Method and system for automatically analyzing cell monoclonal origin and storage medium

Abstract

A cell monoclonal origin automatic analysis method, system and storage medium, the method comprising: performing microscopic imaging on at least one pore site in a pore plate within a first preset time after planking cell sedimentation to obtain at least one candidate image (210); performing microscopic imaging on at least one pore site in the pore plate within a second preset time after planking cell sedimentation to obtain at least one target image (220); and acquiring an image recognition result, and analyzing the cell monoclonal origin according to the image recognition result (230).

Description

Method, system and storage medium for automatic analysis of cell monoclonal origin

[0001] Cross-references

[0002] This application claims priority to Chinese application No. 202211090661.9 filed on September 7, 2022, the entire contents of which are incorporated herein by reference.

[0003] The present invention relates to the field of biomedicine technology, and in particular to a method, font and storage medium for automatic analysis of cell monoclonal origin.

[0004] The monoclonal origin of engineered cells is the premise and guarantee of product expression and quality uniformity. At present, during the drug review process, appropriate experimental methods need to be used to prove the monoclonal origin of the cells used. It can be seen that confirming the monoclonal origin of engineered cells is of great significance to ensuring the product quality of the drug.

[0005] Therefore, it is desirable to provide a method, system and storage medium for automatic analysis of cell monoclonal origin, so as to realize automatic, efficient and accurate monoclonal origin analysis.

[0006]

[0007] One of the embodiments of the present specification provides a method for automatically analyzing the monoclonal origin of cells, the method comprising: performing microscopic imaging on at least one well in a well plate after the plated cells have settled to obtain at least one image; obtaining an image recognition result based on the at least one image, and analyzing the monoclonal origin of the cells according to the image recognition result, wherein the image recognition result comprises at least one of the following: a first recognition result obtained by performing image recognition on at least one target image, the at least one target image being an image obtained by performing microscopic imaging on the at least one well in the well plate within a second preset time after the plated cells have settled, the first recognition result comprising at least the number of cell clusters in each well in the at least one well and at least one of the first recognition frames of the cell clusters; or a second recognition result obtained by performing image recognition on at least one initial image included in at least one candidate image, the at least one candidate image being an image obtained by performing microscopic imaging on at least one well in the well plate within a first preset time after the plated cells have settled, the second recognition result comprising at least the number of cells in each well in the at least one well and at least one of the second recognition frames of the cells.

[0008] One of the embodiments of the present specification provides a system for automatic analysis of cell monoclonal origin, comprising: at least one computer-readable storage medium including a set of instructions, and at least one processing device communicating with the computer-readable storage medium, wherein when executing the set of instructions, the at least one processor is configured to: obtain at least one image by performing microscopic imaging of at least one well position in a well plate after the plated cells have settled; obtain an image recognition result based on the at least one image, and analyze the cell monoclonal origin according to the image recognition result, wherein the image recognition result includes at least one of the following: a first recognition result obtained by performing image recognition on at least one target image, the at least one target image; The candidate image is an image acquired by microscopic imaging of at least one well in the well plate within a second preset time after the plated cells have settled, and the first recognition result includes at least the number of cell clusters in each well of the at least one well and at least one of the first recognition frames of the cell clusters; or, a second recognition result obtained by image recognition of at least one initial image included in at least one candidate image, and the at least one candidate image is an image acquired by microscopic imaging of at least one well in the well plate within a first preset time after the plated cells have settled, and the second recognition result includes at least the number of cells in each well of the at least one well and at least one of the second recognition frames of the cells.

[0009] One of the embodiments of the present specification provides a computer-readable storage medium, wherein the storage medium stores computer instructions. When the computer reads the computer instructions in the storage medium, the computer executes a method for automatically analyzing the monoclonal origin of cells, wherein the method comprises: obtaining at least one image by performing microscopic imaging of at least one well position in a well plate after the plated cells have settled; obtaining an image recognition result based on the at least one image, and analyzing the monoclonal origin of cells according to the image recognition result, wherein the image recognition result comprises at least one of the following: a first recognition result obtained by performing image recognition on at least one target image, wherein the at least one target image is a first recognition result obtained by performing image recognition on the plated cells; The first recognition result comprises at least one of the number of cell clusters in each of the at least one well and at least one of the first recognition frames of the cell clusters; or, a second recognition result is obtained by performing image recognition on at least one initial image included in at least one candidate image, wherein the at least one candidate image is an image obtained by performing microscopic imaging on at least one well in the well plate within a first preset time after the plated cells have settled, and the second recognition result comprises at least one of the number of cells in each of the at least one well and at least one of the second recognition frames of the cells.

[0010] Some embodiments of this specification include at least the following beneficial effects:

[0011] (1) After cell plating, some cells may be suspended inside the wells, and some cells may be attached to the wells. The position of suspended cells usually changes during division, while the position of attached cells is usually fixed during division. By dividing the different states of plated cells in the wells and using different monoclonal origin analysis methods, targeted, effective and accurate monoclonal origin analysis can be performed on plated cells in different states;

[0012] (2) Conventional monoclonal origin identification methods usually perform monoclonal origin judgment before the cells begin to reproduce, or continuously observe the cell reproduction process and perform monoclonal origin judgment when the cells form cell clusters. In some embodiments of this specification, the number of cell clusters is first identified in the middle stage of cell division (such as selecting an arbitrary time point within the 4th to 7th day) to perform a preliminary analysis of the monoclonal origin of the cells; when the identification result is a cell cluster, the target well position is automatically backtracked according to the position information of the cell cluster in the first identification frame or according to template matching, and the corresponding initial image and the corresponding matching area are automatically backtracked; then, based on the matching image corresponding to the matching area, the number of cells is identified again, and the cell monoclonal origin analysis is performed according to the identification result of the matching image. Through backtracking, on the basis of the preliminary analysis of the monoclonal origin, it is possible to further accurately and efficiently judge whether the cells in the target well position have monoclonal origin according to the initial number of cells in the target well position, which has higher reliability, and effectively excludes the situation where multiple cell clusters formed by multiple cell divisions merge into a total cell cluster to affect the accuracy of the monoclonal origin analysis, which has higher reliability;

[0013] (3) According to the degree of cell proliferation, the time period for cell division to form a cell cluster of appropriate size is determined as the second preset time, which can effectively avoid the influence of too small a cell cluster on the recognition effect of the first recognition model, and can effectively avoid the cost consumption caused by too large a cell cluster, thereby improving the accuracy of monoclonal origin analysis. By predicting the time that is most conducive to judging the monoclonal origin of the cell (i.e., the target time), taking pictures near this time can improve the accuracy of judgment, and at the same time, it can reduce the shooting at other times except the target time, effectively saving resources.

[0014] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0015] FIG1 is a schematic diagram of an application scenario of an exemplary automatic analysis system according to some embodiments of this specification;

[0016] FIG2 is an exemplary flow chart of an automatic analysis method for cell monoclonal origin according to some embodiments of the present specification;

[0017] FIG3 is an exemplary flow chart of an automatic analysis method for the monoclonal origin of adherent cells according to some embodiments of the present specification;

[0018] FIG4 is an exemplary flow chart of obtaining a first recognition result according to some embodiments of this specification;

[0019] FIG5 is an exemplary flow chart of obtaining a second recognition result according to some embodiments of this specification;

[0020] FIG6 is an exemplary schematic diagram of at least one candidate image according to some embodiments of the present specification;

[0021] FIG7 is an exemplary schematic diagram of at least one target image according to some embodiments of the present specification;

[0022] FIG8 is an exemplary schematic diagram of obtaining a matching image based on a first recognition frame according to some embodiments of this specification;

[0023] FIG. 9a-FIG. 9d are exemplary schematic diagrams for analyzing the monoclonal origin of cells according to some embodiments of the present specification;

[0024] FIG10 is an exemplary schematic diagram of an application judgment model according to some embodiments of this specification;

[0025] FIG. 11 is an exemplary flowchart of determining a second preset time according to some embodiments of the present specification.

[0026] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0027] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0028] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0029] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0030] FIG. 1 is a schematic diagram of an application scenario of an exemplary automatic analysis system according to some embodiments of the present specification.

[0031] The imaging device 110 can be used to perform microscopic imaging of the plated cells in the well plate, wherein the plated cells are cells that settle in the well plate (e.g., well plate 110-1) after the cells are plated. In some embodiments, the well plate 110-1 can include a 6-well plate, a 48-well plate, a 96-well plate, a 384-well plate, etc. In some embodiments, the well plate 110-1 can also include a U-shaped hole, a V-shaped hole, an F-shaped hole, etc.

[0032] In some embodiments, the plated cells may be in a suspended or adherent state in the wells after settling. In some embodiments, the plated cells may include at least one of suspended cells and adherent cells. In some embodiments, after the cell plate is settled, there may be one cell, multiple cells, or even no cells in any well of the well plate 110-1. In some embodiments, one or more cells in the wells of the well plate 110-1 may split to form one or more cell clusters after a period of time.

[0033] In some embodiments, the imaging device 110 may include an ordinary optical microscope, an inverted microscope, an electron scanning microscope, etc. In some embodiments, the imaging device 110 may perform microscopic imaging of the plated cells in at least one well of the well plate 110-1, and capture corresponding imaging data (e.g., candidate images, etc.). The above description of the imaging device 110 is for illustrative purposes only and is not intended to limit the scope of this specification.

[0034] The processing device 120 may process data and / or information obtained from the imaging device 110, the terminal device 140, and / or the storage device 150. For example, the processing device 120 may obtain imaging data of plated cells in the well plate 110-1 from the imaging device 110 via the network 130; further, the processing device 120 may analyze the monoclonality of the cells based on the imaging data of the plated cells.

[0035] Fig. 2 is an exemplary flow chart of an exemplary automatic analysis method for cell monoclonal origin shown in some embodiments of this specification. In some embodiments, the automatic analysis method can be performed by an automatic analysis system 100 (e.g., a processing device 120). For example, process 200 can be stored in a storage device (e.g., a storage device 150) in the form of a program or instruction, and when the processing device 120 executes the program or instruction, process 200 can be implemented. The operational schematic diagram of process 200 presented below is illustrative. In some embodiments, one or more additional operations not described and / or one or more operations not discussed can be used to complete the process. In addition, the order of operations of process 200 shown in Fig. 2 and described below is not restrictive. In some embodiments, according to actual needs, the processing device 120 can only perform any one of steps 210 and 220, or the processing device 120 can also perform both steps 210 and 220.

[0036] Step 210, performing microscopic imaging on at least one well in the well plate within a first preset time after the plated cells have settled, and acquiring at least one candidate image.

[0037] Plated cells are cells that settle in the well plate after cell plating. In some embodiments, the cells can be subjected to limiting dilution and plated at a concentration of 0.5 to 1 cell / well. In some embodiments, plated cells can include adherent cells and suspended cells. Adherent cells are in an adherent state within the well, and suspended cells are in a suspended state within the well.

[0038] The first preset time refers to the time period during which the plated cells have not yet formed a cell mass of the appropriate size. For an explanation of the cell mass of the appropriate size, see step 220 and its related description. In some embodiments, the first preset time can be set to within 0 to 24 hours after the plated cells have settled. In some embodiments, the first preset time can be within 0 to 5 days after the plated cells have settled. The first preset time can be set according to actual needs and is not limited here.

[0039] In some embodiments, the first preset time may be a period of time before the second preset time. For more information about the second preset time, see below.

[0040] In some embodiments, the first preset time can be divided into multiple time intervals. In some embodiments, the multiple time intervals may include a time period when the plated cells have not yet undergone cell division (hereinafter referred to as the initial time interval), such as within 0 to 24 hours after the plated cells settle. In some embodiments, the multiple time intervals may include a time period when the plated cells begin to divide to a time period when cell clusters of a suitable size are not formed (hereinafter referred to as the intermediate time interval), such as within 1 to 5 days after the plated cells settle.

[0041] In some embodiments, the at least one hole position is any one or more hole positions in the orifice plate. In some embodiments, the at least one hole position can be all the hole positions in the orifice plate, or can be part of the hole positions in the orifice plate.

[0042] The candidate image is a microscopic image (eg, a whole-hole microscopic image) obtained when microscopic imaging is performed on one of the wells within the first preset time. In some embodiments, each well in the at least one well corresponds to a candidate image.

[0043] In some embodiments, at least one candidate image may include at least one initial image. The initial image may be a microscopic image obtained when microscopic imaging is performed on one of the hole positions during an initial time interval. In some embodiments, at least one candidate image may include at least one intermediate image. The intermediate image may be a microscopic image obtained when microscopic imaging is performed on one of the hole positions during an intermediate time interval.

[0044] In some embodiments, the processing device can use one or more imaging devices (e.g., imaging device 110) to perform microscopic imaging on at least one well position in the well plate within a first preset time after the plated cells settle, and take a candidate image corresponding to each well position. In some embodiments, the imaging device can include an ordinary optical microscope, an inverted microscope, an electron scanning microscope, etc. For example, the processing device can use one or more inverted microscopes to perform microscopic imaging on at least one well position in the well plate, and obtain an initial full-well microscopic image corresponding to each well position.

[0045] The processing device can perform microscopic imaging on at least one well position in the well plate in each time interval within the first preset time.

[0046] In some embodiments, the processing device can control one or more imaging devices to perform microscopic imaging on at least one hole position in the well plate in the initial time interval after the plated cells settle, and take the initial image corresponding to each hole position. As an example only, as shown in Figure 6, the processing device can control the imaging device to perform microscopic imaging on hole positions a, b, c, d, e, and f in the initial time interval, and take initial images 610, 620, 630, 640, 650 and 660 corresponding to hole positions a, b, c, d, e, and f respectively. It should be noted that in the initial time interval, the plated cells in at least one hole position in the well plate have not yet begun to divide and proliferate. Accordingly, the number of cells in the initial full-hole microscopic image obtained within the first preset time is the initial number, that is, the initial number of cells planted in each hole position when the cells are plated. As an example only, as shown in Figure 6, the initial number of cells in the initial image 610 is two, and the initial number of cells in the initial image 620 is two.

[0047] In some embodiments, the processing device can control one or more imaging devices to perform microscopic imaging on at least one well in the well plate during an intermediate time interval after the plated cells have settled, and take an intermediate image corresponding to each well.

[0048] In some embodiments, the processing device can control one or more imaging devices to perform multiple microscopic imaging of at least one well in the well plate in multiple time intervals within a first preset time after the plated cells have settled, and take candidate images corresponding to each well in multiple time intervals.

[0049] In some embodiments, the processing device may acquire a local microscopic image set of at least one hole position through an imaging device, and obtain a candidate image corresponding to the hole position by processing the local microscopic image set.

[0050] The local microscopic image set may refer to a set of microscopic images taken when microscopic imaging of a local position in a well is performed by an imaging device.

[0051] In some embodiments, the processing device may preprocess the local microscopic image set. Exemplarily, the preprocessing may include but is not limited to filtering and denoising. In some embodiments, the processing device may process the preprocessed local microscopic image set to obtain a candidate image. Exemplarily, the processing device may use a template matching algorithm to splice the local microscopic image set of each hole position into a corresponding microscopic image. Furthermore, the hole position boundary may be identified using an edge detection algorithm, a shape recognition algorithm, etc., and the spliced ​​microscopic image may be cropped according to the hole position boundary to obtain a candidate image corresponding to the hole position.

[0052] In some embodiments of the present specification, more detailed features of cells can be captured by taking local microscopic images using a high-power microscope, and more detailed features of cells can be displayed in the candidate images obtained through the local microscopic image set.

[0053] Step 220, performing microscopic imaging on at least one well in the well plate within a second preset time after the plated cells have settled, to obtain at least one target image.

[0054] The second preset time refers to the time period during which the plated cells undergo cell division to form a cell mass of appropriate size. Wherein, a cell mass of appropriate size refers to a cell mass of such size that the cell mass can be easily observed and distinguished through a microscope. In some embodiments, a cell mass of appropriate size can be determined based on the number of cell divisions. For example, a cell mass formed when the number of divisions is 4-7 times can be determined as a cell mass of appropriate size.

[0055] The second preset time can be determined in a variety of ways. In some embodiments, the second preset time can be determined based on prior knowledge or historical data. For example, it can be set to any time point within the 4th to 7th day after the plated cells settle. Accordingly, at least one hole position can be microscopically imaged at any time point within the 4th to 7th day after the plated cells settle. It is understood that if the second preset time is set earlier (for example, set to within the 1st to 2nd day after the plated cells settle), the number of cell divisions is small at this time, and cell clusters of appropriate size cannot be formed; if the second preset time is set later (for example, set to the 9th to 14th day after the plated cells settle), the number of cell divisions is large at this time, and the cell cluster area is too large. If there are multiple cell clusters in the hole position, it is not easy to distinguish the cell clusters to determine the number of cell clusters.

[0056] In some embodiments, the second preset time can be determined according to the cell cycle. When the cell cycle is different, the time to form a cell mass of appropriate size may be different, and the second preset time corresponding to cells of different cell cycles may be different. The cell cycle is related to factors such as the cell type. Different cell types have different times for cell division to form a cell mass of appropriate size. In some embodiments, the second preset time can be determined according to the cell type. For example, the cell cycle of CHO cells is 14h, and the cell cycle of Vero cells is 18.5h. Accordingly, it takes about 4 to 5 days for CHO cells to divide and form a cell mass of appropriate size. The second preset time of CHO cells can be set to within the 4th to 5th day, and the Vero cells divide and form a cell mass of appropriate size for about 5 to 6 days. The second preset time of Vero cells can be set to within the 5th to 6th day.

[0057] In some embodiments, the second preset time can be determined according to the time required for the size of the cell mass to reach a preset threshold. It is understandable that the size of the cell mass is affected by the culture environment in addition to the cell cycle of the cell itself. For example, it is affected by the quality of the culture medium, bacteria, temperature, carbon dioxide concentration, etc. In some embodiments, the second preset time can be automatically determined according to the algorithm. In some embodiments, determining the second preset time according to the algorithm may include: intermittently photographing the hole position (such as photographing the hole position once every 12 hours), using an image recognition model to identify the image data taken each time, and determining the size of the cell mass. When the size of the identified cell mass reaches the preset threshold, the time period between the current moment and the start of shooting is automatically determined as the second preset time.

[0058] In some embodiments, the second preset time may also be determined according to the degree of cell proliferation. For more information about this embodiment, see FIG. 11 and its related description.

[0059] In some embodiments of the present specification, the second preset time is determined by the relationship between the cell cluster size and the preset threshold value, which can take into account the influence of the culture environment on the cell proliferation rate to a certain extent, avoid the error caused by determining the second preset time only based on the cell cycle in the subsequent monoclonal origin analysis, and affect the accuracy of the monoclonal origin analysis.

[0060] The target image is a microscopic image obtained when microscopic imaging is performed on one of the well positions within the second preset time. In some embodiments, each well position in the at least one well position corresponds to a target image.

[0061] In some embodiments, the processing device can control one or more imaging devices to perform microscopic imaging on at least one well position in the well plate again within a second preset time after the plated cells settle, and take a target image corresponding to each well position. As an example only, as shown in FIG7 , the processing device can control the imaging device to perform microscopic imaging on well positions a, b, c, d, e, and f respectively within the second preset time, and take target images 710, 720, 730, 740, 750, and 760 corresponding to well positions a, b, c, d, e, and f, respectively. In some embodiments, the processing device can obtain a local microscopic image set of at least one well position through an imaging device, and obtain a target image corresponding to the well position by processing the local microscopic image set. Similar to the acquisition of candidate images, the method for acquiring the target image can refer to the method for acquiring the candidate image, which will not be repeated here.

[0062] When using an imaging device to perform microscopic imaging on at least one well in a well plate, it is not possible to obtain a full-well microscopic image of the entire well at one time due to the small visual range of the imaging device. In some embodiments of the present specification, a full-well microscopic image of the corresponding well is obtained by processing a set of local microscopic images of each well, so that a full-well microscopic image of the entire well can be obtained at one time, so as to understand the cell situation in the entire well and improve the accuracy of monoclonal origin analysis.

[0063] Step 230, obtaining image recognition results, and analyzing the cell monoclonal origin based on the image recognition results.

[0064] In some embodiments, the image recognition result includes at least one of the following: a first recognition result obtained by performing image recognition on at least one target image; or a second recognition result obtained by performing image recognition on at least one initial image included in at least one candidate image.

[0065] There are many ways to obtain the image recognition result. For example, the image recognition result can be obtained by manual recognition.

[0066] In some embodiments, the processing device may process at least one target image using a machine learning model to obtain a first recognition result.

[0067] The first recognition result refers to the result obtained by performing image recognition on the target image. In some embodiments, the first recognition result may include at least the number of cell clusters in each well of at least one well and at least one of the first recognition frames of the cell clusters.

[0068] The number of cell clusters refers to the number of cell clusters identified in a well. In some embodiments, a cell cluster comprising multiple cells can be formed after cell division. In some embodiments, when a single cell is present in a well, a single cell can form a cell cluster after multiple cell divisions, and the number of cell clusters in the well is one. In some embodiments, when multiple cells (e.g., 2, 3, etc.) are present in a well, multiple cells can form multiple cell clusters after multiple cell divisions, and the number of cell clusters in the well is multiple. In some embodiments, after multiple cell divisions, the corresponding cell clusters formed by each of the multiple cells may merge into one to form a total cell cluster, and accordingly, the number of cell clusters formed by multiple cell divisions in the well is one.

[0069] The first identification frame refers to a frame with a certain shape used to mark the cell cluster. As an example only, as shown in FIG7 , 710 is a target image of well a, 710 - 1 is a cell cluster in well a, and 710 - 2 is a first identification frame of the cell cluster 710 - 1.

[0070] In some embodiments, the first identification frame may be a geometric frame in the shape of a rectangle, a circle, etc. In some embodiments, the shape of the first identification frame may also be close to the outline shape of the edge of the cell cluster.

[0071] In some embodiments, the size of the first identification frame is close to the size of the cell cluster, which can reflect the shape and size of the cell cluster. In some embodiments, the first identification frame can also be presented in color. In some embodiments, the first identification frame of the cell cluster can be covered with a first mark, for example, the first mark is added to the red channel so that the first identification frame is presented in red, which is easy to observe and identify. For more information about color presentation, please refer to step 330 and its related description.

[0072] In some embodiments, the first identification frame may include position information and edge information of the cell cluster. The position information indicates the position of the cell cluster in the well position. For example, the position information may be the coordinate information of the first identification frame in the coordinate system of the well position. The edge information indicates the contour shape of the edge of the cell cluster. For example, the edge information may be the distance relationship between the first identification frame and the edge of the cell cluster. In some embodiments, the position information of the cell cluster may be used to determine its matching area in the corresponding initial image. For more information about the matching area, see step 320 and its related description.

[0073] In some embodiments, a first identification frame may include a cell cluster. In some embodiments, a first identification frame may also include multiple cell clusters. For example, multiple cell clusters identified in a well position may be marked with a first identification frame.

[0074] In some embodiments, when a first identification frame can include a cell cluster, the number of cell clusters can be reflected by the number of first identification frames. In some embodiments, the processing device can determine the number of cell clusters by the number of first identification frames. As an example only, as shown in FIG7 , when there is a first identification frame 710-2 in the target image 710 of well a, it can be determined that the number of cell clusters in well a is one.

[0075] In some embodiments, when a first identification frame includes multiple cell clusters, the processing device can determine the number of cell clusters by further identifying the image enclosed by the first identification frame.

[0076] In some embodiments, the first recognition frame may also be marked in the initial image. For example, the initial image of the hole position and the target image marked with the first recognition frame may be simultaneously input into the second recognition model to obtain the initial image marked with the first recognition frame and / or the initial image marked with the second recognition frame. For more information about the second recognition model and the second recognition frame, see below.

[0077] In some embodiments, the first identification result further includes the number of cells. The number of cells refers to the number of cells identified in a well.

[0078] In some embodiments, the processing device may obtain a first recognition model, process at least one target image based on the first recognition model, and obtain a first recognition result. For more information about the first recognition model, see FIG. 4 and its related description.

[0079] In some embodiments, the processing device may process at least one initial image using a machine learning model to obtain a second recognition result.

[0080] The second recognition result refers to the result obtained by performing image recognition on the initial image. In some embodiments, the second recognition result may include at least the number of cells in each well in at least one well and at least one of the second recognition frames of the cells.

[0081] The second identification frame refers to a frame with a certain shape used to mark cells. As an example only, as shown in FIG8 , a is the target well position, 820 is the initial image of the target well position a, and 820-2 and 820-3 are the second identification frames of the cells.

[0082] In some embodiments, similar to the first identification frame, the second identification frame may be a geometric frame in a rectangular, circular, or other shape.

[0083] In some embodiments, similar to the first identification frame, the second identification frame can be presented in color. In some embodiments, the second identification frame of the cell can be covered with a second marker, for example, the second marker is added to the green channel so that the second identification frame is presented in green for easy observation and identification. For more information about color presentation, see step 330 and its related description.

[0084] In some embodiments, the processing device may obtain a second recognition model, and process at least one initial image based on the second recognition model to obtain a second recognition result. For more information about the second recognition model, see FIG. 5 and its related description.

[0085] In some embodiments, the processing device can analyze the monoclonal origin of the cell according to the image recognition result in a variety of ways. For example, the processing device can determine the well position with one second identification frame as a monoclonal well position according to the image recognition result of the initial image.

[0086] For suspended cells, in some embodiments, the processing device may analyze the monoclonality of the cells according to preset rules based on the first recognition result of the target image.

[0087] In some embodiments, the first recognition result may include the number of cells in the cell cluster. In some embodiments, the first recognition result may also include a second recognition frame marking the cells.

[0088] In some embodiments, the preset rule may be a rule algorithm for determining the monoclonal origin of the cell based on the first recognition result of the target image of the suspended cell.

[0089] In some embodiments, the preset rule may be: comparing the identified cell number with a preset cell number range, and determining the well position corresponding to the target image with a cell number within the preset cell number range as a monoclonal well position. In other words, the processing device may directly determine the monoclonal well position based on the relationship between the cell number and the preset cell number range without entering the review process.

[0090] In some embodiments, the preset rule may also be: comparing the identified cell number with a preset cell number range, determining the well position corresponding to the target image whose cell number is within the preset cell number range as the well position to be reviewed; based on the second recognition result of the intermediate image of the well position to be reviewed, determining whether the well position to be reviewed is a monoclonal well position.

[0091] The preset cell number range is related to the cell number. For example, the preset cell number range can be a cell number between 30 and 50. The preset cell number range can be determined in a variety of ways. In some embodiments, the preset cell number range can be determined by the system or manually based on prior knowledge or historical data.

[0092] In some embodiments, when the number of cells in the first recognition result of the target image corresponding to the well position meets the preset cell number range, the well position can be determined as a well position to be reviewed. When the number of cells in the first recognition result of the target image corresponding to the well position is not within the preset cell number range, the well position can be determined as a non-monoclonal well position. Among them, the well position to be reviewed refers to a well position that needs to be judged again for the monoclonal origin of the well position, and the non-monoclonal well position refers to a well position in which the cells in the well position do not have monoclonal origin.

[0093] In some embodiments, for the wells to be reviewed, the monoclonal origin of the wells to be reviewed can be verified and judged. There are many ways to verify and judge. For example, the target image of the well to be reviewed can be manually observed and judged. When it is determined that there is a cell of origin in the target image, it can be judged that the well to be reviewed has a monoclonal origin, that is, a monoclonal well. For another example, the number of cells in the well to be reviewed within the initial time interval can be manually judged. When the number of cells in the well to be reviewed within the initial time interval is one, it can be judged that the well to be reviewed has a monoclonal origin.

[0094] In some embodiments, for a non-monoclonal well position, it can be manually determined whether to verify the monoclonal origin of the well position.

[0095] In some embodiments, the processing device may also perform image recognition on at least one intermediate image acquired within the intermediate time interval of the first preset time, and obtain a second recognition result for each intermediate image. For example, each time after microscopic imaging of well a, the image may be recognized to obtain the number of cells and the second recognition frame of well a at each imaging time point. In some embodiments, the processing device may determine whether the well to be reviewed is a monoclonal well based on the second recognition result of the intermediate image of the well to be reviewed. In some embodiments, when the monoclonal origin of the well to be reviewed is verified and judged, auxiliary judgment may be made based on the number of cells of the well to be reviewed at other imaging time points (i.e., imaging time points within the intermediate time interval). For example, when the number of cells at multiple other imaging time points meets the division law, it may be determined whether the well to be reviewed is a monoclonal well. Through this embodiment, the influence of abnormal division and the like on the judgment of monoclonal origin can be screened out, and the analysis accuracy of monoclonal origin can be effectively improved.

[0096] In some embodiments, for adherent cells, the processing device may determine a well with one cell cluster as a target well based on the first recognition result of the target image, retrospectively determine an initial image of the target well taken within the initial time interval, and perform a monoclonal origin analysis in combination with the second recognition result of the initial image. For more description of this embodiment, see FIG3.

[0097] After cell plating, some cells may be suspended inside the wells, and some cells may be attached to the wells. The position of suspended cells usually changes during division, while the position of attached cells is usually fixed during division. By dividing the different states of plated cells in the wells and using different monoclonal origin analysis methods, targeted, effective and accurate monoclonal origin analysis can be performed on plated cells in different states.

[0098] Fig. 3 is an exemplary flow chart of an exemplary automatic analysis method for the monoclonal origin of adherent cells shown in some embodiments of this specification. In some embodiments, the automatic analysis method can be performed by an automatic analysis system 100 (e.g., a processing device 120). For example, process 300 can be stored in a storage device (e.g., a storage device 150) in the form of a program or instruction, and when the processing device 120 executes the program or instruction, process 300 can be implemented. The operational schematic diagram of process 300 presented below is illustrative. In some embodiments, the process can be completed using one or more additional operations not described and / or one or more operations not discussed. In addition, the order of the operations of process 300 shown in Fig. 3 and described below is not restrictive.

[0099] Step 310: Based on the first recognition result, a well position with one cell cluster is determined as a target well position.

[0100] The target well position may be a well position with one cell cluster. In some embodiments, a cell cluster in the target well position may be formed by the division of a single cell, or may be formed by the fusion of multiple cells after division. For more information about the first recognition result, see step 230 and its related description.

[0101] In some embodiments, the processing device may determine a hole position having one first identification frame as a target hole position according to the number of first identification frames identified in at least one target image. As an example only, as shown in FIG7 , the number of first identification frames in hole position a is one, the number of first identification frames in hole position b is two, the number of first identification frames in hole position c is one, the number of first identification frames in hole position d is two, the number of first identification frames in hole position e is one, and the number of first identification frames in hole position f is one. The processing device may determine hole positions a, hole position c, hole position e, and hole position f as target hole positions.

[0102] In some embodiments, when the processing device uses the first recognition model to obtain the recognition result of the number of cell clusters in each well of at least one well, it can also output the confidence of the number of cell clusters corresponding to each well. For more description of the first recognition model, see FIG. 4 .

[0103] In some embodiments, the processing device may determine the wells with a confidence level less than a preset threshold as verification wells, and verify the number of cell clusters in the verification wells. In some embodiments, the method for verifying the number of cell clusters in the verification wells includes: acquiring a target image of the verification well after a preset time; based on the target image of the verification well, acquiring the number of cell clusters and the first identification frame of the cell clusters in the verification well, and acquiring the confidence level of the number of cell clusters. Among them, the preset threshold and the preset time can be a system default value, an experience value, a manually preset value, etc., or any combination thereof, and can be set according to actual needs, and this specification does not limit this.

[0104] In some embodiments, the processing device may determine a verification well position with one cell cluster and a confidence level greater than a preset threshold as a target well position.

[0105] In some embodiments, after the target well position is determined, culture medium may continue to be added to the target well position, and culture medium may not continue to be added to other well positions except the target well position.

[0106] Conventional monoclonal origin identification methods usually require continuous observation for 14 days before performing monoclonal origin analysis. In some embodiments of this specification, a preliminary analysis of the monoclonal origin of cells can be performed in the middle stage of cell cloning (such as selecting an arbitrary time point within the 4th to 7th day) to avoid the situation in which it is inconvenient to distinguish the number of cell clusters when observing cell clusters in the early and late stages of cell cloning because the size of the cell clusters is not suitable for observation. At the same time, the wells that do not have monoclonal origin in the preliminary analysis results can be screened out to exclude the wells where the number of cell clusters is greater than one. There is no need to supplement the culture medium for the wells that do not have monoclonal origin or continue to shoot and observe, which is conducive to improving the analysis efficiency.

[0107] Step 320 : determining a matching image in at least one candidate image based on the target hole position, and determining a second recognition result corresponding to the matching image.

[0108] In some embodiments, the matching image refers to the initial image corresponding to the target hole position. In some embodiments, the matching image may also be a partial image area of ​​the initial image corresponding to the target hole position.

[0109] The candidate images at least include the initial image. The processing device can determine the matching image in at least one initial image based on the target hole position in a variety of ways. For more information about the candidate images and the initial image, see FIG. 2.

[0110] In some embodiments, when the automatic analysis system 100 acquires the initial image and the target image of at least one hole position, the initial image and the target image may be marked according to the hole position number, and the marking result may be stored in a storage device (e.g., storage device 150) in real time. When the processing device acquires the matching image of the target hole position, the corresponding initial image may be directly retrieved from the storage device by the hole position number corresponding to the target hole position to determine it as the matching image of the target hole position. For example, when the automatic analysis system 100 acquires the initial image and the target image of hole position a, the initial image and the target image taken at hole position a may be marked with the hole position number (e.g., a), and the marking result may be stored in the storage device. Accordingly, when hole position a is the target hole position, the processing device may directly retrieve the corresponding initial image from the storage device by the hole position number of hole position a to determine it as the matching image of hole position a.

[0111] In some embodiments, the processing device may determine a matching area of ​​the target hole position on at least one initial image based on the first recognition frame; and obtain a matching image based on a partial image corresponding to the matching area.

[0112] The matching area can be the area enclosed by the first recognition frame, and the corresponding area on the corresponding initial image. As an example only, as shown in FIG8 , a is the target hole position, 810 is the target image corresponding to the target hole position a, 810-1 is the first recognition frame, and 810-2 is the area enclosed by the first recognition frame 810-1. Accordingly, the area 820-1 corresponding to the area 810-2 on the corresponding initial image 820 is the matching area.

[0113] In some embodiments, when determining a matching image of a target well position in at least one initial image based on the first identification frame, the processing device may determine a matching area in the corresponding initial image based on the position information of the cell cluster in the first identification frame. For example, the processing device may determine a matching area in the corresponding initial image based on the coordinate range of the cell cluster in the target well position. For another example, if the cell cluster is located in the upper left corner of the well position, the matching area may be determined in the upper left corner of the corresponding initial image.

[0114] In some embodiments, the processing device may perform target matching based on the first recognition frame to determine a matching area of ​​the target hole position in at least one initial image. For ease of description, how to determine the matching area using template matching will be further described below through steps S1 and S2.

[0115] S1. Based on the arc edge feature of the target hole position, determine a first hole image and a second hole image from at least one initial image and at least one target image, wherein the first hole image is the initial image corresponding to the target hole position, and the second hole image is the target image corresponding to the target hole position.

[0116] The arc edge feature refers to the feature data at the edge of the target well. In some embodiments, the arc edge feature may include features on details such as texture, curvature, position and shape of edge wear. For example, the arc edge feature of a 96-well plate may be an arc, etc.

[0117] In some embodiments, the arc edge feature of the target hole position can be obtained by an edge detection algorithm. Exemplarily, the edge detection algorithm may include Canny operator edge detection, Hough transform detection, etc., which is not limited in this specification.

[0118] In some embodiments, the processing device can perform template matching based on the arc edge features of the target hole position, determine the hole position having similar arc edge features to the target hole position from at least one initial image and at least one target image, and determine the initial image corresponding to the hole position as the first hole image, and determine the target image corresponding to the hole position as the second hole image. It should be noted that the template matching algorithm here is the same type of algorithm as the template matching algorithm for splicing the local microscopic image set of each hole position into the corresponding full-hole microscopic image, but the specific parameters used are different and the template selection is also different.

[0119] S2. Perform template matching on the first hole image and the second hole image based on the first recognition frame to determine a matching area of ​​the target hole position on the first hole image.

[0120] In some embodiments, after the first hole image and the second hole image are determined, the processing device can determine the area enclosed by the first identification frame in the second hole image based on the first identification frame, and then by performing template matching between the first hole image and the second hole image, it can be determined that the area enclosed by the first identification frame in the first hole image is the matching area of ​​the target hole position on the first hole image. As an example only, as shown in FIG8, 810 is the second hole image (i.e., the target image corresponding to the target hole position a), 820 is the first hole image (i.e., the initial image corresponding to the target hole position a), the processing device can determine the area 810-2 enclosed by the first identification frame 810-1 in the second hole image 810 based on the first identification frame 810-1, and then by performing template matching between the first hole image 820 and the second hole image 810, it can be determined that the area 820-1 enclosed by the first identification frame 810-1 in the first hole image 820 is the matching area of ​​the target hole position a on the first hole image 820.

[0121] In some embodiments, the processing device may determine a matching ratio of the target hole position, and determine a matching area of ​​the target hole position on the at least one initial image based on the first recognition frame and the matching ratio.

[0122] The matching ratio refers to the ratio of the area of ​​the area enclosed by the first identification frame in the first hole image to be adjusted. In some embodiments, the matching ratio may be the ratio of the area of ​​the area enclosed by the first identification frame in the second hole image to the area enclosed by the first identification frame in the first hole image. For example, when the matching ratio is greater than 1, the area enclosed by the first identification frame in the first hole image may be enlarged; when the matching ratio is less than 1, the area enclosed by the first identification frame in the first hole image may be reduced. For another example, when the area enclosed by the first identification frame in the first hole image is directly determined as the matching area, the corresponding matching ratio is 1.

[0123] The matching ratio can be determined in a variety of ways. For example, the matching ratio can be pre-set by the system or manually based on prior knowledge or historical data.

[0124] In some embodiments, the processing device may determine the matching ratio based on the first identification frame coverage of the target well and the cell discreteness. For example, the processing device may determine the calculation result through a preset formula based on the first identification frame coverage of the target well and the cell discreteness; and determine the matching ratio based on the calculation result and the corresponding relationship between the calculation result and the matching ratio. The corresponding relationship between the calculation result and the matching ratio may be preset based on prior knowledge and historical experience.

[0125] In some embodiments, the first identification frame coverage can be determined based on the ratio of the total area of ​​all cell clusters in the well to the total area of ​​the first identification frame.

[0126] In some embodiments, the degree of cell discreteness can be determined based on the ratio of the number of cells in a well that are outside the first identification frame to the total number of cells in the well.

[0127] In some embodiments, the first recognition frame coverage and cell discreteness can be used to measure the recognition effect of the first recognition model on the cell cluster. When the first recognition frame coverage and cell discreteness are higher, it means that the first recognition model has a better recognition effect on the cell cluster.

[0128] In some embodiments, an exemplary preset formula is as follows:

[0129] Among them, u is the calculation result, m is the total number of holes, and p is i is the coverage rate of the first recognition frame of the i-th hole position, q i is the cell discreteness of the ith well, and a and b are preset coefficients.

[0130] The processing device can determine the matching area of ​​the target hole position on at least one initial image in a variety of ways based on the first identification frame and the matching ratio. In some embodiments, the processing device can determine the area enclosed by the first identification frame in the second hole image based on the first identification frame, and process the enclosed area according to the matching ratio to determine the matching area. For example, when the matching ratio is greater than 1, the enclosed area is enlarged according to the matching ratio to obtain the matching area. As an example only, as shown in FIG8, 810 is the second hole image, 820 is the first hole image, 810-2 is the area enclosed by the first identification frame 810-1 in the second hole image 810, and the area 810-2 can be enlarged according to the preset ratio to obtain the matching area 820-4.

[0131] In some embodiments, the processing device may obtain the matching image in a variety of ways based on the partial image corresponding to the matching area. For example, the partial image corresponding to the matching area may be directly determined as the matching image.

[0132] It can be understood that the matching area is a partial area in the initial image, and the pixels of the corresponding image are small, which is not conducive to subsequent processing. In some embodiments, the processing device can adjust the size of the partial image corresponding to the matching area to a preset size to obtain a matching image, so as to facilitate further processing of the matching image. As an example only, as shown in Figure 8, area 820-1 is the matching area of ​​the target hole position on the initial image. In some embodiments, 830 is the partial image corresponding to the matching area 820-1, and the processing device can adjust the size of the partial image 830 to a preset size to obtain a matching image 840. In some embodiments, 850 is the partial image corresponding to the matching area 820-4, and the processing device can adjust the size of the partial image 850 to a preset size to obtain a matching image 860.

[0133] In some embodiments, the preset size may be a system default value, an empirical value, a manually preset value, or any combination thereof, and may be set according to actual needs, and this specification does not limit this. For example, the preset size may be a pixel size of 512*512, etc.

[0134] In some embodiments, the processing device may determine the area to be filled according to the size relationship between the size of the partial image corresponding to the matching area and the preset size. Further, the processing device may fill the area to be filled with all-zero padding, and adjust the size of the partial image corresponding to the matching area to the preset size to obtain the matching image. As an example only, as shown in FIG8 , 830 is the partial image corresponding to the matching area, 840-1 is the area to be filled determined according to the size relationship between the size of the partial image 830 and the preset size, and the matching image 840 may be obtained by filling the area to be filled 840-1.

[0135] In some embodiments, filling may be performed by other methods, which are not limited in this specification.

[0136] In some embodiments, the processing device may further process at least one candidate image corresponding to the target well position based on the judgment model to determine whether the target well position has a cell division abnormality and its abnormality probability; in response to the existence of a cell division abnormality, screen out the target well position, and determine a matching image based on the screened target well position; in response to the absence of a cell division abnormality, determine a matching image based on the target well position. For more information about the judgment model, see FIG. 10.

[0137] In some embodiments, when a target well position is determined to have abnormal cell division, the processing device may screen out the target well position, determine a matching image in at least one candidate image based on the screened target well position (i.e., a target well position without abnormal cell division), and determine a second recognition result of the matching image, and then perform a cell monoclonal origin analysis based on the first recognition result and the second recognition result of the target well position. For more information on determining the matching image, see above. For more information on performing a cell monoclonal origin analysis based on the first recognition result and the second recognition result, see below.

[0138] In some embodiments, the matching image can be used to analyze the monoclonal origin of the cell. In some embodiments, the processing device can input the matching image into a second recognition model to determine a second recognition result corresponding to the matching image. For more information about the second recognition model, see Figure 5 and its related description, which will not be repeated here. For more information about the second recognition result, see step 230 and its related description.

[0139] In some embodiments of the present specification, by determining the matching image through template matching, the input of the second recognition model can include only the matching image of the target well position, thereby removing interference items in the image recognition process, effectively improving the accuracy and efficiency of image recognition. In addition, through template matching, the corresponding relationship between the same well position in the corresponding initial full-well microscopic image and the corresponding target full-well microscopic image can be obtained, so as to backtrack the number of cells in the target well position within the first preset time, which is conducive to the subsequent analysis of the monoclonal origin of the cells.

[0140] Step 330: Analyze the cell monoclonality based on the second identification result and the first identification result.

[0141] In some embodiments, the processing device can analyze the monoclonality of the cells based on the number of cells in the matching image. In some embodiments, the processing device can determine the target wells with zero cells as wells to be reviewed; determine the target wells with one cell as monoclonal wells; and determine the target wells with more than one cell as non-monoclonal wells, thereby analyzing the monoclonality of the cells.

[0142] In some embodiments, the processing device may determine the target well position with zero cells in the initial image as the well position to be reviewed. As an example only, as shown in FIG9a, 910 is the target image of the target well position g, and 920 is the initial image of the target well position g, wherein the target image 910 includes a first identification frame 910-1, 920-1 is another annotation form of the first identification frame 910-1 (it can also be regarded as the circumscribed rectangular frame of the first identification frame 910-1), and the first identification frame 920-1 can be annotated in the initial image 920 or in the target image 910. When the number of cells in the matching image of the target well position g (such as the initial image 920) is zero (the second identification frame is not annotated at this time), the target well position g can be determined as the well position to be reviewed.

[0143] In some embodiments, the processing device may determine a target well position in which the number of cells in the initial image is one as a monoclonal well position, wherein the cells in the monoclonal well position are monoclonal in origin. As an example only, as shown in FIG9b , 930 is a target image of target well position h, and 940 is an initial image of target well position h, wherein the target image 930 includes a first identification frame 930-1, 940-1 is another annotation form of the first identification frame 930-1 (can also be regarded as an external rectangular frame of the first identification frame 930-1), and 940-2 is a second identification frame. When the number of cells in the matching image of the target well position h (such as the initial image 940) is one (a second identification frame 940-2 is annotated at this time), the target well position h can be determined as a monoclonal well position.

[0144] In some embodiments, the first identification frame and the second identification frame can be marked in the initial image and the target image at the same time. As shown in FIG9b, the first identification frame 930-1 and the second identification frame 940-2 can be marked in the target image 930 of the target hole position h at the same time, and the first identification frame 940-1 and the second identification frame 940-2 can be marked in the initial image 940 of the target hole position h at the same time.

[0145] In some embodiments, the processing device can determine a target well position with more than one cell number as a non-monoclonal well position, wherein the cells in the non-monoclonal well position do not have monoclonal origin. For example, when the number of cells is two or three or more than one, the corresponding target well position can be determined to be a non-monoclonal well position, wherein the cells do not have monoclonal origin. As an example only, as shown in FIG9c, 950 is a target image of target well position i, and 960 is an initial image of target well position i, wherein the target image 950 includes a first identification frame 950-1, 960-1 is another annotation form of the first identification frame 950-1 (can also be regarded as an external rectangular frame of the first identification frame 950-1), 960-2, 960-3 are second identification frames, and when the number of cells in the matching image of the target well position i (such as the initial image 960) is two (the second identification frame 960-2 and the second identification frame 960-3 are annotated at this time), the target well position i can be determined as a non-monoclonal well position.

[0146] In some embodiments, when the processing device obtains the first identification frame of the cell cluster through the first identification model, it can also mark the origin position of the cell cluster, wherein the origin position of the cell cluster can be the center coordinates of the cell cluster. As an example only, as shown in FIG9b, 930-2 is the origin position of the cell cluster corresponding to the first identification frame 930-1; as shown in FIG9c, 950-2 and 950-3 are the origin positions of the cell cluster corresponding to the first identification frame 950-1.

[0147] In some embodiments, the processing device may also determine a well position with more than one cell cluster or zero cell cluster as a non-monoclonal well position. As an example only, as shown in FIG9d , 970 is a target image of well position j, and the number of first identification frames in target image 970 is zero. Correspondingly, the number of cell clusters is zero, and well position j can be determined to be a non-monoclonal well position. 980 is a target image of well position k, and the number of first identification frames in target image 980 is two (including first identification frame 980-1 and first identification frame 980-2), and correspondingly, the number of cell clusters is two, and well position k can be determined to be a non-monoclonal well position.

[0148] In some embodiments, the first identification frame and the second identification frame can be presented in the same or different colors. For example, when a single cell cluster is identified and the cell cluster is identified to be formed by the growth and reproduction of one cell, the first identification frame and the second identification frame can be presented in green; when multiple cell clusters are identified or the identified single cell cluster is formed by the growth and reproduction of multiple cells, the first identification frame and the second identification frame can be presented in red, or when the identified single cell cluster is formed by the growth and reproduction of multiple cells, the first identification frame can be presented in green and the second identification frame can be presented in red; when a single cell cluster is identified and the number of cells in the corresponding initial image is zero, the first identification frame and the second identification frame can be presented in yellow, or the first identification frame can be presented in green and the second identification frame can be presented in yellow.

[0149] In some embodiments, the partial image of the initial image in the second recognition frame can be enlarged and displayed in the target image. As an example only, as shown in FIG9b, the second recognition frame 940-2 and the corresponding partial image 940-3 of the initial image 940 can be enlarged and displayed in the target image 930. As shown in FIG9c, the second recognition frame 960-2 and the corresponding partial image 960-4 of the initial image 960 can be enlarged and displayed in the target image 950, and the second recognition frame 960-3 and the corresponding partial image 960-5 of the initial image 960 can be enlarged and displayed in the target image 950.

[0150] In some embodiments, when the second recognition model fails to recognize cells and / or fails to obtain the second recognition frame, in order to facilitate the review of the number of cells in the wells to be reviewed within the first preset time, the matching image corresponding to the matching area of ​​the first recognition frame in the initial image can also be copied and pasted, and covered in the target image. As an example only, as shown in FIG9a, no cells are recognized in the initial image 920 of well g and the second recognition frame is not obtained. Accordingly, the matching image corresponding to the matching area A of the first recognition frame 920-1 in the initial image 920 can be copied and pasted, and covered in the target image 910, that is, the image B enclosed by the first recognition frame 920-1 in the target image 910.

[0151] In some embodiments, the analysis results of the monoclonality of the cells can be displayed synchronously on the target image and / or the initial image of the well. For example, the analysis results can be represented by "True" and "False". "True" can indicate that the cells corresponding to the well have monoclonality or that the cells corresponding to the well may have monoclonality but need to be manually reviewed, and "False" can indicate that the cells corresponding to the well do not have monoclonality.

[0152] In some embodiments, similar to the presentation of the first identification frame and the second identification frame in different colors, the analysis results of monoclonality can also be presented in different colors. For example, a yellow font "True" can be used to indicate that the cells corresponding to the well position may have monoclonality, but manual review is required, a green font "True" can be used to indicate that the cells corresponding to the well position have monoclonality, and a red font "False" can be used to indicate that the cells corresponding to the well position do not have monoclonality. As an example only, as shown in FIG9a , the analysis result 910-2 on the target image 910 of well e and the analysis result 920-3 on the initial image 920 may be “True” in yellow font, indicating that the cells corresponding to the well may be monoclonal, but need to be manually reviewed; as shown in FIG9b , the analysis result 930-3 on the target image 930 of well f and the analysis result 940-4 on the initial image 940 may be “True” in green font, indicating that the cells corresponding to the well are monoclonal; as shown in FIG9c , the analysis result 950-4 on the target image 950 of well g and the analysis result 960-6 on the initial image 960 may be “False” in red font, indicating that the cells corresponding to the well are not monoclonal; as shown in FIG9d , the analysis result 970-1 on the target image 970 of well h and the analysis result 980-3 on the target image 980 of well i may be “False” in red font, indicating that the cells corresponding to the well are not monoclonal.

[0153] It is understandable that after cell plating, there may be one plating cell in the well, or there may be multiple plating cells. After multiple cells have undergone multiple cell divisions, the corresponding cell clusters formed by each of them may merge into one to form a total cell cluster. Correspondingly, the number of cell clusters formed by multiple cell divisions in the well is one. This situation will affect the accuracy and reliability of the cell monoclonal origin analysis. Conventional monoclonal origin identification methods usually perform monoclonal origin judgment when the cells have not yet begun to reproduce, or continuously observe the cell reproduction process and perform monoclonal origin judgment when the cells form cell clusters. In some embodiments of the present specification, the number of cell clusters is first identified in the middle stage of cell division (such as selecting an arbitrary time point within the 4th to 7th day), and the monoclonal origin of the cells is preliminarily analyzed; when the identification result is a cell cluster, the target well is automatically traced back according to the position information of the cell cluster in the first identification frame or according to template matching, and the corresponding initial image and the corresponding matching area are automatically traced back; then, based on the matching image corresponding to the matching area, the number of cells is identified again, and the monoclonal origin of the cells is analyzed according to the identification result of the matching image. Through backtracking, on the basis of the preliminary analysis of monoclonal origin, it is possible to further accurately and efficiently judge whether the cells in the well have monoclonal origin according to the initial number of cells in the target well, with higher reliability, and effectively exclude the situation where multiple cell clusters formed by multiple cell divisions merge into a total cell cluster to affect the accuracy of the monoclonal origin analysis, with higher reliability.

[0154] Fig. 4 is an exemplary flow chart for obtaining the first recognition result according to some embodiments of the present specification. In some embodiments, the acquisition method can be performed by the automatic analysis system 100 (e.g., the processing device 120). For example, process 400 can be stored in a storage device (e.g., storage device 150) in the form of a program or instruction, and when the processing device 120 executes the program or instruction, process 400 can be implemented. The operational schematic diagram of process 400 presented below is illustrative. In some embodiments, the process can be completed using one or more additional operations not described and / or one or more operations not discussed. In addition, the order of the operations of process 400 shown in Fig. 4 and described below is not restrictive.

[0155] Step 410: Obtain a first recognition model.

[0156] The first recognition model can be used to determine a first recognition result of the target image. In some embodiments, the first recognition model can be a semantic segmentation network (e.g., FCN, U-Net series network, etc.), or an instance segmentation network (e.g., Mask R-CNN, SOLO, etc.). In some embodiments, the first recognition result can include at least the number of cell clusters in each well position in at least one well position and at least one of the first recognition frames of the cell clusters. For more description of the first recognition result, see step 230.

[0157] In some embodiments, the input of the first recognition model may be at least one target image, and the output may be a first recognition result.

[0158] In some embodiments, before inputting the at least one target image into the first recognition model, the size of the at least one target image may be adjusted. In some embodiments, the size of the at least one target image may be adjusted to a preset size. The preset size used here is consistent with the preset size for adjusting the size of the partial image corresponding to the matching area. For more information about the preset size, please refer to step 320 and its related description, which will not be repeated here.

[0159] The size of the target image captured is usually large, which will affect the system's operating rate when processing the target image. Some embodiments in this specification can appropriately compress the size of the target image by adjusting the size of the target image, thereby increasing the network's operating rate and accelerating the algorithm. At the same time, the number of models used for processing can be reduced to reduce resource usage, thereby facilitating the deployment of the algorithm to the end device. In the process of identifying cell clusters, it is not necessary to switch models frequently, further improving the running speed of the algorithm.

[0160] In some embodiments, the processing device can obtain a first recognition model by training multiple first training samples with first labels. For example, multiple first training samples with first labels can be input into the initial first recognition model, a loss function can be constructed by the first label and the result of the initial first recognition model, and the parameters of the first recognition model can be iteratively updated based on the loss function. When the loss function of the initial first recognition model meets the preset conditions, the model training is completed and the trained first recognition model is obtained. The preset conditions may be that the loss function converges, the number of iterations reaches a threshold, etc.

[0161] In some embodiments, the first training sample may include a sample target image, and the first label corresponding to the first training sample may be at least one of the number of cell clusters corresponding to the sample target image and a first identification frame of the cell cluster. In some embodiments, the first training sample may be a target image collected during the sample cell division process, and the first label may be obtained by manual annotation.

[0162] Step 420: Process at least one target image based on the first recognition model to obtain a first recognition result.

[0163] In some embodiments, the processing device may obtain the number of cell clusters and a first identification frame of the cell clusters in each well of at least one well based on the output result of the first identification model.

[0164] In some embodiments of the present specification, the recognition of cell clusters and cells in an image can be achieved through a semantic segmentation network, effectively improving the efficiency and accuracy of image recognition.

[0165] FIG. 5 is an exemplary flow chart for obtaining a second recognition result according to some embodiments of the present specification. In some embodiments, the automatic analysis method can be performed by the automatic analysis system 100 (e.g., the processing device 120). For example, the process 500 can be stored in a storage device (e.g., the storage device 150) in the form of a program or an instruction, and when the processing device 120 executes the program or the instruction, the process 500 can be implemented. The operational schematic diagram of the process 500 presented below is illustrative. In some embodiments, the process can be completed using one or more additional operations not described and / or one or more operations not discussed. In addition, the order of the operations of the process 500 shown in FIG. 5 and described below is not restrictive.

[0166] Step 510: Obtain a second recognition model.

[0167] The second recognition model can be used to determine the second recognition result of the candidate image. In some embodiments, the second recognition model is consistent with the first recognition model structure, and can be a semantic segmentation network (e.g., FCN, U-Net series network, etc.), or an instance segmentation network (e.g., Mask R-CNN, SOLO, etc.). In some embodiments, the second recognition result includes at least the number of cells in each well position in at least one well position and at least one of the second recognition frames of the cells. For more description of the second recognition result, see step 230.

[0168] In some embodiments, the input of the second recognition model may be at least one initial image, and the output may be a second recognition result corresponding to the initial image.

[0169] In some embodiments, the input of the second recognition model may be a matching image, and the output may be a second recognition result corresponding to the matching image. It is worth noting that the size of the matching image is a preset size, which is consistent with the size of the target image input by the first recognition model. For more information about the matching image, see FIG. 3 and its related description, which will not be repeated here.

[0170] In some embodiments, the processing device can obtain the second recognition model through training. The training process of the second recognition model is similar to that of the first recognition model, except that the training samples and labels of the two models are different. For a description of the specific training process of the second recognition model, please refer to the training process of the first recognition model in FIG. 4, which will not be repeated here.

[0171] In some embodiments, the second training sample of the second recognition model may include the sample initial image, and the second label corresponding to the second training sample may be the number of cell clusters and the second identification frame of the cell clusters corresponding to the sample initial image. In some embodiments, the second training sample may be a matching image manually collected during the analysis of the monoclonal origin of the sample cells, and the second label may be obtained through manual annotation.

[0172] Step 520: Process at least one initial image based on a second recognition model to obtain a second recognition result.

[0173] In some embodiments, the processing device may obtain the number of cells in each well and a second identification frame of the cells in at least one well based on the output result of the second identification model.

[0174] In some embodiments of the present specification, the recognition of cells in an image can be achieved through a semantic segmentation network, thereby effectively improving the efficiency and accuracy of image recognition.

[0175] FIG. 10 is an exemplary schematic diagram of an application determination model according to some embodiments of the present specification.

[0176] In some embodiments, the processing device may process a plurality of to-be-judged images 1010 corresponding to the target well position based on the judgment model 1020 to determine whether there is a cell division anomaly 1031 and an abnormality probability 1032 at the target well position.

[0177] The judgment model may be a machine learning model. In some embodiments, the judgment model may be a machine learning model of a custom structure as described below. The judgment model may also be a machine learning model of other structures, such as a neural network model.

[0178] In some embodiments, the input of the judgment model 1020 may be a plurality of to-be-judged images 1010 corresponding to the target well position, and the output may be whether there is a cell division abnormality 1031 and an abnormality probability 1032 at the target well position.

[0179] In some embodiments, the plurality of images to be determined 1010 include a candidate image of the target hole position at a first preset time and at least one hole position image obtained by microscopic imaging of the target hole position at at least one other preset time. In some embodiments, at least one hole position image obtained by microscopic imaging of the target hole position at at least one other preset time may include a target image of the target hole position at a second preset time. In some embodiments, at least one hole position image obtained by microscopic imaging of the target hole position at at least one other preset time may include a hole position image of the target hole position between the first preset time and the second preset time.

[0180] In some embodiments, abnormal cell division may include, but is not limited to, interruption of cell division, fusion of two or more cell clusters obtained during the division of two or more cells into one cell cluster, and the like.

[0181] Abnormality probability is used to measure the possibility of abnormal cell division. The higher the abnormality probability, the higher the possibility of abnormal cell division.

[0182] In some embodiments, the judgment model 1020 may include a feature extraction layer 1021 and an abnormality judgment layer 1023 .

[0183] In some embodiments, the feature extraction layer 1021 can determine an image feature vector sequence 1022 based on a plurality of to-be-judged images 1010 of the target hole position. The image feature vector sequence 1022 output by the feature extraction layer 1021 can be used as an input of the abnormality judgment layer 1023. The image feature vector sequence is a sequence composed of image feature vectors corresponding to at least one to-be-judged image, and the sequence can be arranged according to the acquisition time of the to-be-judged image.

[0184] In some embodiments, the abnormality judgment layer 1023 can determine the abnormality judgment result of the target hole position based on the image feature vector sequence 1022. In some embodiments, the abnormality judgment result includes whether there is a cell division abnormality 1031 and a corresponding abnormality probability 1032.

[0185] In some embodiments, the feature extraction layer 1021 may be a convolutional neural network model, etc., and the abnormality judgment layer 1023 may be a long short-term memory network model, etc.

[0186] In some embodiments, the feature extraction layer 1021 and the abnormality judgment layer 1023 can be obtained by joint training of a third training sample with a third label. In some embodiments, the third training sample can be a plurality of sample images to be judged of the sample target wells, and the third label can be whether the sample target wells have cell division abnormalities and the actual abnormal features corresponding to the sample target wells.

[0187] In some embodiments, the third training sample may be obtained by directly reading from a storage device storing a large amount of historical division data. In some embodiments, the training sample may be obtained in any other manner, which is not limited in this embodiment. In some embodiments, the third label may be obtained by manual marking or other means, which is not limited in this embodiment. For example, the abnormal probability may be determined artificially based on the number of pictures showing abnormal cell division in multiple sample images to be judged. When the number of pictures showing abnormal cell division in multiple sample images to be judged of the sample target well position is greater, the higher the abnormal probability may be marked.

[0188] The exemplary joint training process includes: inputting the third training sample into the initial feature extraction layer to obtain the image feature vector sequence output by the initial feature extraction layer, inputting the image feature vector sequence into the initial abnormality judgment layer, and obtaining the presence and probability of abnormality output by the initial abnormality judgment layer. A loss function is constructed based on the output of the initial abnormality judgment layer and the third label, and the parameters of the initial feature extraction layer and the initial abnormality judgment layer are synchronously updated. Through parameter updating, the trained feature extraction layer 1021 and the abnormality judgment layer 1023 are obtained.

[0189] In some embodiments, the processing device may also analyze the monoclonal origin of the target well position based on the judgment model.

[0190] In some embodiments, the processing device may process a plurality of to-be-judged images corresponding to the target well position based on the judgment model, and analyze the monoclonal origin of the target well position.

[0191] In some embodiments, the input of the judgment model may be a plurality of images to be judged corresponding to the target well position, and the output may be whether the target well position has abnormal cell division, the abnormal probability of the target well position, and the monoclonal origin analysis result. In this embodiment, the judgment model may include a feature extraction layer, an abnormal judgment layer, and an analysis layer. In this embodiment, the input, output, and model type of the feature extraction layer and the abnormal judgment layer are the same as those in the embodiment described above, except that the output of the abnormal judgment layer (i.e., the abnormal judgment result) and the target image of the target well position are used as the input of the analysis layer, and the analysis layer directly outputs the analysis result of the monoclonal origin of the target well position. For example, the analysis layer may output the analysis result that the target well position belongs to one of the well positions to be reviewed, the monoclonal well position, and the non-monoclonal well position. Accordingly, when training the feature extraction layer, the abnormal judgment layer, and the analysis layer, the actual analysis result of the monoclonal origin of the target well position may be used as the fourth label. The feature extraction layer, the abnormal judgment layer, and the analysis layer may be obtained by joint training, and the process of joint training is similar to that in the embodiment described above. For more explanations, see above, and no further description is given here. In some embodiments, the fourth label may be obtained by the method disclosed in other embodiments of this specification, and for more explanations, see above.

[0192] In some embodiments of the present specification, the parameters of the judgment model are obtained by the above training method, which is helpful in solving the problem of difficulty in obtaining labels when training the feature extraction layer alone in some cases, and can also enable the feature extraction layer to better obtain image feature vectors reflecting candidate images. The judgment model can efficiently and accurately determine whether the target well position has abnormal cell division and judge the abnormal probability, which helps to screen out the target well position with abnormal cell division and improve the accuracy of monoclonal origin analysis.

[0193] Figure 11 is an exemplary flow chart for determining the second preset time according to some embodiments of this specification. In some embodiments, the determination method can be performed by the automatic analysis system 100 (e.g., the processing device 120). For example, the process 1100 can be stored in a storage device (e.g., the storage device 150) in the form of a program or instruction, and when the processing device 120 executes the program or instruction, the process 1100 can be implemented. The operational schematic diagram of the process 1100 presented below is illustrative. In some embodiments, the process can be completed using one or more additional operations not described and / or one or more operations not discussed. In addition, the order of the operations of the process 1100 shown in Figure 11 and described below is not restrictive.

[0194] Step 1110, obtaining at least one candidate time.

[0195] The candidate time refers to a time period corresponding to a pending option for determining the second preset time.

[0196] The candidate time can be determined in a variety of ways. In some embodiments, the candidate time can be determined based on prior knowledge or historical data. For example, the candidate time can be determined based on the prior relationship between the number of cell divisions and time. In some embodiments, the candidate time can be determined based on the cell cycle. When the cell cycle is long, the candidate time is set later. In some embodiments, the candidate time can also be determined based on the cell type. For example, the cell cycle of CHO cells is 14h, and the cell cycle of Vero cells is 18.5h. Accordingly, it takes about 4 to 5 days for CHO cells to divide and form cell clusters of appropriate size. The candidate time of CHO cells can be set to within the 4th to 5th day, and the Vero cells divide and form cell clusters of appropriate size for about 5 to 6 days. The candidate time of Vero cells can be set to within the 5th to 6th day. In some embodiments, the candidate time can also be determined by other means, and this specification does not limit this.

[0197] Step 1120, determining the degree of cell proliferation of at least one well position at at least one candidate time.

[0198] The degree of cell proliferation can be used to measure the division of cells. In some embodiments, the degree of cell proliferation can be used to measure the number and / or quantity of cell divisions.

[0199] In some embodiments, the degree of cell proliferation can be represented by a grade. A higher grade indicates a greater number and / or quantity of cell divisions.

[0200] In some embodiments, the degree of cell proliferation can be represented by a cell division index. The cell division index can represent the number and number of cell divisions. For example, the cell division index can be represented by an exponent of 2.

[0201] In some embodiments, each well position in at least one well position may determine a cell proliferation degree at each candidate time in at least one candidate time. Each well position in at least one well position may correspond to a cell proliferation degree sequence, and the cell proliferation degree sequence includes the cell proliferation degree of the well position at each candidate time in at least one candidate time.

[0202] The degree of cell proliferation can be determined in a variety of ways. In some embodiments, the processing device can estimate the degree of cell proliferation based on the cell cycle and the candidate time.

[0203] In some embodiments, the processing device may determine the extent of cell proliferation of at least one well site at at least one candidate time based on the prediction model.

[0204] The prediction model may be a machine learning model. In some embodiments, the prediction model may include any one or combination of various feasible models such as a recurrent neural network (RNN) model, a deep neural network (DNN) model, a convolutional neural network (CNN) model, etc.

[0205] In some embodiments, the input of the prediction model may include cell type, culture medium data and at least one candidate time, and the output may be the degree of cell proliferation corresponding to each of the at least one candidate time.

[0206] In some embodiments, the cell types may include but are not limited to CHO cells, Vero cells, etc. In some embodiments, the culture medium data may include but are not limited to culture medium instructions, bacteria, culture medium temperature, carbon dioxide concentration, etc.

[0207] In some embodiments, the prediction model can be obtained by training multiple fourth training samples with fifth labels. In some embodiments, the fourth training sample can include sample cell types, sample culture medium data and sample candidate time, and the fifth label can be the actual cell proliferation degree corresponding to the sample candidate time.

[0208] In some embodiments, the fourth training sample may be obtained by directly reading from a storage device storing a large number of historical splits. In some embodiments, the training sample may be obtained in any other manner, which is not limited in this embodiment.

[0209] In some embodiments, the fifth tag may be obtained by manual marking or other methods, which are not limited in this embodiment.

[0210] In some embodiments of this specification, the degree of cell proliferation is determined by a prediction model, and the degree of cell proliferation can be determined based on a large number of extensive features, breaking the limitations of the traditional rule-based determination of the degree of cell proliferation. The rule-based approach is limited by its complexity and can only be based on fewer features, and is restricted by artificially defined rules, making it difficult to get close to the actual situation. Predictions based on machine learning technology can be based on more and richer features, and the actual degree of cell proliferation can be used as a marker for training, which can make the predicted degree of cell proliferation have a higher accuracy.

[0211] Step 1130 , determining a second preset time from at least one candidate time based on the degree of cell proliferation in at least one well position.

[0212] The processing device can determine the second preset time from at least one candidate time in a variety of ways based on the degree of cell proliferation of at least one well position. In some embodiments, the processing device can determine the candidate time corresponding to the degree of cell proliferation that satisfies the preset proliferation condition as the second preset time. For example, the preset proliferation condition can be that the cell division index is a certain specified value. The preset proliferation condition can be set according to actual needs and is not limited here.

[0213] In some embodiments, the processing device can determine the cell cluster size corresponding to each candidate time in at least one candidate time based on the degree of cell proliferation; determine the candidate time when the cell cluster size meets the preset size condition, and determine the second preset time based on the candidate time.

[0214] In some embodiments, the processing device can construct a preset comparison table based on the historical division data of different cell types, the preset comparison table includes reference cell sizes of different reference cell types at different division time periods; determine the average cell size of cells at the candidate time based on the candidate time, the actual cell type and the preset comparison table; determine the cell cluster size corresponding to the candidate time based on the cell proliferation degree and the average cell size. For example, the product of the cell proliferation degree and the average cell size can be determined as the cell cluster size corresponding to the candidate time.

[0215] The preset size condition refers to a range condition related to the size of the cell cluster. In some embodiments, the preset size condition may be that the size of the cell cluster is within a specific size range. In some embodiments, the preset size condition may be determined based on the recognition effect of the first recognition model. For example, images corresponding to cell clusters of various sizes may be processed based on the first recognition model, and multiple cell cluster sizes with better model recognition effects may be selected to form a size range as the preset size condition.

[0216] In some embodiments, the recognition effect of the first recognition model on the cell cluster may be related to the first recognition frame coverage and the cell discreteness. When the first recognition frame coverage is higher and the cell discreteness is higher, it means that the recognition effect of the first recognition model on the cell cluster is better. For more information about the first recognition frame coverage and the cell discreteness, see step 320.

[0217] In some embodiments, the processing device may also determine a candidate time corresponding to a cell proliferation degree in which the cell cluster size meets a preset size condition as a target time, and determine a second preset time based on the target time and the candidate time range.

[0218] In some embodiments, the alternative time range can be a period of time before and after the target time. For example, if the target time is the sixth day, the preset time range can be within 24 hours before the target time (i.e., within the fifth day) and within 24 hours after the target time (i.e., within the seventh day). The corresponding second preset time range is from within 24 hours before the target time to within 24 hours after the target time, i.e., the fifth day to the seventh day.

[0219] In some embodiments, the processing device may also determine the imaging frequency of the microscopic imaging within the second preset time. For example, the imaging frequency of the microscopic imaging within the second preset time may be set to be higher.

[0220] In some embodiments of the present specification, the time period for cell division to form a cell cluster of appropriate size is determined as the second preset time according to the degree of cell proliferation, which can effectively avoid the influence of too small a cell cluster on the recognition effect of the first recognition model, and can effectively avoid the cost consumption caused by too large a cell cluster, and thus improve the accuracy of monoclonal origin analysis. By predicting the time that is most conducive to judging the monoclonal origin of the cell (i.e., the target time), shooting near this time can improve the accuracy of the judgment, and can reduce shooting at other times except the target time, effectively saving resources.

[0221] It should be noted that the above descriptions of processes 200, 300, 400, 500 and 1100 are only for illustration and description, and do not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to processes 200, 300, 400, 500 and 1100 under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0222] Some embodiments of the present specification provide a system for automatically analyzing the origin of a cell monoclonality, which can be deployed on a processing device (eg, processing device 120). In some embodiments, a system for automatically analyzing the origin of a cell monoclonality can include a first acquisition module, a second acquisition module, and an analysis module.

[0223] The first acquisition module can be used to perform microscopic imaging of at least one well position in the well plate within a first preset time after the plated cells have settled, and acquire at least one candidate image.

[0224] The second acquisition module can be used to perform microscopic imaging of at least one well position in the well plate within a second preset time after the plated cells have settled, and acquire at least one target image.

[0225] In some embodiments, the method for determining the second preset time includes: obtaining at least one candidate time; determining the degree of cell proliferation of at least one hole position at at least one candidate time; and determining the second preset time from at least one candidate time based on the degree of cell proliferation of at least one hole position.

[0226] The analysis module is used to obtain image recognition results and analyze the cell monoclonal origin based on the image recognition results.

[0227] In some embodiments, when performing image recognition on at least one target image, the analysis module can be used to: obtain a first recognition model; process at least one target image based on the first recognition model to obtain a first recognition result, wherein the first recognition result includes at least the number of cell clusters in each well of at least one well and at least one of the first identification frames of the cell clusters.

[0228] In some embodiments, at least one candidate image includes at least one initial image. When performing image recognition on at least one candidate image, the analysis module can be used to: obtain a second recognition model; process at least one initial image based on the second recognition model to obtain a second recognition result, and the second recognition result includes at least the number of cells in each well in at least one well and at least one of the second identification frames of the cells.

[0229] In some embodiments, the plated cells include suspended cells, and in order to analyze the monoclonal origin of the cells according to the image recognition result, the analysis module is used to analyze the monoclonal origin of the cells according to the preset rules according to the first recognition result. In some embodiments, the first recognition result includes the number of cells, and the preset rule is: comparing the number of cells and the range of cell numbers, and determining the well position corresponding to the target image whose number of cells is within the preset range of cell numbers as the monoclonal well position.

[0230] In some embodiments, the plated cells include adherent cells. In order to analyze the monoclonal origin of the cells according to the image recognition result, the analysis module is used to: based on the first recognition result, determine the well position with one cell cluster as the target well position; based on the target well position, determine a matching image in at least one candidate image, and determine a second recognition result corresponding to the matching image; based on the second recognition result and the first recognition result, analyze the monoclonal origin of the cells.

[0231] In some embodiments, to determine a matching image based on the target hole position in at least one candidate image, the analysis module is used to: determine a matching area of ​​the target hole position on at least one initial image based on the first identification frame; and obtain a matching image based on a partial image corresponding to the matching area.

[0232] In some embodiments, to determine a matching area of ​​a target hole position on at least one initial image based on a first identification frame, the analysis module is used to: determine a matching ratio of the target hole position; and determine a matching area of ​​the target hole position on at least one initial image based on the first identification frame and the matching ratio.

[0233] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0234] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0235] In addition, it will be understood by those skilled in the art that various aspects of this specification may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of this specification may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of this specification may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0236] A computer storage medium may include a propagated data signal containing computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0237] The computer program code required for the operation of each part of this specification can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, Visual Basic, Fortran2003, Perl, COBOL2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program code can be run entirely on the user's computer, or run on the user's computer as an independent software package, or run partially on the user's computer and partially on a remote computer, or run entirely on a remote computer or processing device. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0238] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

A method for analyzing cell monoclonal origin, characterized in that: The method includes: performing microscopic imaging on at least one well in a well plate after the plated cells have settled to obtain at least one image; obtaining an image recognition result based on the at least one image, and analyzing the monoclonal origin of the cells according to the image recognition result, wherein the image recognition result includes at least one of the following: a first recognition result obtained by performing image recognition on at least one target image, the at least one target image is an image obtained by performing microscopic imaging on the at least one well in the well plate within a second preset time after the plated cells have settled, the first recognition result at least includes the number of cell clusters in each well in the at least one well and at least one of the first recognition frames of the cell clusters; and a second recognition result obtained by performing image recognition on at least one initial image included in at least one candidate image, the at least one candidate image is an image obtained by performing microscopic imaging on at least one well in the well plate within a first preset time after the plated cells have settled, the second recognition result at least includes the number of cells in each well in the at least one well and at least one of the second recognition frames of the cells. The method according to claim 1, characterized in that Performing image recognition on the at least one target image includes: obtaining a first recognition model, which is a machine learning model; processing the at least one target image based on the first recognition model to obtain the first recognition result; and / or performing image recognition on the at least one candidate image includes: obtaining a second recognition model, which is a machine learning model; processing the at least one initial image based on the second recognition model to obtain the second recognition result. The method according to claim 1, characterized in that The plated cells include suspended cells, and the analyzing the monoclonal origin of the cells according to the image recognition result includes: analyzing the monoclonal origin of the cells according to preset rules according to the first recognition result; wherein the first recognition result also includes the number of cells, and the preset rules include: comparing the number of cells with a preset cell number range, and determining the well position corresponding to the target image whose cell number is within the preset cell number range as the monoclonal well position. The method according to claim 1, characterized in that The plated cells include suspended cells, and the analyzing of the monoclonal origin of the cells according to the image recognition result includes: analyzing the monoclonal origin of the cells according to preset rules according to the first recognition result; wherein the first recognition result also includes the number of cells, the candidate image includes at least one intermediate image, and the preset rules include: comparing the number of cells with a preset cell number range, and determining the well position corresponding to the target image whose cell number is within the preset cell number range as the well position to be reviewed; based on the second recognition result of the intermediate image of the well position to be reviewed, judging whether the well position to be reviewed is a monoclonal well position. The method according to claim 1, characterized in that The plated cells include adherent cells, and the analyzing the monoclonal origin of the cells according to the image recognition result includes: based on the first recognition result, determining the well position with one cell cluster as the target well position; based on the target well position, determining a matching image in the at least one candidate image, and determining a second recognition result corresponding to the matching image; based on the second recognition result and the first recognition result, analyzing the monoclonal origin of the cells. The method according to claim 5, characterized in that Determining a matching image in the at least one candidate image based on the target hole position includes: determining a matching area of ​​the target hole position on the at least one initial image based on the first recognition frame; and obtaining the matching image based on a partial image corresponding to the matching area. The method according to claim 6, characterized in that The determining, based on the first recognition frame, a matching area of ​​the target hole position on the at least one initial image comprises: performing template matching based on the first recognition frame to determine a matching area of ​​the target hole position on the at least one initial image, wherein the template matching comprises: determining a first hole image and a second hole image from the at least one initial image and the at least one target image based on an arc edge feature of the target hole position; and performing template matching on the first hole image and the second hole image based on the first recognition frame to determine a matching area of ​​the target hole position on the first hole image. The method according to claim 5, characterized in that Determining a matching image in the at least one candidate image based on the target hole position includes: processing a plurality of images to be judged corresponding to the target hole position based on a judgment model to determine whether the target hole position has a cell division abnormality and its abnormality probability, wherein the plurality of images to be judged include the candidate image of the target hole position at the first preset time and at least one hole position image obtained by microscopic imaging of the target hole position at at least one other preset time, and the judgment model is a machine learning model; in response to the presence of the cell division abnormality, screening out the target hole position; in response to the absence of the cell division abnormality, determining the matching image in the at least one candidate image based on the target hole position. The method according to claim 8, characterized in that The judgment model includes a feature extraction layer and an abnormality judgment layer; the feature extraction layer is used to determine an image feature vector sequence based on the multiple images to be judged corresponding to the target hole position; the abnormality judgment layer is used to determine whether the target hole position has the cell division abnormality and the abnormality probability based on the image feature vector sequence. The method according to claim 8, characterized in that The judgment model is also used to determine the monoclonal origin analysis result of the target well position. The judgment model includes a feature extraction layer, an abnormality judgment layer and an analysis layer. The feature extraction layer is used to determine the image feature vector sequence based on the multiple images to be judged corresponding to the target well position; the abnormality judgment layer is used to determine the abnormality judgment result of the target well position based on the image feature vector sequence, and the abnormality judgment result includes whether the cell division abnormality exists and the abnormality probability; the analysis layer is used to determine the monoclonal origin analysis result of the target well position based on the abnormality judgment result and the at least one target image of the target well position. The method according to claim 5, characterized in that The analyzing the cell monoclonality based on the second recognition result and the first recognition result includes: determining a target well with zero cells as a well to be verified; determining a target well with one cell as a monoclonal well, wherein the cells in the monoclonal well are of monoclonal origin; and determining a target well with more than one cell as a non-monoclonal well, wherein the cells in the non-monoclonal well are not of monoclonal origin. The method according to claim 6, characterized in that Determining a matching area of ​​the target hole position on the at least one initial image based on the first recognition frame includes: determining a matching ratio of the target hole position; and determining the matching area of ​​the target hole position on the at least one initial image based on the first recognition frame and the matching ratio. The method according to claim 12, characterized in that Determining the matching ratio of the target well position includes: determining the matching ratio based on the first recognition frame coverage rate and cell discreteness of the target well position. The method according to claim 1, characterized in that The method for determining the second preset time includes: obtaining at least one candidate time; determining the cell proliferation degree of the at least one hole position at the at least one candidate time; and determining the second preset time from the at least one candidate time based on the cell proliferation degree of the at least one hole position. The method according to claim 14, characterized in that The determining the second preset time from the at least one candidate time based on the cell proliferation degree of the at least one hole position comprises: determining the cell cluster size corresponding to each candidate time in the at least one candidate time based on the cell proliferation degree; determining the candidate time at which the cell cluster size meets a preset size condition, and determining the second preset time based on the candidate time. The method according to claim 15, characterized in that The determining the candidate time when the size of the cell cluster meets the preset size condition, and determining the second preset time based on the candidate time, further includes: determining the candidate time when the size of the cell cluster meets the preset size condition as the target time; and determining the second preset time based on the target time and the alternative time range. The method according to claim 1, characterized in that The method further includes: marking the first recognition frame and the second recognition frame in the initial image and / or the target image at the same time. The method according to claim 17, characterized in that The method further includes: enlarging a partial image of the initial image within the second recognition frame and displaying it in the target image. A cell monoclonal origin analysis system, comprising: At least one computer-readable storage medium comprising a set of instructions, and at least one processing device in communication with the computer-readable storage medium, wherein when executing the set of instructions, the at least one processor is configured to: obtain at least one image by performing microscopic imaging of at least one well position in a well plate after the plated cells have settled; obtain an image recognition result based on the at least one image, and analyze the monoclonal origin of the cells according to the image recognition result, wherein the image recognition result includes at least one of the following: a first recognition result obtained by performing image recognition on at least one target image, the at least one target image being a first recognition result obtained by performing image recognition on at least one target image after the plated cells have settled. an image acquired by microscopic imaging of at least one well in the well plate within a second preset time, the first recognition result comprising at least the number of cell clusters in each well of the at least one well and at least one of the first recognition frames of the cell clusters; or a second recognition result obtained by image recognition of at least one initial image included in at least one candidate image, the at least one candidate image being an image acquired by microscopic imaging of at least one well in the well plate within a first preset time after the plated cells have settled, the second recognition result comprising at least the number of cells in each well of the at least one well and at least one of the second recognition frames of the cells. A computer-readable storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes a cell monoclonal origin analysis method, the method comprising: Acquire at least one image by microscopically imaging at least one well in the well plate after the plated cells have settled; An image recognition result is obtained based on the at least one image, and the monoclonal origin of the cells is analyzed according to the image recognition result, wherein the image recognition result includes at least one of the following: a first recognition result obtained by performing image recognition on at least one target image, wherein the at least one target image is an image obtained by microscopic imaging of at least one well in the well plate within a second preset time after the plated cells have settled, and the first recognition result includes at least the number of cell clusters in each well of the at least one well and at least one of the first identification frames of the cell clusters; or a second recognition result obtained by performing image recognition on at least one initial image included in at least one candidate image, wherein the at least one candidate image is an image obtained by microscopic imaging of at least one well in the well plate within a first preset time after the plated cells have settled, and the second recognition result includes at least the number of cells in each well of the at least one well and at least one of the second identification frames of the cells.

Images