TW202032574A - Method and system for classifying cells and medical analysis platform - Google Patents

Abstract

The present disclosure provides a method for classifying cells. The method includes: performing an image training adjustment operation with a medical analysis platform so that the medical analysis platform includes at least one deep-learning model for performing cell image classification; obtaining at least one cell image data corresponding to one cell; inputting the at least one cell image data to the at least one deep-learning model in the medical analysis platform; processing, by the medical analysis platform, a plurality of cell mass images corresponding to the at least one cell image data to obtain a plurality of feature values; and combining the feature values to generate a classification data corresponding to the feature values for determining the type of the cell. In addition, a system for classifying cells and a medical analysis platform are disclosed herein.

Description

Cell classification method, system and medical analysis platform

This disclosure relates to a classification method and system, and particularly to a cell classification method and system.

Artificial reproduction technology continues to develop. The assisted artificial reproductive technology, in which embryos are cultivated through in vitro fertilization and then replanted into the uterus, provides infertile couples with a choice of reproduction. Therefore, embryo quality is a key factor for the successful implantation of embryos in the uterus. Traditionally, doctors or embryologists judge embryo quality by naked eyes. However, different physicians or embryologists may make different judgments on the same embryo image due to personal knowledge and experience. Therefore, how to improve the accuracy and efficiency of judging embryo quality is an important issue.

Some embodiments of the present disclosure relate to a cell classification method. The method includes: performing an image training adjustment operation by a medical analysis platform, so that the medical analysis platform includes at least one deep learning model for cell image classification; obtaining at least one cell image data corresponding to a cell; A cell image data is input into the at least one deep learning model in the medical analysis platform Type; processing multiple cell cluster images corresponding to at least one cell image data through the medical analysis platform to obtain multiple feature data; and combining the feature data to generate a classification data corresponding to the feature data for determining the The type of cell.

Other embodiments of this disclosure are related to a cell classification system. The system includes a terminal and a medical analysis platform. The terminal is used to obtain at least one cell image data corresponding to a cell. The medical analysis platform is connected to the terminal for classifying the at least one cell image data corresponding to the cell received from the terminal, and returning a classification data corresponding to the cell to the terminal. The medical analysis platform includes a processor. The processor is used to process the plural cell cluster images of the at least one cell image data to obtain plural characteristic data, and to combine the characteristic data to generate a classification data corresponding to the characteristic data for determining the type of the cell.

In order to make the above and other purposes, features, advantages and embodiments of this case more obvious and understandable, the description of the attached symbols is as follows:

100‧‧‧Cell classification method

S110, S120, S130, S140, S150, S160, S170, S180‧‧‧Step

400‧‧‧Cell Classification System

410‧‧‧Terminal

412‧‧‧Image acquisition device

414‧‧‧Input/Output Device

416‧‧‧Communication device

420‧‧‧Internet

430‧‧‧Medical Analysis Platform

432‧‧‧Server

434‧‧‧Processor

In order to make the above and other objectives, features, advantages and embodiments of the present disclosure more comprehensible, the description of the accompanying drawings is as follows: Figure 1 is the flow of a cell classification method according to an embodiment of the present disclosure Figure; Figure 2 is a schematic diagram of a statistical classification data according to an embodiment of the present disclosure; Figure 3 is a schematic diagram illustrating a classification of cell image data according to another embodiment of the present disclosure; and Figure 4 It is a schematic diagram illustrating a cell classification system according to an embodiment of the disclosure.

The following will illustrate the spirit of the present disclosure with drawings and detailed descriptions. Any person with ordinary knowledge in the relevant technical field who understands the preferred embodiments of the present disclosure can change and modify the techniques taught in the present disclosure. Not departing from the spirit and scope of this disclosure.

It should be understood that in the description herein and all subsequent patent scopes, when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or may There is an insert component. In contrast, when an element is said to be "directly connected" or "directly coupled" to another element, there is no intervening element. In addition, "electrical connection" or "connection" can also refer to the mutual operation or interaction between two or more elements.

It should be understood that in the description herein and all subsequent patent scopes, although words such as "first", "second", ... can be used to describe different elements, these elements should not be limited by these terms. These terms are only used to distinguish a single element from another element. For example, a first element may also be referred to as a second element, and similarly, a second element may also be referred to as a first element without departing from the scope of the embodiment.

It should be understood that in the description of this article and all subsequent patent scopes, the terms "include", "include", "have", "contain", etc. used in this article are all open terms, namely Means "including but not limited to".

It should be understood that in the description herein and all subsequent patent scopes, the use of "and/or" includes any one or more of the related listed items and all combinations thereof.

It should be understood that in the description of this article and all subsequent patent scope Unless otherwise defined, all terms used (including technical and scientific terms) have the same meanings as understood by those skilled in the art to which this disclosure belongs. It is further clear that, unless explicitly stated here, these terms, such as those defined in commonly used dictionaries, should be interpreted as having meanings consistent with their meanings in the context of the relevant field, and should not be idealized or excessively Explain formally.

If any element in the scope of the patent does not clearly state that "the device is used to implement a specific function, or that the "step is used to implement" a specific function, it should not be interpreted as a means function term.

Please refer to Figure 1. FIG. 1 is a flowchart of a cell classification method 100 according to an embodiment of the present disclosure. As shown in Figure 1, the cell classification method 100 first executes step S110. In step S110, the cell images used as training images and their corresponding classification data are collected. In some embodiments of this case, the above-mentioned collection steps can use a microscope or other photographic devices to collect image data of embryonic cells that have developed to the blastocyst stage, and then medical professionals such as obstetrics and gynecology Doctors, embryologists, etc., mark the classification data of the quality of blastocyst cells in each cell image data.

Specifically, to determine the quality of blastocyst cells, three types of characteristics must be observed separately and independently. The first category is the fullness of the blastocoel (Blastocoel) and the state of the zona pellucida, which can be divided into 1 to 6 levels according to the degree of expansion of the blastocoel. Grade 1 means that the volume of the blastocyst cavity is less than 50% of the total embryo volume, while grade 6 means that the hatched blastocysts have all overflowed from the zona pellucida, which is the most mature type of blastocyst. The second category is the number and arrangement of Inner Cell Mass (ICM), which can be divided into A, B, and C levels. Among them, grade A means that the number of cells is large and tightly connected, and the quality is the best, while grade C means that the number of cells is small. The third category is the number and arrangement of trophoblast cells (Trophectoderm cell, TE) Situations can be divided into A, B, and C levels. Among them, grade A means it is made up of very many cells with the best quality, while grade C means it is made up of very few and large cells. Therefore, in some embodiments, the cell image data of each blastocyst will have 3 characteristic classifications. For example, "5AA" represents the state of expansion degree 5, inner cell mass grade A, and trophoblast cell grade A. Since the inner cell mass is the main structure for the development of the fetus, the trophoblast cells are the main structure for the development of the placenta. Therefore, the quality of the cells is critical to the development of the fetus. At the same time, the better the quality of the blastocyst, the better the implantation rate, which also reduces the probability of miscarriage and ectopic pregnancy.

Next, in step S120, the image obtained in S110 for training is preprocessed for subsequent analysis and training. In some embodiments of this case, the image is converted into a 264 pixel by 198 pixel image file with the original image's proportions preserved, and then the histogram can be used for equalization to average the brightness of the image while also enhancing Contrast of local area.

On the other hand, in step S130, a pre-trained neural network model is constructed. In some embodiments, the above construction steps first take the convolutional layer and its weights in the pre-trained neural network in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) as a model The base. For example, the neural network may be a residual neural network model ResNet50 (Residual Neural Network) pre-trained using the Imagenet database. Then, in some embodiments of this case, in order to determine the quality of blastocysts, a new neural network layer, such as a Full Connected layer (FC), is connected to the aforementioned model base to form a neural network Classifier to classify blastocyst quality. In addition, one or more machine learning software libraries such as Keras and TensorFlow can be used to implement neural network classifiers. It should be noted that the state of the case is here The surface is not restricted. The neural network classifier is not limited to a single neural network, but can be any other suitable type of classifier, or a combination of each other. For example, the neural network classifier may be a combination of a Convolutional Neural Network (CNN) and a Recurrent Neural Network (RNN).

Then, input the pre-processed cell image data into the pre-trained neural network model, and perform step S140 to perform an image training and adjustment operation to generate a deep learning model based on the neural network. In some embodiments, the above-mentioned training and adjustment operation first divides the preprocessed cell image data and its corresponding classification data into a training set, a verification set, and a test set. Then input the cell image data in the training set into the convolutional neural network model. The input image is processed alternately through a convolutional layer and a pooling layer to find a feature map or feature value, and at least one fully connected layer is used to process the feature map to generate a predictive classification corresponding to the cell image data. In some embodiments, because the output predicted classification is in error with the corresponding actual classification data marked by the medical professional, the loss function and optimizer are defined, and the backpropagation algorithm is used to calculate the difference between the predicted classification and the actual classification data. The error is propagated back to the input layer of the neural network, and the model is trained through the previous process until the optimal combination of parameters and model structure is found. Finally, use all cell image data and classification data, combined with the aforementioned optimal parameters and model structure to train and output the final model.

It should be noted that, taking some embodiments of this case as an example, in step S140, by connecting different neural network layers (such as the blastocyst cavity and zona pellucida of the blastocyst, the inner cell mass and the trophoblast cell) ) Can produce different deep learning models. Specifically, a cell image data can be input into three deep learning models that connect different fully connected layers to perform neural network classifiers, and perform In the process of training and optimization, three final models are finally output. It should be understood that the embodiments described in this case are only for ease of understanding, but the implementation of this case is not limited thereto.

Then, in step S150, a medical analysis platform is used to provide cell quality discrimination services. In some embodiments of the present case, the medical analysis platform includes at least one deep learning model described in step S140 to perform cell image classification. In addition, the medical analysis platform can also perform image training and adjustment operations on one or more deep learning models according to step S140.

After step S150, the platform can perform cell quality determination operations by receiving cell images and inputting the images into the deep learning model in the platform, that is, performing cell image classification operations as described in some embodiments of this case .

The above-mentioned cell images can be obtained in different ways, that is, according to actual subsequent analysis needs, a single or multiple cell images can be obtained by operating the image obtaining device at different times, with different methods or parameters. In some embodiments, the image acquisition device may be a camera, a microscope, a CCD camera or any device with the same function. In some embodiments of the present case, by adjusting the shooting focal length of a microscope, multiple cell image data corresponding to different focal planes in the longitudinal direction of the blastocyst can be obtained from deep to shallow to form multiple depth-sequence cell images. The method of adjusting the microscope can be a method of programmatically automatically controlling the focal length of the microscope or manually adjusting and shooting. For example, every time the focal length is reduced by 10 microns, a cell image is taken and stored in a storage device. After adjusting the focal length of the microscope within a specific focal length, the photos taken at this specific focal length are composed into a set of depth sequence cell images arranged according to the focal length, and they are input into the deep learning model of the medical analysis platform for processing.

In other embodiments, the image acquisition device can also be controlled programmatically. By means of setting or manually recording time, sequentially obtain multiple cell image data corresponding to blastocyst cells at different time points to form multiple time-series cell images. For example, an image of blastocyst cells is taken every 10 minutes and stored in a storage device. After a specific period of time, the photos in this specific period of time are assembled into a group of time-series cell images in the order of shooting time, and they are input into the deep learning model of the medical analysis platform for processing.

The above-mentioned cell image data can be stored in a hard disk, a memory, a database, a cloud data storage platform or any storage device, and be read out for analysis or processing at an appropriate time.

Then, in step S160, input the cell image data to be identified into the medical analysis platform. Specifically, the cell image data of a single sheet or sequence is input into at least one deep learning model in the medical analysis platform. In some embodiments of this case, in order to provide classification data of blastocyst cells, the characteristic classification data of blastocyst cavity expansion degree, inner cell mass and trophoblast layer are respectively required. Therefore, it is necessary to input the at least one blastocyst image data into the deep learning model corresponding to the expansion degree of the blastocyst cavity, the inner cell mass and the trophoblast cell layer. It should be understood that the embodiments described in this case are only for ease of understanding, but the implementation of this case is not limited thereto.

Further, in step S170, the medical analysis platform will perform a cell quality judgment operation based on the input cell image data. In some embodiments, multiple cell cluster images corresponding to at least one input cell image data to be identified are processed through a medical analysis platform to obtain multiple feature data. Specifically, the at least one multiple cell cluster image of the input cell image data to be determined is processed through at least one deep learning model in the medical analysis platform.

In some embodiments, the at least one deep learning model includes a The convolutional neural network model, so the above operation can use the convolutional neural network model to capture the complex feature value of the cell cluster in the cell image to be judged. For example, in some embodiments, a piece of blastocyst image data to be discriminated is input into the convolutional neural network of three deep learning models corresponding to the expansion degree of the blastocyst cavity, the inner cell mass and the trophoblast cells. In the input layer of the road. Then, at least one convolutional layer of the convolutional neural networks is used to capture the feature value of the cell image. These feature values can be shape, brightness or texture. Finally, through at least one hidden layer of the convolutional neural network, these feature values are connected to the fully connected layer (neural network classifier) to generate corresponding first feature classification data. For example, this blastocyst image has three deep learning models corresponding to the expansion of the blastocyst cavity, the inner cell mass and the trophoblast cells, and the corresponding three first feature classification data can be "4", "A", "B" and so on.

Then, the medical analysis platform combines the first feature classification data to generate classification data corresponding to the cell. As in the above embodiment, the medical analysis platform combines these independent first feature classification data ("4", "A", "B", etc.) to form "4AB" as the classification generated based on this cell image data data.

Then, in step S180, the classification data corresponding to the cell is output as the result of cell quality discrimination. In some embodiments, the medical analysis platform can transmit the judgment result to a device connected to the platform through the network, so that the operator using the device can use the judgment result to make subsequent judgments, for example: a blastocyst with this cell quality Whether the cells are suitable for implantation in the mother.

In some embodiments, the medical analysis platform can provide a cell quality discrimination service for multiple pieces of cell image data of a cell according to step S150. Please refer to Figure 2. Figure 2 is a statistical analysis based on an embodiment of the present disclosure Schematic diagram of class data. As shown in Figure 2, input multiple cell images 1~cell images n of the same cell into the convolutional neural network CNN1~CNNn to generate classification data 1~classification data n. Then, the number of occurrences of various classification data is counted, and the classification data with the most frequency is used as the final judgment result of the quality of this cell. For example, in some embodiments of this case, a total of 30 cell image data of a certain blastocyst are input to the medical analysis platform, and 30 classification data corresponding to the 30 cell image data are obtained after the neural network classifier , The classified data of "4AB" appears 15 times, the classified data of "4AA" appears 5 times, and the classified data of "4BB" appears 5 times. The medical analysis platform outputs the classification data "4AB" with the most occurrences as the category of this cell. It should be understood that the embodiments described in this case are only for ease of understanding, but the implementation of this case is not limited thereto.

In other embodiments, the cell classification method 100 can further process the depth sequence cell image and/or the time sequence cell image. Please refer to Figure 3. FIG. 3 is a schematic diagram illustrating the classification of cell image data according to another embodiment of the disclosure. Taking the processing of time-series cell images as an example, step S160 is executed to input the time-series cell image data to a medical analysis platform that further includes a recurrent neural network model in steps S130 to S160, which combines convolutional neural networks and recurrent neural networks The network model includes n-level networks. As shown in Figure 3, the part framed by the dotted line represents a hierarchical network, and each cell image in the time-series cell image is sequentially input to a corresponding hierarchical network.

Following the above, in step S170, in some embodiments, in each hierarchical network, at least one convolutional layer of the convolutional neural network model is used to capture the feature value of the cell cluster in the cell image, where the feature value can be Eigenvector Or matrix. In some embodiments, these feature vectors may be the first feature classification data, which is related to the graphical features of the spatial variation of the cell image data. After that, the first feature classification data is input to the recurrent neural network to capture and change with time. And the second feature classification data related to the generated graphic feature. At the same time, the second feature classification data will also be input to the recurrent neural network of the next level and the last level. Through the foregoing operations, the second feature classification data describing this cell can be passed down in the neural network, and the analysis result of a single-level network is not only generated by the convolutional neural network of that level network, but also Will be affected by the results of the previous several levels of network analysis. In the end, the recurrent neural network in the last-level network can generate pairs based on the feature vector provided by the convolutional neural network in the last-level network and the second feature classification data provided by the recurrent neural network at the previous levels. Classification data of this cell.

For example, please refer to Figure 3 again. At the frequency of taking one cell image of a cell every 10 minutes, 288 cell images are collected sequentially as cell image data within 2 days. Then, input the cell image data to the medical analysis platform for analysis operation. In the above analysis operation, 288 cell images were input into the corresponding 288 hierarchical networks according to the time sequence obtained, and then the feature vector of each image was obtained through the convolutional neural network in the hierarchical network. For example, a 1x2048 matrix is used as the first feature classification data. After that, the first feature classification data is input to the recurrent neural network in the hierarchical network to generate the second feature classification data, and the second feature classification generated by each hierarchical network will be input to the next layer and the 288th Hierarchical network. Finally, the 288th level network generates corresponding data based on the first feature classification data (feature vector) provided by its inner convolutional neural network and the second feature classification data provided by the previous 287 level networks. Classification of cells.

In some embodiments, by inputting the second feature classification data extracted from each level network to the next level and the last level of the recurrent neural network, the deep learning model of the medical analysis platform considers different time or different The image characteristics of a blastocyst cell in the deep formation sequence can improve the accuracy of the cell quality judgment.

It should be noted that the deep learning model based on the combination of the convolutional neural network and the recurrent neural network as described above is trained and generated in the same way as the deep learning generated based on the convolutional neural network in step S140. The models are similar, so I won't repeat them here.

Please refer to Figure 4. FIG. 4 is a schematic diagram illustrating a cell classification system 400 according to an embodiment of the disclosure. In some embodiments, the cell classification method 100 shown in FIG. 1 can be implemented by the cell classification system 400. But the realization of the cell classification method provided in this case is not limited to this.

As shown in FIG. 4, the cell classification system 400 includes a terminal 410, a network 420, and a medical analysis platform 430. The terminal 410 and the medical analysis platform 430 are connected to each other through the network 420. The terminal 410 is used to obtain at least one cell image data corresponding to a cell. In some embodiments of this case, the cell image data may include a blastocyst cell. Through the network 420, the terminal 410 transmits the cell image data to the medical analysis platform 430 for analysis. The medical analysis platform 430 is connected to the terminal 410 to classify the at least one cell image data corresponding to the cell received from the terminal 410, and return a classification data corresponding to the cell to the terminal 410.

As shown in FIG. 4, the terminal 410 may include an image acquisition device 412, an input/output device 414, and a communication device 416. The image acquisition device 412 is coupled to the input/output device 414 and the communication device 416. Input/output device 414 Coupled to the communication device 416. In some embodiments, the terminal 410 can be implemented by a personal computer, a tablet computer, a mobile device, or any device with the same function.

The medical analysis platform 430 may include a server 432 and a processor 434. The server 432 is coupled to the processor 434. In some embodiments, the server 432 may refer to a physical processor with functions such as related communication, data storage, or data processing, and is not limited herein. The processor 434 can be an integrated circuit such as a microcontroller, a microprocessor, a digital signal processor, an application specific integrated circuit (ASIC), a logic circuit or other Similar elements or a combination of the above elements are implemented.

Please refer to Figure 1 and Figure 4 together. The following will describe the cell classification method 100 implemented by the cell classification system 400 according to an embodiment of the present case.

After step S110 is performed, the medical analysis platform 430 can perform the operation of providing subsequent cell quality judgment in the subsequent step S150 by the processor 434 according to the at least one deep learning model according to the steps S120 to S140.

According to step S120, the processor 434 is used for preprocessing the collected cell image data, such as histogram equalization, so that the cell image data have the same graphic size and the contrast for subsequent analysis. After that, according to step S130, a pre-trained neural network model is constructed in the processor 434. The implementation is as described above, and will not be repeated here.

Then, according to step S140, the processor 434 is used to train the plurality of cell image data to be trained and the plurality of feature data corresponding to the cell image data to be trained. In some embodiments, the characteristic data may be classification data about the quality of blastocyst cells, such as the aforementioned classification data "5AA". In addition, when performing the training adjustment operation in step S140, some practical In an embodiment, all collected cell image data and corresponding classification data can be randomly allocated into training set and validation set. The processor 434 uses the data in the training set to train the pre-training model and generate a deep learning model. After that, the processor 434 inputs the data in the verification set to the deep learning model to generate feature data corresponding to the cell image data in the verification set. When the feature data generated by the processing of the verification cell image data by the processor 434 is different from the verification feature data corresponding to the verification cell image data, the parameters and/or model structure of the deep learning model in the processor 434 will be adjusted. Improve the accuracy of the deep learning model to determine cell quality.

Then, according to step S150, the medical analysis platform 430 that has a deep learning model and can provide cell quality judgment services is connected to the terminal 410 through the network 420. In some embodiments, the terminal 410 may be a computer or an image processing system built in a hospital or an artificial reproduction center, but this case is not limited to this.

The image obtaining device 412 in the terminal 410 is used to obtain the at least one cell image data. In some embodiments, the image acquisition device 412 captures cell image data of a cell, where the cell image data can be a single image, or the image acquisition device 412 sequentially acquires serial images at different time points or different slice depths. After that, the image obtaining device 412 transmits the obtained cell image data to the medical analysis platform 430 for analysis through the communication device 416 via the network 420 according to step S160.

Next, the server 432 in the medical analysis platform 430 is used to receive the cell image data transmitted from the communication device 416, and according to step S170, input the cell image data to the processor 434 for cell quality determination.

In some embodiments, the processor 434 is further used to process the plural cell cluster images of the at least one cell image data to obtain plural characteristic data, and use The characteristic data are combined to generate a classification data corresponding to the characteristic data for judging the type of the cell. In some embodiments of the present case, as described above, the plurality of cell cluster images may refer to the blastocoel and zona pellucida, inner cell cluster and trophoblast cells of blastocyst cells.

Specifically, the processor 434 is further configured to perform a first feature recognition operation based on a neural network on the image of the cell clusters to generate first feature data for the processor 434 to combine the feature data. In some embodiments, the neural network may be a convolutional neural network or any neural network model that extracts feature values or feature vectors from images. As in an embodiment of this case, the processor 434 performs the first feature recognition operation on the cell image data of a blastocyst cell. It is worth noting that this first feature identification operation corresponds to the expansion degree of blastocyst cells, inner cell mass and trophoblast cells, and generates corresponding first feature data as feature data, such as "4", "A", "B" etc. Finally, the processor 434 combines the characteristic data "4", "A" and "B" to form "4AB" as the classification data generated based on the blastocyst image data. Other implementation manners are as described in step S170, which will not be repeated here.

In another embodiment, when an image acquisition device 412 sequentially acquires cell image data at different time points or at different slice depths, the processor 434 is further used to perform neural-based imaging on the cell clusters in the cell image data. The second feature recognition operation of the network generates second feature data for the processor 434 to combine the feature data. The neural network model is not limited here. Specifically, please refer to Figure 3. Take, for example, cell image data obtained based on different slice depths. In some embodiments of the present case, the processor 434 inputs the blastocyst image data into a deep learning model that also includes a recurrent neural network according to the depth of the slice from deep to shallow. First, the processor 434 performs the first feature recognition operation through the convolutional neural network To obtain the feature vector of each image pattern in the cell image data as the first feature data. Then, by inputting the input of the first characteristic data into the corresponding recurrent neural network to perform the second characteristic identification operation, the corresponding second characteristic data is analyzed as the characteristic data. Finally, the processor 434 combines the characteristic data corresponding to the expansion degree, the inner cell mass and the trophoblast cell to generate the classification data of the blastocyst.

After the medical analysis platform 430 completes the cell quality identification service, according to step S180, the processor 434 transmits the classification data corresponding to the cells obtained by the image acquisition device 412 via the network 420 via the server 432 to the terminal. 410. The terminal 410 is received by the communication device 416 and then displayed on the input/output device 414 for relevant professionals to determine the next action based on the classified data.

In some embodiments, after the input/output device 414 receives and displays the classified data corresponding to the cell image data to relevant professionals, the professionals can input their judgments on the cell image data as feedback data based on their own knowledge and experience. In some embodiments, the input/output device 414 is further used to send feedback data to the medical analysis platform 430 through the communication device 416. When the feedback data does not match the classification data provided by the medical analysis platform 430, the processor 434 uses the feedback data and at least one cell image data to perform training again, so as to modify or update the deep learning model for performing cell quality judgment operations. In addition, in some embodiments, the server 432 medical analysis platform 430 can provide cell quality identification services to a plurality of different terminals 410 to determine at least one cell image data of at least one cell. These different The terminal 410 receives the corresponding feedback classification data to train the deep learning model.

Through the operations of the various embodiments above, a cell can be realized The classification method is to generate classification data of the result of analyzing a cell by at least one deep learning model, and in the processing process, a neural network can be trained based on a large amount of data to generate the analysis result. In addition, since the correlation between the cell images in the cell image data is also considered, the accuracy of cell quality determination can also be improved. In addition, through the aforementioned cell classification system, remote users can use the service of cell quality identification through the Internet and provide feedback at the same time.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention. Anyone who is familiar with this technique can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection of the present invention The scope shall be subject to those defined in the attached patent scope.

100‧‧‧Cell classification method

S110, S120, S130, S140, S150, S160, S170, S180‧‧‧Step

Claims (11)

A cell classification method includes: performing an image training and adjustment operation by a medical analysis platform, so that the medical analysis platform includes at least one deep learning model for cell image classification; and obtaining at least one cell image data corresponding to a cell; Inputting the at least one cell image data into the at least one deep learning model in the medical analysis platform; processing the plural cell cluster images corresponding to the at least one cell image data through the medical analysis platform to obtain plural characteristic data; and combining the features Data to generate a classification data corresponding to the characteristic data for judging the cell type. The cell classification method according to claim 1, wherein the step of obtaining the at least one cell image data corresponding to the cell comprises: obtaining the cells corresponding to different focal lengths in the longitudinal direction by adjusting the shooting focal length of an image obtaining device A plurality of cell image data on a plane form a plurality of depth sequence cell images; wherein the depth sequence cell images are respectively input to the at least one deep learning model in the medical analysis platform for processing. The cell classification method according to claim 1, wherein the step of obtaining the at least one cell image data corresponding to the cell comprises: sequentially obtaining a plurality of cell image data corresponding to the cell at different time points to form a plurality of cells Time series cell imaging; The time-series cell images are respectively input to the at least one deep learning model in the medical analysis platform for processing. The cell classification method according to claim 1, wherein the step of processing the cell mass images corresponding to at least one cell image data through the medical analysis platform comprises: processing the at least one deep learning model in the medical analysis platform The cell cluster images should be at least one cell image data, wherein the at least one deep learning model includes a convolutional neural network model; the convolutional neural network model is used to retrieve the plural number of the cell clusters corresponding to the cell image Characteristic value; and generating plural first characteristic classification data corresponding to the cell clusters according to the characteristic values; wherein the medical analysis platform is used for combining the characteristic data according to the first characteristic classification data. The cell classification method according to claim 4, further comprising: sequentially inputting the first feature classification data into at least one deep learning model in the medical analysis platform, wherein the at least one deep learning model further comprises a recurrent neural network Model; through the recurrent neural network model to generate a plurality of second feature classification data corresponding to the cell clusters, wherein the medical analysis platform is further used to combine the feature data according to the second feature classification data. A cell classification system includes: a terminal machine for obtaining at least one cell image data corresponding to a cell; and a medical analysis platform connected with the terminal machine for matching the cell received from the terminal machine The at least one cell image data is classified, and a classification data corresponding to the cell is returned to the terminal, wherein the medical analysis platform includes: a processor for processing a plurality of cell clusters of the at least one cell image data The image is used to obtain plural characteristic data and combine the characteristic data to generate a classification data corresponding to the characteristic data for judging the cell type. The cell classification system according to claim 6, wherein the processor is further configured to perform at least one first feature recognition operation based on a neural network on the cell cluster images respectively to generate a plurality of first feature data for the The processor combines the first characteristic data. The cell classification system according to claim 7, wherein the processor is further used to train a plurality of cell image data to be trained and a plurality of feature data corresponding to the cell image data to be trained; When the plurality of feature data generated by processing the plurality of verification cell image data is different from the verification feature data corresponding to the verification cell image data, the parameters of the processor are adjusted. The cell classification system described in claim 6, wherein When the image obtaining device sequentially obtains the plurality of cell cluster images of the at least one cell image data at different time points or at different slice depths, the processor is further used to execute at least one neural network-based image on the cell cluster images respectively Two feature identification operations are used to generate a plurality of second feature data for the processor to combine the feature data. The cell classification system according to claim 6, wherein the terminal includes: an image acquisition device for acquiring the at least one cell image data; and an input/output device connected to the image acquisition device and a communication device, Used for receiving the at least one cell image data from the image acquisition device, and after receiving the classification data corresponding to the at least one cell image data, the input/output device is used for returning a feedback data to the Medical analysis platform; wherein when the feedback data does not match the classification data, the processor retrains with the feedback data and the at least one cell image data. A medical analysis platform includes: a server for receiving at least one cell image data corresponding to a cell; and a processor connected to the server for the at least one cell based on at least one deep learning model A cell image data generates a classification data corresponding to the cell; wherein the at least one deep learning model is based on a plurality of cell shadows for training Image data and its classification information are generated by training, and/or generated by receiving at least one cell image data corresponding to at least one cell from at least one terminal and the feedback classification data training.

Images