JP6721895B2 - Cell analysis result output device, cell analysis result output method and program - Google Patents

Description

The present invention relates to a cell analysis result output device, a cell analysis result output method, and a program.

BACKGROUND ART Conventionally, cell staining technology has been widely used in fields such as detection of spontaneous cancer in cancer research, detection of xenogeneic cells other than therapeutic cells in regenerative medicine, and clinical research of mesenchymal stem cells (MSCs). Came.

When such a technique for staining cells is applied, the cells are destroyed by the staining. Furthermore, staining reagents are expensive and the staining of cells themselves is tedious.
Therefore, the inventors of the present invention have continued research and development on a technique for non-destructive analysis of cells so that it can be inexpensively and simply realized (see Patent Documents 1 to 5).

JP, 2011-232051, A JP, 2011-229413, A JP, 2011-229410, A JP, 2011-229409, A Re-table 2010/098105

However, even if it is said as a technique for analyzing cells, the analysis contents are various, and even if the same analysis contents are present, there are various kinds of analysis methods and conditions. Therefore, it is necessary to generate different types of analysis models for each analysis content, analysis method, and various conditions.
There has been a demand for the user to easily and effectively recognize the result of such an analysis process regarding various cells having different analysis contents, analysis methods, various conditions, and the like.

The present invention has been made in view of such a situation, and allows a user to easily and effectively recognize the result of an analysis process relating to various cells having different analysis contents, analysis methods, various conditions, and the like. With the goal.

In order to achieve the above object, the cell analysis result output device of one embodiment of the present invention is
Based on the data of one or more cell images included in the unit, with one or more cell images selected under a predetermined condition as a unit from among the cell images obtained by capturing a cell population including a plurality of cells as a subject. , Information including the values of N kinds (N is an integer value of 1 or more) of characteristic parameters relating to the morphological characteristics of the cells alone or the morphological characteristics of the cell population is held as characteristic group information,
Based on at least one of the N types of feature parameters included in the feature group information to be compared, the comparison target information including M types (M is an integer value independent of N) of parameter values is Acquired,
Comparison target information composed of values of the M types of parameters is acquired based on at least one of the N types of feature parameters included in the feature group information of the comparison target,
When the distance from the predetermined viewpoint between the comparison target information and the comparison partner information is calculated,
The image information including the first symbol indicating the comparison target information and the second symbol indicating the comparison target information, the display form of which changes according to the calculated distance, is output as output information. Output means to
Equipped with.

A cell analysis result output method and program according to an aspect of the present invention are a method and a program corresponding to the cell analysis result output device according to an aspect of the present invention described above.

According to the present invention, it is possible to make the user easily and effectively recognize the result of the analysis processing on various cells having different analysis contents, analysis methods, various conditions, and the like.

It is a block diagram which shows the structure of the hardware of the cell analyzer which concerns on one Embodiment of this invention. It is a functional block diagram which shows the functional structure for performing a cell analysis process among the functional structures of the cell analyzer 1 of FIG. FIG. 3 is a functional block diagram showing details of the functional configuration of a feature group information generation unit in the functional configuration of the cell analysis device 1 of FIG. 2. 4 is a schematic diagram illustrating a process of an image digitizing unit of the feature group information generating unit of FIG. 3. FIG. It is a figure explaining ten concrete types of a characteristic parameter. It is a figure which shows an example of the cell image used as the basis of the production|generation of characteristic group information. It is a figure which shows a part of feature group information concretely. It is a figure which shows a part of feature group information concretely. It is a schematic diagram for demonstrating the difference of the element extracted for analysis evaluation as feature group information. It is a figure which shows typically the feature group information produced|generated from a cell image based on the emphasis and combination of the effective thing in the cell population. The various specific examples of the pattern of information emphasis in distribution are shown. Various specific examples of the method of processing the collective data in consideration of the change over time are shown. It is a figure which shows various specific examples of the processing method of group information in consideration of the point of adding heterogeneous information. It is a figure which shows an example of the feature group information processed by adding heterogeneous information and time-dependent consideration. It is a figure which shows the difference between the conventional sample data and the sample data produced|generated from the feature group information of this embodiment. It is a figure showing an example of sample data used for generation of an analytical model for predicting a change in cell quality. It is a figure which shows the prediction precision at the time of predicting a cell quality change using the sample data of FIG. It is a figure which shows the result of having actually predicted the cell quality change by the analytical model produced|generated using the sample data of the standardization type E of FIG. It is a figure showing an example of a plurality of sample data used for generation of a predetermined analysis model. Two examples of methods of generating a predetermined analysis model are shown using the sample data of FIG. An example of the expression form of the analysis model different from the examples described in FIGS. 17 to 19 is shown. FIG. 22 is a diagram showing an example of the expression form of the analysis model, which is different from FIG. 21. FIG. 23 is a diagram showing an example of the expression form of the analysis model, which is different from the examples shown in FIGS. 21 and 22. FIG. 24 is a diagram showing an example of the expression form of the analysis model, which is different from the examples shown in FIGS. 21 to 23. It is a figure which shows the structural example of the feature group information used as the basis of generation of sample data and evaluation object information. It is a figure which shows the structural example of the feature group information of FIG. 25 which considered the time change. FIG. 27 is a diagram illustrating an example of a method of comparing sample data and evaluation target information using the feature group information having the configuration of FIG. 26. It is a figure which shows an example of the screen which shows output information. It is a figure which shows an example of the screen which shows output information.

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a hardware configuration of a cell analyzer 1 according to an embodiment of the present invention.
The cell analyzer 1 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a bus 14, an input/output interface 15, an output unit 16, and an input unit. 17, a storage unit 18, a communication unit 19, and a drive 20.

The CPU 11 executes various processes according to a program recorded in the ROM 12 or a program loaded from the storage unit 18 into the RAM 13.
The RAM 13 also appropriately stores data and the like necessary for the CPU 11 to execute various processes.

The CPU 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input/output interface 15 is also connected to the bus 14. An output unit 16, an input unit 17, a storage unit 18, a communication unit 19, and a drive 20 are connected to the input/output interface 15.

The output unit 16 includes a display, a speaker, and the like, and outputs various kinds of information as images and sounds.
The input unit 17 is composed of a keyboard, a mouse and the like, and inputs various information.
The storage unit 18 includes a hard disk, a DRAM (Dynamic Random Access Memory), or the like, and stores various data.
The communication unit 19 controls communication performed with other devices (an external device 81 and a preprocessing device 82 in FIG. 2, which will be described later in this embodiment) via a network (not shown).

A removable medium 31 is properly attached to the drive 20 as needed. The program read from the removable medium 31 by the drive 20 is installed in the storage unit 18 as needed. Further, the removable medium 31 can also store various data stored in the storage unit 18 similarly to the storage unit 18.

FIG. 2 is a functional block diagram showing a functional configuration for executing a cell analysis process among the functional configurations of the cell analysis device 1 of FIG.
The cell analysis process is a result of performing a predetermined analysis on the cell based on data of an image including one or more cells as a subject (hereinafter referred to as “cell image”) and a predetermined analysis model, and the analysis result. Is a series of processes until output.
Here, the predetermined analysis regarding cells means that analysis based on a cell image in which the cells are the subject is sufficient. That is, the predetermined analysis regarding a cell is a broad concept including not only analyzing the type and state of the cell but also analyzing the outside (surrounding environment) of the cell.

When the cell analysis processing is executed, mainly in the CPU 11, as shown in FIG. 2, a cell image input unit 41, a feature group information generation unit 42, a noise removal unit 43, and an analysis model generation unit 44. The evaluation target information acquisition unit 45, the evaluation target information acquisition unit 46, the analysis unit 47, and the output information generation unit 48 function.
Further, a feature group information DB 61, a cell evaluation DB 62, and an analysis model holding unit 63 are provided in one area of the storage unit 18.

The cell image input unit 41 inputs cell image data. Here, the source of the cell image is not particularly limited, and may be an external device 81 connected via the Internet or the like (not shown), or a pre-processing device 82 held by the administrator of the cell analysis device 1 or the like. May be
That is, if a person who provides a cell image is hereinafter referred to as a “user”, the user operates the external device 81 to carry out the cell image via the Internet or the like when the user does not have the cell analyzer 1. The data of 1 is provided to the cell analyzer 1. On the other hand, when the user holds the cell analysis device 1, the user also holds the pretreatment device 82 and provides the cell analysis device 1 with the data of the cell image generated by the pretreatment device 82.
The pre-processing device 82 includes a cell adjustment unit 71 that adjusts cells, and a cell imaging unit 72 that images the cells and outputs them as cell image data.

The feature group information generation unit 42 binarizes the data of the cell image, extracts the cell object from the resulting data (hereinafter referred to as “binarized data”), and forms one or more types of the cell object. Data composed of respective values of parameters indicating respective physical characteristics, that is, digitized data is generated.

In addition, hereinafter, a parameter indicating a predetermined morphological feature is particularly referred to as a “feature parameter”. The value of the characteristic parameter is also called “morphological characteristic amount”.

Here, one cell image is an image of a cell population composed of a plurality of cells, for example, a cell population cultured in the same container (the number of cells may change with time) as an object. In many cases. In this case, one cell image includes a plurality of cell objects, and each value of one or more types of characteristic parameters is calculated for each of the plurality of cell objects.

In the case of a cell image with such a cell population as a subject, the value of the feature parameter of the cell alone (morphological feature amount) is also significant information, but the information regarding the feature parameter of the cell population (hereinafter referred to as “population information”). Calling) is also useful information.
The population information includes, for example, a histogram of predetermined characteristic parameters in the cell population, and a statistical value (a numerical value such as an average value or a variance value) of the predetermined characteristic parameters in the cell population.

In the present embodiment, the value of the characteristic parameter of the single cell (morphological characteristic amount of the single cell) and the population information are collectively referred to as “feature group information”.

Here, the feature group information does not always include only the numerical values obtained from one cell image.
For example, a cell population in a predetermined container may be a subject, and cell images may be captured multiple times over time. In such a case, the feature group information may be generated based on each data of a plurality of cell images having different imaging times.
Further, for example, in the case where the feature group information is generated based on each data of a plurality of cell images separately photographed in different containers for the same type of cells separately cultured in different containers (well etc.) There is also.
In this way, the feature group information can be generated based on each data of the arbitrary number of cell images.

The noise removing unit 43 removes noise data as sample data used to generate the analysis model.
The data that becomes noise is, for example, data of a cell image when a procedure is bad during cell culture, that is, sample data extracted from feature group information and the like generated based on the data of the cell image. is there.

The analysis model generation unit 44 generates sample data based on at least a part of the feature group information stored in the feature group information DB 61.
Here, "to generate sample data based on at least a part of the feature group information" means not only to extract some of the constituent elements (values of feature parameters) of the feature group information but also to select an arbitrary number of pieces. It also means to generate new data using any kind of components of.

The analysis model generation unit 44 uses one or more sample data to generate a model (hereinafter, referred to as “analysis model”) for performing a predetermined analysis on the cells in the cell image.

Here, the content of the analysis is wide-ranging, even if it is referred to as a predetermined analysis regarding cells. For example, just to give a representative analysis content, there are three types: a type that classifies cell types, a type that determines the state of cultured cells, and a type that analyzes the external environment in which cells are cultured. Exists.
Therefore, there may be multiple types of analysis models for each type of analysis content. That is, in each of the analytical model for analyzing the cell type, the analytical model for judging the state of the cultured cells, and the analytical model for analyzing the external environment in which the cells are cultured, Often different types are used.
Even if the analysis contents are the same, there are various analysis methods and various conditions. Therefore, different types of analysis models are often used depending on the analysis method and various conditions.

If the type of analysis model is different, the type and structure of the sample data used to generate it are often different.
That is, the values of all types of feature parameters included in the feature group information are not always used as they are as sample data.
That is, a combination in which the type and number of characteristic parameters differ depending on the analysis content, analysis method, various conditions, etc. may be extracted, and sample data may be generated based on each extracted combination.

Specifically, for example, the average value of the size in the cell population (first characteristic parameter), the variance value of the size in the cell population (second characteristic parameter), and the average value of the length in the cell population (third characteristic) It is assumed that there is feature group information including a parameter) and a variance value of the length in the cell population (fourth feature parameter).
In this case, sample data may be generated from the respective values of the first characteristic parameter and the third characteristic parameter in order to generate an analytical model for classifying cell types.
On the other hand, sample data may be generated from the respective values of the second characteristic parameter, the third characteristic parameter, and the fourth characteristic parameter in order to generate an analytical model for determining the state of the cells being cultured. ..

Here, the values of the characteristic parameters of the characteristic group information may be used as they are as the constituent elements of the sample data. However, as will be described later with reference to FIGS. 25 to 27, a parameter value different from the characteristic parameter of the characteristic group information may be newly generated as a constituent element of the sample data.
That is, a sample composed of each value of M kinds of parameters (M is an integer value of 1 or more independent of N) based on N kinds (N is an integer value of 1 or more) of characteristic parameters of the feature group information. Data is generated.
The M types of parameters are concepts independent of the N types of characteristic parameters. That is, in many cases, the type and number of M types of parameters forming the sample data differ depending on the analysis content, analysis method, various conditions, and the like.

Furthermore, the feature group information on which the sample data is based is not always generated from one cell image. That is, the feature group information is generated from a unit of one or more cell images selected under a predetermined condition.
Here, the predetermined condition is not particularly limited, and for example, a condition of “each cell image in the same container at times t1 to t5” may be adopted. When this condition is adopted, one sample data is generated based on the feature group information obtained from the five cell images at the times t1 to t5 in the same container.

Therefore, the sample data may include data (numerical values) at different times.
Specifically, for example, it is assumed that there is feature group information including the first to fourth feature parameters in the above example.
The value of the third characteristic parameter is the parameter of the sample data, both for generating an analytical model for classifying cell types and for generating an analytical model for determining the state of cultured cells. Shall be adopted as the value of.
Even in this case, for generation of the analysis model for classifying the cell types, the analysis of determining the state of the cultured cells while each value of the time t1 and the time t2 is included in the sample data. For model generation, the value at time t4 may be included in the sample data.
That is, even if the characteristic parameters are the same, different values at different times are independent data having different meanings. For this reason, sample data may be generated in consideration of the difference in time change for each analysis content, analysis method, various conditions, and the like.

In summary, the feature group information is N-dimensional or more multidimensional information including at least N types of feature parameters as elements.
Data composed of M types of parameter values generated based on at least one of N types of feature parameters included in such feature group information is sample data.
That is, the M kinds of parameters of the sample data are independent of the N kinds of characteristic parameters, and of course, some kinds of the characteristic parameters of the N kinds of characteristic parameters can be included, but other parameters can also be included. May be included.
Here, other parameters than the characteristic parameters will be referred to as “incidental parameters” below. As the additional parameter, for example, a time parameter can be adopted. Further, for example, the parameters defined by the external environment, various conditions, etc. are also the auxiliary parameters. Further, for example, in the sample data of the examples of FIGS. 25 to 27 described later, the parameters indicating the respective axes forming the predetermined three-dimensional space are the constituent elements. The parameter indicating each axis is also a kind of additional parameter.

As described above, in the present embodiment, when the analysis model is generated, the feature group information itself is not used as it is as the sample data, but the sample data having M kinds of parameters independent of the feature group information as elements is used.
As the M types of parameters, it is possible to employ an arbitrary number of combinations of arbitrary types according to the analysis content, analysis method, various conditions, and the like. That is, it is possible to individually generate sample data of different types and different structures suitable for the analysis content, the analysis method, the predetermined condition, etc. from one feature group information. As a result, it becomes possible to generate an analysis model suitable for the analysis content, the analysis method, the predetermined condition, and the like. That is, it is possible to efficiently generate sample data (one type of cell analysis information) that is suitable for each analysis model.
As a result, it becomes possible to easily and appropriately perform comprehensive analysis on cells with respect to various analysis contents, analysis methods, predetermined conditions, and the like.

That is, the analysis model generation unit 44 in FIG. 2 searches the feature group information DB 61 for the feature group information regarding the cell population of the sample for each analysis content, analysis method, various conditions, etc., and N included in the feature group information. Based on at least one kind of the characteristic parameters of the kinds, sample data having each value of the M kinds of parameters as elements is generated.
Then, the analysis model generation unit 44 uses one or more sample data having different types and structures for each analysis content, analysis method, various conditions, and the like to generate a plurality of analysis models for each analysis content, analysis method, various conditions, etc. Generate each.
Each of these analysis models is held in the analysis model holding unit 63 in association with information indicating the type and structure of each sample data.

It should be noted that the various analysis models held in the analysis model holding unit 63 are appropriately updated by the analysis model generation unit 44 at an appropriate timing after the feature group information about the new cell image is stored in the feature group information DB 61. To be done.

The evaluation target information acquisition unit 45 acquires the evaluation target information corresponding to the cell image from the external device 81 or the cell adjustment unit 71 of the preprocessing device 82, and stores it in the cell evaluation DB 62.
The evaluation target information corresponding to the cell image indicates a result of the destructive cell evaluation performed by using a predetermined evaluation method on a cell population (for example, a cell population in a predetermined container) that is a subject of the cell image. It is information. That is, the information analyzed and evaluated by a means different from the cell image for the predetermined cell population is the evaluation target information.
The destructive cell evaluation is not particularly limited, and any one such as genome, gene expression, protein, metabolite, mutual spoofing, living transplantation result, therapeutic result, etc. can be adopted.
The evaluation target information is preferably stored in the cell evaluation DB 62 as digitized information. This is because the value of the evaluation target information can be included as the element value (value of the predetermined parameter) of the sample data (multidimensional information) when generating the predetermined analysis model.

The evaluation target information acquisition unit 46 acquires, from the feature group information DB 61, the feature group information of the cells to be analyzed and evaluated in a form suitable for the analysis evaluation analysis model. Note that the information acquired by the evaluation target information acquisition unit 46 in this manner is hereinafter referred to as “evaluation target information”.

The form suitable for the analytical model for analytical evaluation means the same form as the form of the sample data used when the analytical model is generated (that is, the form in which the M types of parameters are the same type). ..
That is, as described above, the form of the sample data used at the time of generation may differ for each type of analysis model.
Therefore, the evaluation target information also needs to be in the same form as the sample data used when the analysis model to be used is generated.

That is, the evaluation target information acquisition unit 46 searches the feature group information DB 61 for the feature group information of the cell to be analyzed and evaluated, and based on at least one of the N types of feature parameters included in the feature group information. , M types of parameter values are acquired as evaluation target information.
In this case, "at least one of N types of feature parameters included in the feature group information" and "M types of parameters" are the same types as those used when the sample data of the analysis model to be used is generated. Is.

The analysis unit 47 executes an analysis process on cells to be analyzed and evaluated based on the evaluation target information and the analysis model. As described above, the analysis content, analysis method, various conditions, etc. are not particularly limited, and various kinds of analysis can be performed.

The output information generation unit 48 generates, as output information, information including cells related to analysis and evaluation (e.g., evaluation target information) and information including analysis processing results for the cells. Then, the output information generation unit 48 outputs the output information to the output unit 16 or the external device 81.
A specific example of the output information will be described later with reference to FIGS. 28 and 29.

The functional configuration example of the cell analysis device 1 has been described above with reference to FIG.
Next, the detailed configuration of the feature group information generation unit 42 in the cell analyzer 1 having the functional configuration of FIG. 2 will be described with reference to FIG.

FIG. 3 is a functional block diagram showing details of the functional configuration of the feature group information generation unit 42 in the functional configuration of the cell analysis device 1 of FIG.

As shown in FIG. 3, in the feature group information generation unit 42, the image digitization unit 91, the plural-image feature group information generation unit 92, the heterogeneity information addition unit 93, the temporal change information addition unit 94, and the feature group. The information output unit 95 functions in the CPU 11 (FIG. 1).
Further, in one area of the storage unit 18 (FIG. 1), the digitized information storage unit 101, the first feature group information storage unit 102, the second feature group information storage unit 103, and the third feature group information storage unit And 104 are provided.

The image digitizing unit 91 generates feature group information for each processing unit by performing a predetermined image processing on the data of the processing unit with the data of one cell image as the processing unit.
Feature group information generated from a single cell image, that is, information consisting of an aggregate of values (numerical values) of various feature parameters of each cell object included in the cell image is accumulated in the digitized information accumulating unit 101. To be done. That is, the digitized information storage unit 101 stores a plurality of units of feature group information, with one cell image as a unit.

FIG. 4 is a schematic diagram illustrating the processing of the image digitizing unit 91.
In the example of FIG. 4, the data of one cell image G1 is given to the image digitization unit 91 as a processing unit.
The image digitization unit 91 generates binarized data G2 from the data of the cell image G1. The image digitization unit 91 extracts "XXX" (XXX is an arbitrary integer value) cell object candidates from the binarized data. Here, the reason why the object is not a cell object but a candidate is that an object that becomes noise other than a cell (hereinafter, referred to as “noise object”) is included.

The image digitization unit 91 attaches a unique ID (Cell ID) to each of the cell object candidates and obtains various characteristic parameters. That is, each cell object is digitized. In the example of FIG. 4, the values of nine types of characteristic parameters (Parameters 1 to 9) are obtained for each cell object.
A list of nine types of feature parameter values (numerical data) for each cell object is the feature group information I1 of the cell image G1.

However, the feature group information I1 of the cell image G1 also includes numerical data (values of nine types of feature parameters) regarding the noise object.
Therefore, in step S1, the image digitization unit 91 determines a noise object from the cell object candidates according to an appropriate algorithm, and excludes the digitized data of the noise object.
In the example of FIG. 4, the cell object candidates whose IDs are “Cell 001” and “Cell 002” are determined to be noise objects. Therefore, the digitized data regarding the noise object is excluded from the feature group information I1 of the cell image G1, and the feature group information I2 of the cell image G1 is generated.

Further, in step S2, the image digitization unit 91 determines a noise object from the cell object candidates according to an appropriate algorithm different from step S1 and excludes the digitized data of the noise object.
In the example of FIG. 4, the cell object candidates whose IDs are “Cell 003” and “Cell 004” are determined to be noise objects. Therefore, the digitized data on the noise object is excluded from the feature group information I1 of the cell image G1, and the feature group information I3 of the cell image G1 is generated.

In step S3, the image digitization unit 91 determines the cell object candidate from which the noise object is excluded as a cell object, and stores the feature group information I3 of the cell image G1 in the digitized information storage unit 101.

In the example of FIG. 4, the number of types of characteristic parameters is nine, but the number of types is not particularly limited to this.
For example, FIG. 5 is a diagram illustrating ten specific types of characteristic parameters.
In FIG. 5, one line indicates a predetermined type of characteristic parameter. The item in the first column shows the parameter number. The item in the second column shows the name of the characteristic parameter. The item in the third column shows an example (method of obtaining) of the characteristic parameter. The item in the fourth column shows the description of the characteristic parameter.

“Total area” is a characteristic parameter indicating the area of a cell (more precisely, “cell object”, but simply referred to as “cell” in this paragraph).
“Area” is a characteristic parameter indicating the area excluding the intracellular holes. Here, the hole is a portion where the brightness of the cell image is equal to or higher than a threshold value due to the contrast (a portion that becomes a state close to white in phase difference observation).
“Perimeter” is a characteristic parameter indicating the length of the outer circumference of the cell.
“Length” is a characteristic parameter that indicates the maximum value (the total length of the cell) of the line that crosses the cell.
"Breadth" is a characteristic parameter indicating the maximum value (cell width) of the line orthogonal to "Length".
"Inner radius" is a characteristic parameter indicating the radius of the inner circumference of the cell.
“Elliptical form factor” is a characteristic parameter obtained by dividing the value of “Length” by “Breadth”.
“Fiber Width” is a characteristic parameter indicating the width (length in the direction orthogonal to Fiber Length) when the cells are assumed to be pseudo linear.
“Fiber Length” is a characteristic parameter indicating the length of a cell assuming that the cell is pseudo linear.
“Shape Factor” is a characteristic parameter indicating the circularity of cells (roundness of cells).

Note that the feature parameters in the example of FIG. 5 are merely examples, and any type can be adopted as long as it is a parameter indicating a predetermined morphological feature that can be quantified from a cell object.
Further, since one cell image includes a cell population, although not shown, population information is also obtained and included in the feature group information. For example, the average value or the variance value of the above-mentioned “Total area” may be included in the feature group information.
In addition, a histogram of a predetermined morphological feature amount of a plurality of cells can be obtained, and the value of each bin in the histogram can be included in the feature group information.

Returning to FIG. 3, the multi-image feature group information generation unit 92 determines a first processing unit including a plurality of cell images according to a predetermined rule, and digitizes the feature group information included in the first processing unit as digitized information. It is acquired from the storage unit 101.
The predetermined rule is not particularly limited, but here, a rule that a plurality of cell images imaged under the same condition and at the same time become the first processing unit is adopted.
Next, the multi-image feature group information generation unit 92 combines the feature group information acquired from the digitized information storage unit 101 with each other and combines the digitized data to arrange the feature group information of the first processing unit. To generate.
The feature group information of the first processing unit is stored in the first feature group information storage unit 102.

The heterogeneous information adding unit 93 determines a second processing unit including a plurality of cell images according to a predetermined rule, and outputs each feature group information included in the second processing unit from the first feature group information storage unit 102. get.
The predetermined rule is not particularly limited, but here, the rule that a plurality of cell images imaged at a plurality of conditions and at the same time become the second processing unit is adopted. Next, the heterogeneous information adding unit 93 For each of the acquired feature group information, the digitized data is combined or organized to generate the feature group information of the second processing unit to which the heterogeneity information is added. Specific examples of the heterogeneity information and the feature group information of the second processing unit will be described later with reference to FIG.
The feature group information of the second processing unit is stored in the second feature group information storage unit 103.

The temporal change information adding unit 94 determines a third processing unit including a plurality of cell images according to a predetermined rule, and outputs each feature group information included in the third processing unit from the second feature group information storage unit 103. get.
The predetermined rule is not particularly limited, but here, a rule is adopted in which a plurality of cell images captured under a plurality of conditions and a plurality of times (time varying at different times) serve as a third processing unit. The temporal change information adding unit 94 generates the characteristic group information of the third processing unit to which the concept of the temporal change is added, by combining or organizing the respective numerical data with respect to the acquired respective characteristic group information. Note that a specific example of the feature group information of the third processing unit to which the concept of change over time is added will be described later with reference to FIGS. 9 to 12.
The feature group information of the third processing unit is stored in the third feature group information storage unit 104.

The feature group information output unit 95 acquires the feature group information from the third feature group information storage unit 104 at an appropriate timing, removes noise, adds tag information, and the like, and then stores the feature group information DB 61. ..

As described above in detail with reference to FIG. 3, the feature group information generation unit 42 performs the image processing on the data of the single cell image in the image digitization unit 91, thereby digitizing the features. It becomes group information.
Therefore, the processing targets of the plural-image feature group information generation unit 92 to the feature group information output unit 95 are not the image data (image data) but the feature group information that is digitized data. Therefore, the plural-image feature group information generation unit 92 to the feature group information output unit 95 are collectively referred to as a “digitized information processing unit 111” as appropriate.

Here, a specific example of the feature group information stored in the feature group information DB 61 will be described with reference to FIGS. 6 to 8.
FIG. 6 shows an example of a cell image which is a basis for generating the feature group information.
The cell image CA is an image of a cell population of the type “CA” cultured in the same container.
The cell image CB is an image of a cell population of the type “CB” cultured in the same container.
The cell image CC is an image of a cell population of the type “CC” cultured in the same container.
The cell image CD is an image of a cell population of the type “CD” cultured in the same container.
Note that in the example of FIG. 6, only one cell image is shown for each type “CA” to “CD”, but in reality, for each type, at different timings from time t1 to t10. It is assumed that there are 10 cell images captured respectively.
The “time” mentioned here does not mean an absolute time (world standard time) but means a relative time from a predetermined reference time (culture start time), such as one hour after the start of culture.

7 and 8 specifically show a part of the feature group information.
The unit of the characteristic group information is not particularly limited as described above, but in the example of FIGS. 7 and 8, the cell type is 1 unit, and the characteristic group information of 4 units is arranged in each column.
That is, a part of the feature group information of the cells of the type “CA” is arranged in the first column from the left (excluding the item name). Part of the characteristic group information of cells of the type “CB” is arranged in the second column from the left. Part of the characteristic group information of cells of the type “CC” is arranged in the third column from the left. Part of the characteristic group information of cells of the type “CD” is arranged in the fourth column from the left.

In the example of FIG. 7, a part of the feature group information, which has the statistical value of the predetermined feature parameter in the cell population of each cell image as an element, is shown. The leftmost column in FIG. 7 contains element names of the feature group information.
Here, “pas” indicates (relative) time. That is, kpas (k is an arbitrary integer value from 1 to 10) is a value obtained from the cell image imaged at the time k.
“AVE” indicates the average value, and “SD” indicates the standard deviation.
For example, the element (Normal_Area_AVE_01pas) in the first row from the top indicates the average value of “Area” of the normal cell population included in the cell image at time t1.
Further, for example, the element (Normal_Area_SD_02pas) in the twelfth row from the top indicates the standard deviation of “Area” of the normal cell population included in the cell image at time t2.

In the example of FIG. 8, a part of the feature group information having the bin of the histogram for the predetermined feature parameter in the cell population of each cell image as an element is shown. The leftmost column in FIG. 8 includes element names of the feature group information.
Here, "bin" indicates a bin number. The numbering method is not particularly limited, but in the example of FIG. 8, the histogram is composed of 12 divisions (12 bins), and the numbers are assigned in order from the left.
For example, the element in the first row from the top (Area_1pas_1bin) indicates the value of the first bin from the left in the histogram of "Area" of the cell population included in the cell image at time t1.

Here, in FIGS. 7 and 8, each element (each feature parameter) of the feature group information is given a different color and shade according to its numerical value.
The color and shade pattern of a certain region may be different or characteristic for each cell type “CA” to “CD”.
Here, "a certain area" may be a continuous area formed of a plurality of continuous items in the column direction, or an area formed of a plurality of non-continuous items. is there. For example, the continuous area (the area of Normal_Area_AVE) of the first to tenth rows from the top of FIG. 7 may be the “certain area”. Further, for example, the non-continuous portion at time t1 and t3 (the non-continuous area having the item names 01pass1 and 03pass3) may be the “certain area”.
Here, for example, it is assumed that in the “some part” of the feature group information, there is a color or shade pattern that enables easy classification of cell types “CA” to “CD”. In this case, it is possible to easily classify the cell types “CA” to “CD” by comparing the patterns of colors and shades of “a certain area” (calculation of the degree of similarity in the actual processing). it can.
That is, when generating or updating the analytical model for classifying the cell types “CA” to “CD”, sample data is generated based on each item value (value of each characteristic parameter) of “a certain area”.
In addition, when a cell image having an unknown cell as a subject is an evaluation target and the unknown cell is classified into any of the types “CA” to “CD”, “a certain area” in the feature group information of the evaluation target. The evaluation target information is acquired based on each item value (value of each characteristic parameter) of. That is, the evaluation target information having the same form as the sample data is acquired.

Again, the “certain area” in the feature group information differs depending on the type of analysis model, that is, the analysis content, analysis method, various conditions, and the like.

Here, consider the data amount of the feature group information.
When there is no particular limitation on the number of dimensions N (data volume) of the feature group information, the set information of all values of all feature parameters obtained from the cell image (hereinafter referred to as “whole information”) is used as the feature group information. Can be adopted as
However, in reality, there may be a limit on the number of dimensions N (data capacity) of the feature group information. In such a case, it is necessary to adopt a part of the set information of all values of all feature parameters obtained from the cell image as the feature group information.
On the other hand, the feature group information with too small information amount hinders highly accurate analysis.
That is, it is necessary to generate feature group information that includes group information suitable for the analysis content, the analysis method, various conditions, etc. from the group information of all possible combinations.

FIG. 9 is a schematic diagram for explaining a difference in elements extracted for analysis and evaluation as the feature group information.
For example, in the example of FIG. 9, it is assumed that the cell types “A”, “B”, and “C” are classified.

In the example of FIG. 9(A), among the whole information, the values of the characteristic parameters of the two cells alone are arbitrarily selected as the shape, and randomly extracted in the time direction as the characteristic group information. Has been done. Specifically, for example, the feature group information of the cell of the type “A” is arbitrary from a large number of cells (objects) in a cell image captured at an early timing to two cell simple objects (objects). Is generated from the extracted values of the characteristic parameters of the two single cells.
In this case, many cell objects are included in one cell image even if they are of the same type “A”, and the values of the characteristic parameters are different. Therefore, the values of the feature parameters of the two arbitrarily extracted cells are not always representative values of the type "A". In addition, the value of the characteristic parameter may change with time, and the value of the characteristic parameter of the two single cells extracted at random time timings is not always the representative value of the type “A”.
From the above, when only the values of the characteristic parameters of some arbitrarily selected cells are used as the characteristic group information without considering the time change, the cell types “A”, “B”, and It is often difficult to accurately classify “C”.

Therefore, for example, the values of the respective characteristic parameters may be extracted based on the viewpoint of the cell population included in one cell image and the viewpoint of the temporal change, and the aggregate thereof may be used as the characteristic group information.
For example, in the example of FIG. 9B, in the case where the representative value of the feature amount in the cell population is extracted as the shape and the representative value of the feature amount is regularly extracted in terms of the shape in the overall information. The information over time (aggregation in the time direction) is adopted as the feature group information.
For example, in the example of FIG. 9C, in the case of the whole information, as the shape, the entire information of the feature amount in the cell population is extracted, and in terms of time, the entire information of the feature amount is regularly extracted. The information over time (aggregation in the time direction) is adopted as the feature group information.
Here, as shown in FIG. 9, the information amount (the number of dimensions N) of the feature group information increases in accordance with FIG. 9(A)→FIG. 9(B)→FIG. 9(C).

However, as described above, the number of dimensions N (data capacity) of the feature group information is often limited. In order to cope with such a restriction, it is necessary to extract suitable parameter values by organizing and selecting information from the whole information and generating feature group information as shown in the example of FIG. 9D. There is.
That is, for example, as shown in the example of FIG. 9(D), as the shape of the overall information, the effective information of the feature quantities in the cell population that has been processed by being emphasized and combined is extracted and temporally extracted. For the feature group information, time-lapse information (aggregation in the time direction) in the case where effective ones of the feature quantities in the cell population and processed by combination are periodically extracted is adopted as the feature group information.

Further, with reference to FIGS. 10 to 12, a method of generating the feature group information based on the emphasis and combination of effective ones of the feature amounts in the cell population will be described.

FIG. 10 is a diagram schematically showing characteristic group information generated from a cell image based on the emphasis and combination of effective ones of the characteristic amounts in the cell population.
In the example of FIG. 10, it is assumed that the same type of cell population is cultured in a predetermined container and the number of cells increases with time.
That is, in FIG. 10, only one cell is shown in the cell image Ga, only two cells are shown in the cell image Gb, and only four cells are shown in the cell image Gc. Of course, as shown in the cell image G of FIG. 10, the cell population is actually contained in the predetermined container. That is, the cells in FIG. 10 do not show one cell, but show a ratio (=1:2:4) that increases with time.

The feature group information Ia is obtained from the cell image Ga captured at the first time. The feature group information Ia includes, as N types of feature parameters (elements), at least three morphological feature quantities of a shape, a movement, and a surface of each cell, and population information of a cell population. Here, as the group information, there are statistical values such as an average value and a variance value, a bin value of a histogram, and the like.
For example, as the population information, numerical data Ika obtained from the "shape" histogram of the cell population, numerical data Iua obtained from the "movement" histogram of the cell population, and numerical values obtained from the "surface" histogram of the cell population The data Iha is included in the feature group information Ia.

Here, among the bins forming each histogram, the hatched bin is the information included in the feature group information Ia (one predetermined feature parameter if one bin is one element). ..
Here, in the “shape” histogram of the cell population, the three bins on the left side of the center characterize the “shape”. In other words, in the entire histogram, the "shape" feature is faint, so the three bins on the left side of the center are extracted to "emphasize" the "shape" feature.
On the other hand, unlike the “shape”, the five bins on the right side are extracted as the bins whose “motion” characteristics should be “emphasized”. Further, as the bins whose “surface” characteristics should be “emphasized”, six bins at both ends are extracted, unlike “shape” and “motion”.
The information (extracted bins) emphasized for each of these features (shape, movement, surface) is the numerical data Ika, the numerical data Iua, and the numerical data Iha, respectively.
Then, the characteristic group information Ia is created by the “combination” of the numerical data Ika, the numerical data Iua, and the numerical data Iha.

The feature group information Ib is obtained from the cell image Gb taken at the second time. The feature group information Ib includes, as N types of feature parameters (elements), at least three morphological feature quantities of a shape, a movement, and a surface of each cell, and population information of a cell population.
For example, as the population information, numerical data Ikb obtained from the histogram of the “shape” of the cell population, numerical data Iub obtained from the histogram of the “movement” of the cell population, and numerical values obtained from the histogram of the “surface” of the cell population. The data Ihb is included in the feature group information Ib.
That is, the pieces of information (extracted bins) emphasized for each feature (shape, movement, surface) are the numerical data Ikb, the numerical data Iub, and the numerical data Ihb, respectively. Then, the characteristic group information Ib is created by the “combination” of the numerical data Ikb, the numerical data Iub, and the numerical data Ihb.

The feature group information Ic is obtained from the cell image Gc taken at the third time. The feature group information Ic includes, as N types of feature parameters (elements), at least three morphological feature quantities of the shape, movement, and surface of each cell, and population information of the cell population.
For example, as population information, numerical data Ikc obtained from the histogram of the "shape" of the cell population, numerical data Iuc obtained from the histogram of the "movement" of the cell population, and numerical values obtained from the histogram of the "surface" of the cell population. The data Ihc is included in the feature group information Ic.
That is, the information (extracted bin) emphasized for each feature (shape, motion, surface) is the numerical data Ikc, the numerical data Iuc, and the numerical data Ihc, respectively. Then, the characteristic group information Ic is created by the “combination” of the numerical data Ikc, the numerical data Iuc, and the numerical data Ihc.

As described above, the pattern of information emphasis in the distribution, that is, the pattern of which bin in the histogram is extracted and emphasized is different for each feature (shape, motion, surface).
Such patterns are not limited to the three types shown in FIG. 10, and any type can be set.
FIG. 11 shows various specific examples of the information emphasis pattern in the distribution.
FIG. 11A shows a pattern using all distribution information of the histogram.
Here, as a histogram, absolute value information on the number (a number obtained by directly converting each number obtained from the cell image G in FIG. 11 into a histogram) and standard normalization information using standard values in the database (in FIG. 11). And a histogram obtained by normalizing each number obtained from the cell image G).
FIG. 11B shows a pattern using the average value/median value of the histogram.
FIG. 11C shows a pattern using information within the 80% confidence interval in the histogram (distribution).
FIG. 11D shows a pattern using information above the average value in the histogram (distribution).
FIG. 11E shows a pattern using information other than the 80% confidence interval in the distribution.

When generating the feature group information (or extracting the sample data described later), from all of these patterns, the most suitable index for each feature (shape, motion, surface) or for the analysis content, analysis method, and various conditions ( Pattern) and the point in time are used. The time point will be described later with reference to FIG.

Here, in the example of FIG. 3, each of the feature group information Ia to Ic in FIG. 10 is digitized information created from a single cell image in the image digitization unit 91, and the digitized information storage unit It is stored in 101.
Therefore, as described above with reference to FIG. 3, if another cell image (for example, a cell image of a plurality of wells) under the same condition exists at the same time, the multi-image feature group information generation unit 92 causes the other The feature group information of the first unit, which is combined with or organized by the feature group information of the cell image, is stored in the first feature group information storage unit 102.
The heterogeneity information adding unit 93 will be described later with reference to FIG. 12, and the processing of the temporal change information adding unit 94 will be described.
The aging information adding unit 94 performs processing (combining, organizing, etc.) in consideration of the aging of the group information among the feature group information Ia to Ic, so that the feature group information of the third processing unit. To generate.

FIG. 12 shows various specific examples of the method of processing the collective data in consideration of the change over time.
Note that, while FIG. 12 described above is an example based on cell images at three different timings, this FIG. 12 shows cell images Ga at four different timings (24h, 48h, 72h, and 96h). The example is based on Gb, Gc, and Gd. However, the concept of each processing method shown in FIG. 12 is similarly applied to the example of FIG.

In the example of FIG. 12, first, the standardized rearrangement type A and the standardized rearrangement type B can be selectively adopted.
The standardized rearrangement type A is a pattern that employs the absolute value information on the number shown in FIG. 11 (a histogram obtained by directly converting each number obtained from the cell image G of FIG. 11) as a distribution (histogram). Is.
The standardized rearrangement type B adopts the standard normalization information (standardized each histogram obtained from the cell image G of FIG. 11 and made into a histogram) using the standard value in the database shown in FIG. 11 described above. Pattern.

FIG. 12A schematically shows the feature group information in consideration of the change with time from time t1 to time t5 when the pattern using the entire distribution information of the histogram of FIG. 11A is adopted.
FIG. 12B schematically shows the feature group information in consideration of the temporal change from time t1 to t5 when the pattern using the average value/median value of the histogram of FIG. 11B is adopted. is there.
FIG. 12C schematically shows the feature group information in consideration of the change with time from time t1 to t5 when the pattern using the information within the 80% confidence interval in the histogram (distribution) of FIG. 11C is adopted. It is shown in the figure.
FIG. 12D schematically shows the feature group information in consideration of the change with time from time t1 to t5 when the pattern using the information above the average value in the histogram (distribution) of FIG. 11D is adopted. It is shown in.
FIG. 12(E) schematically shows the feature group information in consideration of the change over time from time t1 to t5 when the pattern using information other than the 80% confidence interval in the distribution of FIG. 11(E) is adopted. It is a thing.

Next, the heterogeneous information adding unit 93 of FIG. 3 will be specifically described with reference to FIG.
FIG. 13 shows various specific examples of a method of processing group information in consideration of the addition of heterogeneous information.
Heterogeneous information addition means addition of information indicating the relationship between a plurality of cell images captured at the same time and under a plurality of conditions. That is, it means that the feature group information obtained respectively from a plurality of cell images imaged at a plurality of conditions and at the same time is processed (combined, organized, etc.) in consideration of the condition change.

FIG. 13 shows an example in which cells of the same type A (three different containers) are subjected to culture condition A, culture condition B, and culture condition C, respectively, and then cultured.

That is, for the culture condition A, the feature group information IgAt at time t1, the feature group information IgAt2 at time t2, and the feature group information IgAt3 at time t3 are collected (processed) to form feature group information IgA.
Similarly, for the culture condition B, the feature group information IgBt1 at the time t1, the feature group information IgBt2 at the time t2, and the feature group information IgBt3 at the time t3 are collected (processed) to form the feature group information IgB. ..
Regarding the culture condition C, the feature group information IgCt1 at the time t1, the feature group information IgCt2 at the time t2, and the feature group information IgCt3 at the time t3 are collected (processed) to form the feature group information IgcC.

Further, the feature group information IgA of the culture condition A, the feature group information IgB of the culture condition B, and the feature group information IgC of the culture condition C are combined into higher-order information to form feature group information Ig1.
That is, the heterogeneous information is added in consideration of the change in the condition of culture.
Furthermore, the higher-order information is prioritized, and the feature group information Ig2 as high-dimensional information expressing the cell quality (here, the quality of the cell A) is obtained.

Compared with the characteristic group information at a predetermined time of one predetermined condition (for example, the characteristic group information IgAt1 at the time t1 of the culture condition A), the characteristic group information Ig2 takes into consideration heterogeneous information addition and change over time. Thus, the information is a better representation of the quality of cell A.

In the example of FIG. 13, it is described that the heterogeneous information is added after the processing (combining, rearranging, etc.) considering the temporal change, but the heterogeneous information is added at each time t1 to t3. After that, they may be processed (combined, organized, etc.) in consideration of changes with time such as putting them together.
That is, although the functional block diagram of FIG. 3 is a functional block diagram of the latter example, the arrangement order of the temporal change information adding unit 94 and the heterogeneous information adding unit 93 may be changed in accordance with FIG. good.

FIG. 14 shows a collection of 240 types of feature group information representing each cell quality that has been processed in this manner with the addition of heterogeneous information and changes taken into consideration, that is, feature group information representing the quality of a plurality of cells. Shows an example.
That is, in the example of FIG. 14, one predetermined column is feature group information as high-dimensional information expressing one type of cell quality.
In the example of FIG. 14, high-dimensional information expressing 240 types of cell qualities is not simply arranged, but the high-dimensional information having similar degrees of similarity is arranged closer to each other by clustering or the like described later. It is a thing. From this point of view, it is possible to understand that the bifurcation diagram showing closeness/distantness of similarity is a kind of analysis model, and in such a case, the data in FIG. 14 is the sample data from which the analysis model is derived. It can be understood as a group.

Heretofore, a specific example of a series of flows for generating the feature group information has been described with reference to FIGS.
Next, with reference to FIGS. 15 to 27, a series of flows from generating sample data from the feature group information and generating an analysis model using the sample data will be specifically described.

FIG. 15 is a diagram showing the difference between the conventional sample data and the sample data generated from the feature group information of the present embodiment. The sample data of the example of FIG. 15 is an analysis model for predicting a change in cell quality. Shall be used to generate.

Conventionally, the present inventors have also used the sample data P12 in consideration of the change over time in the characteristics of the cell population. However, the sample data P12 is obtained by simply arranging the data group P11 at each time of the average value of the cell morphology of the cell population, which is obtained from the cell image G at each time.
On the other hand, in the present embodiment, the aggregate P21 in which the information for each pattern shown in FIG. 11 is extracted at each time is obtained from the distribution (histogram) of the cell morphology of the cell population. The feature group information P22 is generated based on the aggregate P21. Then, the feature group information P22 is collected (arranged and processed as necessary), and high-dimensional feature group information P23 is obtained.
From this feature group information P23, it is possible to adopt any number of data of any combination as sample data.

FIG. 16 shows an example of sample data used to generate an analytical model for predicting changes in cell quality.
The example of FIG. 16 shows an example of sample data for each of the qualities A to E.
In FIG. 16, the minority feature amount information P1 corresponds to the conventional sample data P12 of FIG.
The high-dimensional feature amount information P2 corresponds to the high-dimensional feature group information P23 of this embodiment shown in FIG.

FIG. 17 shows the prediction accuracy when the cell quality change is predicted using the sample data of FIG.
“Average of cell morphology” indicates prediction accuracy when the minority feature amount information P1 of FIG. 16 (conventional sample data P12 of FIG. 15) is used.
The “standardized type A” is time-series information of a pattern using all distribution information of the histogram of FIG. 12A in the high-dimensional feature amount information P2 of FIG. 16 (high-dimensional feature group information P23 of FIG. 15). The part of is extracted as sample data, and the prediction accuracy when the sample data is used is shown.
“Standardized type B” means a pattern in which the average value/median value of the histogram of FIG. 12B is used in the high-dimensional feature amount information P2 of FIG. 16 (high-dimensional feature group information P23 of FIG. 15). The prediction accuracy is shown when the sequence information portion is extracted as sample data and the sample data is used.
The “standardized type C” is within the 80% confidence interval in the histogram (distribution) of FIG. 12C, out of the high-dimensional feature amount information P2 of FIG. 16 (high-dimensional feature group information P23 of FIG. 15). The time-series information part of the pattern using information is extracted as sample data, and the prediction accuracy is shown when the sample data is used.
The “standardized type D” is the information above the average value in the histogram (distribution) of FIG. 12D among the high-dimensional feature amount information P2 of FIG. 16 (high-dimensional feature group information P23 of FIG. 15). The time-series information part of the pattern using is extracted as sample data, and the prediction accuracy is shown when the sample data is used.
The “standardized type E” uses information other than the 80% confidence interval in the distribution of FIG. 12E among the high-dimensional feature amount information P2 of FIG. 16 (high-dimensional feature group information P23 of FIG. 15). The prediction accuracy is shown when the time-series information part of the pattern is extracted as sample data and the sample data is used.

In the example of FIG. 17, it can be seen that the standardized type E has the best prediction accuracy.
That is, in the present embodiment, from the high-dimensional feature group information P23, an information group of an arbitrary number and arbitrary combination (including the total number, that is, including the high-dimensional feature group information P23 itself) can be used as sample data. it can.
Therefore, when generating a predetermined analysis model, if sample data (standardized type E in the example of FIG. 17) suitable for the analysis content, analysis method, various conditions, etc. in which the analysis model is used can be found, the sample data By using, it becomes possible to generate an analysis model suitable for the analysis content, analysis method, various conditions, and the like.

FIG. 18 shows the result of actually predicting the change in cell quality by the analytical model generated using the standardized type E sample data in FIG. 17 in this way.
In FIG. 18, the row direction indicates the correct answer label, and the column direction indicates the analysis result label.
55 sets of analyzes were performed on a set of cells of a predetermined type (quality) and predetermined cell culture conditions, and the analysis results are shown by square marks. That is, when the square mark of the analysis result is attached to the item where the label of the correct answer and the label of the analysis result match, it indicates that the prediction is successful.
As shown in FIG. 18, it can be seen that the prediction can be performed with extremely high accuracy.

FIG. 19 shows an example of a plurality of sample data used for generating a predetermined analysis model different from the examples described in FIGS. 15 to 18.
As described above, the plurality of sample data Iga to Igj are not image data but numerical data, and specifically, are multidimensional information including M types of parameters as elements.
Here, a single numerical value element that constitutes the sample data is also a value indicating the morphological characteristic amount of the cell and is also significant information.
However, the sample data Iga to Igj formed by the combination of each of these elements, as a whole, serve as more characteristic and significant information in comparison with other sample data.
In the example of FIG. 19, each element is represented by a color and a density corresponding to the value thereof, and a “pattern” formed as a set of colors and a density of each of these elements corresponds to each of the sample data Iga to Igj. It can be seen that it is significant information that shows the characteristics well.

Here, it is assumed that a comprehensive score is given to each of the plurality of sample data Iga to Igj from a predetermined viewpoint.
The predetermined viewpoint is not particularly limited, but for the sake of convenience of description, it is assumed that, for example, the viewpoint of classifying cell types into type A and type B is adopted. The closer the score is to 100, the more the characteristics peculiar to the type A are included (the characteristics peculiar to the type B are smaller). On the contrary, it is assumed that the closer the score is to 0, the more the characteristics peculiar to the type B are included (the characteristics peculiar to the type A are smaller).
That is, in the example of FIG. 19, the sample data Iga has a score of “90”, which is the highest value, so that the sample data Iga includes many features peculiar to the type A. On the contrary, since the sample data Ige to Igh have the lowest score of “0”, the sample data Ige to Igh are data including many characteristics peculiar to the type B. Since the sample data Igi has a median value of “50”, the sample data Igi is a data including the characteristic peculiar to the type A and the characteristic peculiar to the type B.

The method of giving a score to each of the plurality of sample data Iga to Igj is not particularly limited.
For example, in the case where corresponding evaluation target information (known information obtained by evaluation such as destructive test) for each of the plurality of sample data Iga to Igj is stored in the cell evaluation DB 62 of FIG. You may employ the method of giving a score based on.

FIG. 20 shows two examples of methods for generating a predetermined analysis model using the sample data Iga to Igj shown in FIG.
Specifically, in combination with the above example, the example of FIG. 20 is an example of generation of an analytical model for classifying cell types into type A and type B.

FIG. 20A is an example of generating an analysis model using machine learning.
In the example of FIG. 20A, the analysis model generation unit 44 executes machine learning using a plurality of sample data Iga to Igj, and scores each value X of M parameters of the sample data as an input parameter. A function that outputs Y (Y=aX+bX+cX in the example of the figure) is generated or updated.

Here, it is assumed that the feature group information is generated from a cell image including an unknown type of cell as a subject and stored in the feature group information DB 61 (FIG. 2).
The evaluation target information acquisition unit 46 acquires, from the characteristic group information, data of the same form (M parameters are the same) as the sample data Iga to Igj as the evaluation target information from the evaluation target information acquisition unit 46.

In this case, the analysis unit 47 calculates the score Y of the evaluation target information by substituting each value X of the M parameters of the evaluation target information as an input parameter into the function.
When the score Y of the evaluation target information is close to 100, it means that the cells corresponding to the evaluation target information are analyzed as being highly likely to be type A.
When the score Y of the evaluation target information is close to 0, it means that the cell corresponding to the evaluation target information is analyzed as being highly likely to be of type B.
When the score Y of the evaluation target information is an intermediate value, it means that the cell corresponding to the evaluation target information cannot be judged to be the type A or the type B.

That is, “a function that outputs the score Y of the evaluation target information with each value X of the M parameters of the evaluation target information as an input parameter” is an analysis model for classifying the cell type into the type A and the type B. Is an example.

FIG. 20(B) is an example of generation of an analytical model using a clustering method.
In the example of FIG. 20B, the analysis model generation unit 44 classifies the plurality of sample data Iga to Igj into classes by using a predetermined algorithm. The type and number of classes that can be obtained as a result of class classification depend on a plurality of sample data Iga to Igj and a predetermined algorithm, but here, it is assumed that they are classified into three classes CA, CB, and CN.

Here, similar to the above-described example of FIG. 20A, the feature group information is generated from a cell image including cells of an unknown type as a subject and is stored in the feature group information DB 61 (FIG. 2). And
The evaluation target information acquisition unit 46 acquires, from the characteristic group information, data of the same form (M parameters are the same) as the sample data Iga to Igj as the evaluation target information from the evaluation target information acquisition unit 46.

In this case, the analysis unit 47 classifies the evaluation target information into any of the three classes CA, CB, and CN.
When the evaluation target information is classified into the class CA, it means that the cells corresponding to the evaluation target information are analyzed as being highly likely to be type A.
When the evaluation target information is classified into the class CB, it means that the cells corresponding to the evaluation target information are analyzed as being highly likely to be of type B.
Further, when the evaluation target information is classified into the class CN, it means that the cell corresponding to the evaluation target information cannot be judged to be the type A or the type B.

That is, "a model for classifying the evaluation target information into any of the three classes CA, CB, and CN" is an example of an analysis model for classifying the cell types into the type A and the type B.
Here, the method of class classification is not particularly limited. For example, the degree of similarity between each of the sample data Iga to Igj and the evaluation target information is obtained, and the evaluation target information is classified based on the similarity or the like. Techniques can also be adopted. When this method is adopted, the analysis model is represented by “a plurality of sample data Iag to Ig and three classes CA, CB, CN generated based on these”.

Note that in FIG. 20B, a simple class classification is described for convenience of explanation.
An actual analytical model is represented by stacking regions (regions of elliptical solid lines in FIG. 21) based on each sample data as shown in FIG. 21, for example.
That is, FIG. 21 shows an example of the expression form of the analysis model.
Circles in this analysis model indicate sample data or evaluation target information.

The evaluation target information qa is included in the region based on the sample data of all cells of type A. Therefore, the cells corresponding to the evaluation target information qa have been analyzed as being highly likely to be of type A.
The evaluation target information qb is included in the region based on the sample data of many cells of type A. Therefore, the cells corresponding to the evaluation target information qb are analyzed as being highly likely to be of type A (however, lower than the evaluation target information qa).

On the other hand, the evaluation target information qc is included in the area based on the sample data of all cells of type B. Therefore, the cells corresponding to the evaluation target information qc have been analyzed as having a very high possibility of type B.
The evaluation target information qd is included in the region based on the sample data of many cells of type B. Therefore, the cells corresponding to the evaluation target information qc have been analyzed as having a high possibility of type B (however, lower than the evaluation target information qc).

Further, the evaluation target information qe is included in the overlapping area of the region based on the sample data of the type A cells and the region based on the sample data of the type B cells. Therefore, it is determined that the cell corresponding to the evaluation target information qe cannot be determined to be the type A or the type B.

As described above, analysis of cells is diverse in terms of analysis content, analysis method, various conditions, and the like. Therefore, it is necessary to use sample data and evaluation target information that are suitable for the analysis content, analysis method, various conditions, etc., and this embodiment makes this possible.
In other words, the sample data of the present embodiment is not simply the feature group information extracted as it is, but its form (the number of parameters M or the combination of parameter types) is set so as to be suitable for the analysis content, analysis method, various conditions, and the like. Etc.) changes.
Then, an analysis model is generated from one or more sample data that matches the analysis content, analysis method, various conditions, and the like. Therefore, the analysis model is also adapted to the analysis content, analysis method, various conditions, and the like.
Then, when the analysis is performed on the unknown cell, the evaluation target information acquisition unit 46 acquires the evaluation target information having the same form as the sample data as the data of the unknown cell. That is, the evaluation target information that matches the analysis content, analysis method, various conditions, etc. is acquired.
The analysis unit 47 uses the evaluation target information and the analysis model (sample data) that are suitable for the analysis content, the analysis method, various conditions, and the like, and executes the analysis process on the unknown cell.
In this way, unknown cells can be easily and appropriately analyzed regardless of differences in analysis content, analysis method, various conditions, and the like.
That is,

Furthermore, the analysis processing by the analysis unit 47 and the output information output from the output information generation unit 48 will be described in more detail below.

Here, as described above, the sample data is a set of numerical values of M kinds of parameters (various characteristic parameters and various incidental parameters), that is, digitized information. Therefore, it is easy to represent the analytical model generated from the aggregate of the plurality of sample data as the numerical information.
The evaluation target information is also used for cells to be analyzed. The evaluation target information is digitized information having the same form as the sample data.
Therefore, it is easy to generate the output information indicating the result of the analysis process using the evaluation target information and the analysis model (sample data) as a list (a list or the like) of digitized information.
However, the user who has requested the analysis of the cells cannot easily recognize the analysis result or the like even if only the list (list or the like) of such quantified information is presented.
That is, as an output form to the user, it is not enough to enumerate the digitized information (list, etc.), and it is preferable that there is a form that is easy to visually recognize.
In order to enable such an output form that is easily visible to the user, the expression form of the analysis model becomes important.
For example, the above-described expression form of the analysis model shown in FIG. 21, that is, the form in which the analysis model is formed by stacking the regions (the elliptical solid line regions in FIG. 21) based on each sample data is an output form that is easy for the user to visually recognize. This is an example.

FIG. 22 is a diagram showing an example of the expression form of the analysis model, which is different from the example shown in FIG.
In the example of FIG. 22, each sample data is plotted as a point on a plane composed of predetermined two axes (X axis, Y axis). Each sample data is color-coded (classified by different hatching in the figure) according to the group to which it belongs. That is, the peripheral area in which the sample data of the same color is distributed shows the area of the group corresponding to the color.
For example, in the fourth quadrant in the figure, the sample data of the first color (in the figure, the diagonal lines having the narrowest intervals) are concentrated near the sample data SA.
That is, it can be said that the region of the fourth quadrant in the figure is mainly the region of the first group corresponding to the first color.
Similarly, it can be said that the region of the first quadrant (peripheral region of the sample data SB) in the figure is mainly the region of the second group corresponding to the second color (the diagonal line having the second narrowest interval in the figure). ..
It can be said that the peripheral area on the positive side of the Y axis (the peripheral area of the sample data SC) is mainly the area of the third group corresponding to the third color (the diagonal line having the widest interval in the drawing).
It can be said that the peripheral region on the negative side of the X axis (peripheral region of the sample data SD) is mainly the region of the fourth group corresponding to the fourth color (white in the figure).

Further, the evaluation target information q can also be plotted on the plane showing the analysis model of FIG.
Accordingly, the user easily and promptly finds that the arrangement position of the evaluation target information q is the peripheral region on the positive side of the Y axis, and thus the cells corresponding to the evaluation target information q are likely to belong to the third group. Can be seen.

Here, the expression form (color, symbol shape, etc.) of each plot of sample data and evaluation target information can be arbitrarily changed. Furthermore, if it is sample data obtained from the same cell image group, the expression form can be arbitrarily changed according to the analysis content, the analysis method, various conditions, and the like.
For example, when expressing an analytical model for classifying cell types, it is possible to adopt an expression form in which each plot point is color-coded for each type.
On the other hand, for example, when expressing an analysis model for determining the state of cells, another form of expression in which the shape of the symbol is changed at each plot point for each state can be adopted. For example, it is possible to adopt an expression form in which the state is “◯” (circle mark) when it is “good”, “Δ” (triangle mark) when it is “medium”, and “Δ” when it is “bad”.
In this case, the user can easily and quickly compare the cell types by comparing the colors of the plot points, and can easily and immediately compare the cell states by comparing the shapes of the symbols on the plot. it can.

Furthermore, the X axis and the Y axis are arbitrary variable axes, and can be easily changed by a user's operation or the like.
Specifically, as described above, the sample data and the evaluation target information can be understood to be M-dimensional multidimensional information having M types of parameters as elements.
Therefore, the plot points on the plane of FIG. 22 can be easily determined only by generating the two elements X and Y based on the sample data and the M kinds of parameters of the evaluation target information.
Here, the X axis and the Y axis are not particularly limited and may be arbitrary.
For example, an arbitrary two-dimensional axis of the M dimensions of the sample data or the evaluation target information can be directly used as the X axis and the Y axis.
Further, for example, of M dimensions, an arbitrary number of arbitrary dimensions are combined to generate a new X axis, and independently of that, an arbitrary number of arbitrary dimensions are combined to generate a new Y axis. Can also

Furthermore, the expression form of the analysis model does not have to be a two-dimensional space, and may be a space of any dimension.
For example, FIG. 23 is a diagram showing an example of the expression form of the analysis model, which is different from the examples shown in FIGS. 21 and 22.
That is, the analysis model is represented in the two-dimensional space in the examples of FIGS. 21 and 22, whereas it is represented in the three-dimensional space in the example of FIG.
Here, the three axes forming the three-dimensional space are arbitrary variable axes, just as in the case of the two-dimensional space, and can be easily changed by a user's operation or the like.
In principle, reduction of dimension means reduction (missing) of the amount of information with respect to the original information. That is, in any of the spaces in FIGS. 21 to 23, an M-dimensional space (when M is 4 or more) consisting of M kinds of parameters of the sample data is projected, etc. It is a space where information is missing. Therefore, in the three-dimensional space of FIG. 23, as compared with the two-dimensional space of FIG. 21 and FIG. 22, the amount of missing information decreases as the number of dimensions increases, so that the user can easily obtain a larger amount of information. You can get it instantly.

Furthermore, the expression form of the analysis model does not have to be a space in which the sample data and the evaluation target information are plotted as points, and it is sufficient if the similarity between the sample data and the evaluation target information can be visually recognized.
Here, the similarity between the two data can be represented as a distance between the two data, for example. That is, the closer the data is, the more similar the data is.
Since the expression form that allows such a sense of distance to be easily visually recognized is a point (plot) in space, first, the expression form of the analysis model of FIGS. 21 to 23 described above has been described.
Next, another expression form in which the similarity of sample data and evaluation target information can be easily visually recognized will be described with reference to FIG.

FIG. 24 is a diagram showing an example of the expression form of the analysis model, which is different from the examples shown in FIGS. 21 to 23.
In the example of FIG. 24, one line with an end represents one piece of data.
In other words, an aggregate of each line indicating each of the plurality of sample data (hereinafter, the line is referred to as “branch” and an expression form of the aggregate of branches is referred to as “branch branch diagram”) uses the plurality of sample data. The analysis model generated by the above is shown.
As described above, the closer the distance is, the more similar the distances are to each other. Therefore, it is possible to grasp a group of branches having a short distance as one group.
Here, as shown in FIG. 22, since there are several branching points of the branch, it is possible to easily generate a hierarchical group for each unit of the branching point of the branch.
For example, in the example of FIG. 24, the cells are classified into normal cells and senescent cells in the first hierarchy.
Further, in each of the second layers, they are classified into a plurality of groups depending on the difference in culture conditions.

Furthermore, the evaluation target information q can also be easily plotted as one branch on the branch branch diagram representing the analysis model of FIG.
Accordingly, the user can easily and easily find that the cells corresponding to the evaluation target information q are highly likely to belong to the group of the culture condition A among the normal cells from the plot position of the branch of the evaluation target information q in the branch bifurcation diagram. It can be seen immediately.

Here, it should be noted that the examples of FIGS. 21 to 24 are expressed only from the viewpoint that the user can easily visually recognize the positional relationship between each sample data (outline of the analysis model) and the evaluation target information. It was done.
That is, as described above, the sample data and the evaluation target information are originally M-dimensional multidimensional information. What is expressed as a space obtained by compressing the M dimension into two dimensions or three dimensions is an example of FIGS. 21 to 24.
That is, the analysis unit 47 of FIG. 2 does not actually calculate the similarity based on the two-dimensional or three-dimensional data, but calculates the similarity between each M-dimensional sample data and the M-dimensional evaluation target information. I am calculating.
However, for convenience of explanation, assuming that M=3, an outline of a calculation method of the similarity between each sample data and the evaluation target information will be described with reference to FIGS. 25 to 27.
That is, in the examples of FIGS. 25 to 27, it should be noted that the M=3D space happens to be a different space, which is originally different from the 3D space of FIG.

FIG. 25 shows a configuration example of the feature group information which is the basis of the generation of the sample data and the evaluation target information.
In FIG. 25, items in the horizontal direction in FIG. 25 indicate respective elements (feature parameters) from a predetermined viewpoint (mean value, balance, distribution form) in a cell population of predetermined cells. That is, the feature group information on which the sample data is based includes the value of each item in the horizontal direction in the drawing as the value of each element.
Items in the vertical direction in the figure indicate individual elements (feature parameters) in a predetermined cell. That is, the feature group information on which the sample data is based includes the value of each item in the vertical direction in the drawing as the value of each element.
Note that, in the example of FIG. 25, the feature group information ICt is shown for a cell image captured at a predetermined time, using a cell population of the same type cultured in a predetermined container as a subject.

The feature group information ICt is configured as information in a two-dimensional space including a horizontal axis indicating elements of a predetermined viewpoint in the cell population and a vertical axis indicating individual elements of the cell.
Here, the unit information Iu is a value group (information group on the vertical axis) of the characteristic parameter for each of the predetermined cells at a predetermined time for the predetermined cell and a predetermined viewpoint in the cell population including the predetermined cell. It is information generated based on the value group of the characteristic parameter (information group on the horizontal axis).
By using such unit information Iu, it is possible to grasp the position of the predetermined cell in the cell population from a predetermined viewpoint at a predetermined time.
For example, from the unit information Iu indicated by the diagonal line at the upper left corner of FIG. 25, the average of each characteristic parameter of the population cell at a predetermined time for the value (for example, length) of each characteristic parameter of the cell 1 (predetermined cell). It becomes possible to make comparisons in terms of values (for example, the average value of the length of the entire population).

Here, for the cell population contained in the same container, it is assumed that the cell image is captured n times in the time direction (each of time t1 to time tn).
In this case, the feature group information ICt of FIG. 25 is the cell image captured at a predetermined time tk (k is an arbitrary integer within the range of 1 to n) for the cell population contained in the same container. It shows the situation.
That is, as shown in FIG. 26, n pieces of feature group information ICt1 to ICtn showing the states of the cell images captured at times t1 to tn for the cell population contained in the same container are obtained.
The feature group information IC is configured by sequentially stacking these n pieces of feature group information ICt1 to ICtn in the time direction.

That is, one or more unit information Iu is arranged in a three-dimensional space composed of a vertical axis indicating individual information of cells, a horizontal axis indicating information of a cell population from a predetermined viewpoint, and a depth axis indicating time change. The information group is the feature group information IC.
Here, the form of the unit information Iu is not particularly limited, but for convenience of explanation, it is assumed that it takes a binary value indicating whether it is “significant information (1)” or “not significant information (0)”.
For example, when a predetermined analysis content such as classifying cell types is used, a value that assists in classifying cell types in the values of each element (various characteristic parameters) in each unit information Iu May be included. If such a value is included, the unit information Iu becomes “significant information (1)”. On the other hand, if a value that assists in classifying cell types is not included, the unit information Iu is “not significant information (0)”.
In other words, even for cells of the same type contained in the same container, "significant information (1)" is obtained according to the characteristics of individual cells, the positioning from a predetermined viewpoint in the cell population, the time change, etc. In some cases, it may be “not significant information (0)”.

The analysis model generation unit 44 (FIG. 2) constructs one feature group information IC with n cell images taken at times t1 to tn of cells of the same type contained in the same container as one unit. Then, only the unit information Iu that is “significant information (1)” is extracted from the one-unit feature group information IC.
The analysis model generation unit 44 generates the aggregate of the plurality of unit information Iu thus extracted as sample data.
Here, as shown in FIG. 26, the unit information Iu can be represented by position coordinates in a three-dimensional space that constructs the feature group information IC, that is, three-dimensional information.
That is, as shown in FIG. 27, the unit information Iu of “significant information (1)” (hatched unit information Iu in FIG. 27) arranged in the three-dimensional space that constructs the feature group information Ic, respectively. The collection of is the sample data.
For example, it is assumed that the sample data ICS1 on the upper right of FIG. 27 indicates the type A. Further, for example, it is assumed that the sample data ICS2 at the lower center of FIG. 27 indicates the type B.
That is, an aggregate of these sample data ICS1, ICS2, etc. becomes an analysis model for classifying cell types into type A and type B.

In other words, the unit information Iu is, in the example of FIG. 25, a feature including at least 10 elements (feature parameters) of a predetermined cell simple substance and 10 elements (feature parameters) of a predetermined viewpoint of population information. The group information is the basis. Each element of this is collected from the viewpoint of "significant information (1)" or "is not significant information (0)", and the unit information Iu of "significant information (1)" constructs the feature group information IC, respectively. Position coordinates in a three-dimensional space, that is, three-axis elements.
That is, it is equivalent to that three-dimensional sample data is extracted from at least 20-dimensional feature group information.

Here, also for a cell population of an unknown type contained in the same container, n cell images taken at times t1 to tn become one unit, and this one unit is given as an analysis target. ..
In this case, the evaluation target information acquisition unit 46 acquires data constructed by the same method as the sample data ICS1 and ICS2 described above as the evaluation target information Ict shown in the upper left of FIG.

The analysis unit 47 obtains the degree of similarity by calculating the distance between the evaluation target information Ict and the sample data ICS1. Similarly, the analysis unit 47 obtains the degree of similarity by calculating the distance between the evaluation target information Ict and the sample data ICS2.
The method for calculating the distance is not particularly limited, and for example, the Euclidean distance or Mahalanobis distance method can be adopted.
In the example of FIG. 27, since the distance between the evaluation target information Ict and the sample data ICS2 is shorter than the distance between the evaluation target information Ict and the sample data ICS1, the evaluation target information Ict is similar to the sample data ICS2. Is judged.
That is, the cells corresponding to the evaluation target information Ict are analyzed as being highly likely to be of type B indicated by the sample data ICS2.

The output information generation unit 48 generates output information including the result analyzed by the analysis unit 47 in this way and the evaluation target information acquired by the evaluation target information acquisition unit 46, and outputs the output unit 16 or an external device. Output to 81.

FIG. 28 shows an example of a screen showing output information.
On the screen G0 in the example of FIG. 28, the analysis results such as the type of cells to be analyzed (cells in culture) are displayed not only as a list of numerical values but also as a belonging space map or a belonging map that is easily visible to the user. To be done.

As described above with reference to FIG. 23, the belonging space map plots various sample data and evaluation target information indicating cells to be analyzed (cells in culture) on dots in a three-dimensional space of arbitrary axes. This is the map. In the example of FIG. 28, the plot points (circles) described as “Query” indicate the evaluation target information.

On the screen G0, as shown in FIG. 29, it is possible to display the belonging map instead of the belonging space map.
As described above with reference to FIG. 23, the affiliation map is a branch branch diagram showing that various sample data are regarded as one branch (line), and that the closer the branch is, the more similar the branch data is. In the example of FIG. 27, the plot points (circles) described as “Query” indicate the evaluation target information.
In this way, the user can instantly and easily visually recognize the analysis result of the cell to be evaluated (what kind of cell is likely to belong, etc.) simply by looking at the belonging space map and the belonging map. , Convenience.

Another characteristic is that the classification result is presented by probability (probability).
That is, it is not determined as type A, but is presented in terms of the degree of similarity with the sample of type A (probability of becoming type A).
This provides more accurate and meaningful information for the user. For example, as described above, when there is a possibility of both the type A and the type B, it is assumed that the result determined to be the type A is presented although the possibility of the type A is slightly high. In this case, if it is actually the type B, it means that the wrong analysis result has been transmitted, and the user may have trouble later.
On the other hand, in the present embodiment, it is suggested that there are both possibilities of type A and type B. With this, the user can consider the possibility of becoming the type B, so that it is possible to easily deal with the case of the type B actually.

Another characteristic is that not only the “evaluation probability” indicating the type (cytoma) but also the “homology score” are presented.
That is, for convenience of explanation, the analysis method for classifying cell types has been mainly described, but the similarity of “origin” can also be output by similarly comparing and analyzing sample data and evaluation target information. It is possible. This “origin” similarity is the homology score.
For example, when comparing the sample data of the second and third places in the ranking, the evaluation probability (similarity of the type) is “78” for the sample data of the second place, while the sample data of the third place is significantly different from “75”. Absent. However, in the homology score (similarity of origin), the sample data at the second place is “80”, whereas the sample data at the third place is “70”, which is a large difference. Therefore, the user can easily determine to some extent that the “origin” of the cell to be analyzed is “bone marrow” rather than “fat”.

Although one embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and modifications, improvements, etc. within a range in which the object of the present invention can be achieved are included in the present invention. is there.

In other words, the cell analysis result output device to which the present invention is applied may have the following configuration, and the embodiment thereof is not particularly limited. That is, the cell analysis device 1 described above is an example of a cell analysis result output device.

That is, the cell analysis result output device to which the present invention is applied,
Based on the data of one or more cell images included in the unit, with one or more cell images selected under a predetermined condition as a unit from among the cell images obtained by capturing a cell population including a plurality of cells as a subject. , Information including the values of N kinds (N is an integer value of 1 or more) of characteristic parameters relating to the morphological characteristics of the cells alone or the morphological characteristics of the cell population is held as characteristic group information,
Based on at least one of the N types of feature parameters included in the feature group information to be compared, the comparison target information including M types (M is an integer value independent of N) of parameter values is Acquired,
Comparison target information composed of values of the M types of parameters is acquired based on at least one of the N types of feature parameters included in the feature group information of the comparison target,
When the distance from the predetermined viewpoint between the comparison target information and the comparison partner information is calculated,
The image information including the first symbol indicating the comparison target information and the second symbol indicating the comparison target information, the display form of which changes according to the calculated distance, is output as output information. Output means (for example, the output information generation unit 48 in FIG. 2)
Prepare for it.
As a result, the user can easily and effectively recognize the result of the analysis process regarding various cells having different analysis contents, analysis methods, various conditions, and the like.

When the series of processes is executed by software, a program forming the software is installed in a computer or the like from a network or a recording medium.
The computer may be a computer embedded in dedicated hardware. Further, the computer may be a computer capable of executing various functions by installing various programs, for example, a general-purpose personal computer.

The recording medium containing such a program is not only configured by the removable medium 31 of FIG. 1 which is distributed separately from the apparatus main body to provide the program to the user, but is also installed in the apparatus main body in advance. It is composed of a recording medium provided to the. The removable medium 31 is composed of, for example, a magnetic disk (including a floppy disk), an optical disk, a magneto-optical disk, or the like. The optical disk is configured by, for example, a CD-ROM (Compact Disk-Read Only Memory), a DVD (Digital Versatile Disk), or the like. The magneto-optical disk is composed of MD (Mini-Disk) or the like. Further, the recording medium provided in advance to the user in a state of being incorporated in the apparatus main body is, for example, the ROM 112 in FIG. 3 in which the program is recorded, the ROM 212 in FIG. 5, the storage unit 118 in FIG. It is composed of a hard disk and the like included in the unit 216.

The step of writing the program recorded in the recording medium is not limited to the processing performed in time series according to the order, but the processing performed in parallel or individually is not necessarily performed in time series. It also includes.

DESCRIPTION OF SYMBOLS 1... Cell analysis device, 11... CPU, 16... Output unit, 18... Storage unit, 31... Removable media, 41... Cell image input unit, 42... Feature group Information generation unit, 43... Noise removal unit, 44... Analysis model generation unit, 45... Evaluation target information acquisition unit, 46... Evaluation target information acquisition unit, 47... Analysis unit, 48... .. Output information generation unit, 61... Feature group information DB, 62... Cell evaluation DB, 63... Analysis model holding unit, 81... External device, 82... Preprocessing device, 91. ..Image digitizing unit 92...Multiple image feature group information generating unit 93...Heterogeneous information adding unit 94...Change information adding unit 95...Feature group information output unit 95 101... Digitized information storage unit, 102... First feature group information storage unit, 103... Second feature group information storage unit, 104... Third feature group information storage unit 104

Claims (6)

Based on the data of one or more cell images included in the unit, with one or more cell images selected under a predetermined condition as a unit from among the cell images obtained by capturing a cell population including a plurality of cells as a subject. , Information including the values of N kinds (N is an integer value of 1 or more) of characteristic parameters relating to the morphological characteristics of the cells alone or the morphological characteristics of the cell population is held as characteristic group information,
The number M of types extracted from the N types of characteristic parameters is changed to any one of 1 to N according to the characteristics of cells to be compared or the characteristics of a predetermined analysis,
Information composed of the values of the M types of parameters extracted from the N types of characteristic parameters on the basis of at least one of said N types of characteristic parameters included in the feature group information regarding cells of the comparison is , it is obtained as compared information,
Information composed of the values of the M kinds of parameters extracted from the N kinds of characteristic parameters based on at least one kind of the N kinds of characteristic parameters included in the characteristic group information regarding the cells of the comparison partner , is obtained as compared to the other information,
When the distance from the predetermined viewpoint between the comparison target information and the comparison partner information is calculated,
The image information including the first symbol indicating the comparison target information and the second symbol indicating the comparison target information, the display form of which changes according to the calculated distance, is output as output information. Output means to
A cell analysis result output device comprising: The output means is
In a space constructed from S (S is an integer value of M or less) axes defined based on the M kinds of parameters,
The cell analysis according to claim 1, wherein information of the image showing a state in which the first symbol and the second symbol are arranged at positions that change according to the calculated distance is output as the output information. Result output device. The cell analysis result output device according to claim 2, wherein the first symbol and the second symbol are points arranged in the space. The cell analysis result output device according to claim 2, wherein the first symbol and the second symbol are branches arranged in the space. In the cell analysis result output method executed by the device that outputs the analysis result of cells,
Based on the data of one or more cell images included in the unit, with one or more cell images selected under a predetermined condition as a unit from among the cell images obtained by capturing a cell population including a plurality of cells as a subject. , Information including the values of N kinds (N is an integer value of 1 or more) of characteristic parameters relating to the morphological characteristics of the cells alone or the morphological characteristics of the cell population is held as characteristic group information,
The number M of types extracted from the N types of characteristic parameters is changed to any one of 1 to N according to the characteristics of cells to be compared or the characteristics of a predetermined analysis,
Information composed of the values of the M types of parameters extracted from the N types of characteristic parameters on the basis of at least one of said N types of characteristic parameters included in the feature group information regarding cells of the comparison is , it is obtained as compared information,
Information composed of the values of the M kinds of parameters extracted from the N kinds of characteristic parameters based on at least one kind of the N kinds of characteristic parameters included in the characteristic group information regarding the cells of the comparison partner , is obtained as compared to the other information,
When the distance from the predetermined viewpoint between the comparison target information and the comparison partner information is calculated,
The image information including the first symbol indicating the comparison target information and the second symbol indicating the comparison target information, the display form of which changes according to the calculated distance, is output as output information. Output step to
A method for outputting cell analysis results including. To the computer that executes the control that outputs the result of the cell analysis,
Based on the data of one or more cell images included in the unit, with one or more cell images selected under a predetermined condition as a unit from among the cell images obtained by capturing a cell population including a plurality of cells as a subject. , Information including the values of N kinds (N is an integer value of 1 or more) of characteristic parameters relating to the morphological characteristics of the cells alone or the morphological characteristics of the cell population is held as characteristic group information,
The number M of types extracted from the N types of characteristic parameters is changed to any one of 1 to N according to the characteristics of cells to be compared or the characteristics of a predetermined analysis,
Information composed of the values of the M types of parameters extracted from the N types of characteristic parameters on the basis of at least one of said N types of characteristic parameters included in the feature group information regarding cells of the comparison is , it is obtained as compared information,
Information composed of values of the M kinds of parameters extracted from the N kinds of characteristic parameters based on at least one kind of the N kinds of characteristic parameters included in the characteristic group information regarding cells of the comparison partner , is obtained as compared to the other information,
When the distance from the predetermined viewpoint between the comparison target information and the comparison partner information is calculated,
The image information including the first symbol indicating the comparison target information and the second symbol indicating the comparison target information, the display form of which changes according to the calculated distance, is output as output information. Output step to
A program that executes control processing including.