance formula is:

In a specific example, in the above method for capturing circulating tumor cells, performing manifold surface optimization of gaussian probability density on the classification feature vector to obtain an optimized classification feature vector includes: optimizing the manifold curved surface of the Gaussian probability density of the classification feature vector by using the following optimization formula to obtain an optimized classification feature vector; wherein, the optimization formula is:

Wherein,and->Is the mean and standard deviation of the feature value sets of each position in the classification feature vector, +.>Is the +.o of the classification feature vector>Characteristic value of individual position, and->Is the +.f of the optimized classification feature vector>Characteristic values of the individual positions.

In a specific example, in the above method for capturing circulating tumor cells, the optimizing classification feature vector is passed through a classifier to obtain a classification result, where the classification result is used to indicate whether to enhance the electric field gradient strength, and the method includes: performing full-connection coding on the optimized classification feature vector by using a plurality of full-connection layers of the classifier to obtain a coding classification feature vector; and passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.

It will be appreciated by those skilled in the art that the specific operation of the steps in the above method for capturing circulating tumor cells has been described in detail above with reference to the descriptions of the microfluidic chip for capturing circulating tumor cells of fig. 1 to 5, and thus, repetitive descriptions thereof will be omitted.

The present application also provides a computer program product comprising instructions which, when executed, cause an apparatus to perform operations corresponding to the above-described methods.

In one embodiment of the present application, there is also provided a computer readable storage medium storing a computer program for executing the above-described method.

It should be appreciated that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the forms of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects may be utilized. Furthermore, the computer program product may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

Methods, systems, and computer program products of embodiments of the present application are described in terms of flow diagrams and/or block diagrams. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

The basic principles of the present application have been described above in connection with specific embodiments, however, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are merely examples and not limiting, and these advantages, benefits, effects, etc. are not to be considered as necessarily possessed by the various embodiments of the present application. Furthermore, the specific details disclosed herein are for purposes of illustration and understanding only, and are not intended to be limiting, as the application is not intended to be limited to the details disclosed herein as such.

The block diagrams of the devices, apparatuses, devices, systems referred to in this application are only illustrative examples and are not intended to require or imply that the connections, arrangements, configurations must be made in the manner shown in the block diagrams. As will be appreciated by one of skill in the art, the devices, apparatuses, devices, systems may be connected, arranged, configured in any manner. Words such as "including," "comprising," "having," and the like are words of openness and mean "including but not limited to," and are used interchangeably therewith. The terms "or" and "as used herein refer to and are used interchangeably with the term" and/or "unless the context clearly indicates otherwise. The term "such as" as used herein refers to, and is used interchangeably with, the phrase "such as, but not limited to.

It is also noted that in the apparatus, devices and methods of the present application, the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent to the present application.

The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or terminal device comprising the element.

The foregoing description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the application to the form disclosed herein. Although a number of example aspects and embodiments have been discussed above, a person of ordinary skill in the art will recognize certain variations, modifications, alterations, additions, and subcombinations thereof.

Claims

1. A microfluidic chip adapted for capturing circulating tumor cells, comprising:

The control result generation module is used for enabling the optimized classification feature vector to pass through a classifier to obtain a classification result, and the classification result is used for indicating whether the electric field gradient strength is enhanced or not;

wherein, the difference measurement module includes:

a euclidean distance calculating unit for calculating a plurality of euclidean distances between every two contextual movement track sub-vectors of the plurality of contextual movement track sub-vectors in the following distance formula;

wherein, the distance formula is:

，

wherein,and->Representing the plurality of contextual motion trail sub-vectors every two contextual motion trail sub-vectors,and->Representing the contextual motion trail subvector +.>And->Characteristic values of the respective positions in>Representing cosine distances between every two contextual motion trail subvectors of the plurality of contextual motion trail subvectors; and

an arrangement unit, configured to arrange the plurality of euclidean distances to obtain the classification feature vector;

wherein, manifold curved surface optimizing module is used for: optimizing the manifold curved surface of the Gaussian probability density of the classification feature vector by using the following optimization formula to obtain an optimized classification feature vector;

wherein, the optimization formula is:

，

2. The microfluidic chip for capturing circulating tumor cells according to claim 1, wherein the space enhancement module is configured to: each layer of the convolutional neural network model using the spatial attention mechanism performs the following steps on input data in the forward transfer process of the layer:

convolving the input data to generate a convolved feature map;

pooling the convolution feature map to generate a pooled feature map;

non-linearly activating the pooled feature map to generate an activated feature map;

calculating the mean value of each position of the activation feature map along the channel dimension to generate a spatial feature matrix;

calculating a Softmax-like function value of each position in the space feature matrix to obtain a space score matrix; and

calculating the position-wise dot multiplication of the space feature matrix and the space score matrix to obtain a feature matrix;

the characteristic matrix of the last layer output of the convolutional neural network model using the spatial attention mechanism is the characteristic matrix of the spatial enhanced cell motion trail.

3. The microfluidic chip for capturing circulating tumor cells according to claim 2, wherein the context encoding module comprises:

the query vector construction unit is used for carrying out one-dimensional arrangement on the plurality of motion track sub-feature vectors to obtain a motion track global feature vector;

the self-attention unit is used for calculating the product between the global motion track feature vector and the transpose vector of each motion track sub-feature vector in the plurality of motion track sub-feature vectors to obtain a plurality of self-attention association matrixes;

the normalization unit is used for respectively performing normalization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices;

the attention calculating unit is used for obtaining a plurality of probability values through a Softmax classification function by each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and

and the weighting unit is used for weighting each motion track sub-feature vector in the motion track sub-feature vectors by taking each probability value in the probability values as a weight so as to obtain the context motion track sub-vectors.

4. The microfluidic chip for capturing circulating tumor cells according to claim 3, wherein the control result generation module comprises:

the full-connection coding unit is used for carrying out full-connection coding on the optimized classification feature vector by using a plurality of full-connection layers of the classifier so as to obtain a coding classification feature vector; and

and the classification unit is used for passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.

5. A method for capturing circulating tumor cells, comprising:

acquiring a cell motion trail graph acquired by a microscope;

the optimized classification feature vector passes through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the electric field gradient strength is enhanced or not;

the method for calculating the euclidean distance between every two contextual motion trail sub-vectors of the contextual motion trail sub-vectors to obtain a classification characteristic direction consisting of a plurality of euclidean distances comprises the following steps:

calculating a plurality of Euclidean distances between every two contextual movement track sub-vectors of the contextual movement track sub-vectors according to the following distance formula;

wherein, the distance formula is:

，

arranging the plurality of Euclidean distances to obtain the classification feature vector;

and performing manifold curved surface optimization of Gaussian probability density on the classification feature vector to obtain an optimized classification feature vector, wherein the method comprises the following steps of: optimizing the manifold curved surface of the Gaussian probability density of the classification feature vector by using the following optimization formula to obtain an optimized classification feature vector; wherein, the optimization formula is:

，

6. The method of claim 5, wherein the step of generating a spatially enhanced cell motion trajectory feature matrix from the cell motion trajectory graph using a convolutional neural network model of spatial attention mechanisms comprises: each layer of the convolutional neural network model using the spatial attention mechanism performs the following steps on input data in the forward transfer process of the layer:

convolving the input data to generate a convolved feature map;

pooling the convolution feature map to generate a pooled feature map;

7. The method of claim 6, wherein the expanding the plurality of motion trajectory submatrices into a plurality of motion trajectory subfeature vectors, respectively, and then passing through a context encoder based on a transducer to obtain a plurality of context motion trajectory subvectors, comprises:

one-dimensional arrangement is carried out on the plurality of motion trail sub-feature vectors so as to obtain a motion trail global feature vector;

calculating the product between the global motion track feature vector and the transpose vector of each motion track sub-feature vector in the plurality of motion track sub-feature vectors to obtain a plurality of self-attention association matrixes;

respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices;

obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and

And weighting each motion track sub-feature vector in the motion track sub-feature vectors by taking each probability value in the probability values as a weight so as to obtain the context motion track sub-vectors.