WO2019175959A1

WO2019175959A1 - Method for assessing cell state and device for assessing cell state

Info

Publication number: WO2019175959A1
Application number: PCT/JP2018/009698
Authority: WO
Inventors: 礼子赤澤
Original assignee: 株式会社島津製作所
Priority date: 2018-03-13
Filing date: 2018-03-13
Publication date: 2019-09-19

Abstract

A microscope observation unit (1) acquires hologram data pertaining to a culture plate (12) that includes a colony (13), and a phase information calculation unit (22) and a moving image creation unit (23) perform prescribed processing on the data to thereby create a phase image. A texture feature value computation unit (241) and a texture feature image creation unit (242) perform statistical texture analysis on each small region of the phase image and create multiple types of texture feature images. A cell region extraction unit (243) extracts a colony region from a texture feature image, and an index value calculation unit (244) computes index values such as the mean or standard deviation from the feature values included in the colony region on the texture feature image. A state assessment processing unit (245) assesses whether there are undifferentiated aberrant cells in the colony by machine learning in which multiple types of index values are used. This makes it possible to accurately and non-invasively assess whether cells, or cells in a colony, are in an undifferentiated aberrant state.

Description

Cell state determination method and cell state determination device

The present invention relates to a cell state determination method and a cell state determination apparatus for non-invasively determining a cell state in a process of culturing pluripotent stem cells (ES cells and iPS cells), and more specifically, for observing cells. The present invention relates to a cell state determination method and a cell state determination device suitable for determining whether a pluripotent stem cell is in an undifferentiated state or an undifferentiated deviation state based on the observed image.

In the field of regenerative medicine, research using pluripotent stem cells such as iPS cells and ES cells has been actively conducted in recent years. In the research and development of regenerative medicine using such pluripotent stem cells, it is necessary to culture a large amount of undifferentiated cells that maintain pluripotency. For this reason, it is necessary to select an appropriate culture environment and to stably control the environment, and it is necessary to check the state of cells in culture at a high frequency. For example, if cells in a cell colony deviate from an undifferentiated state, in this case, all the cells in the cell colony have the ability to differentiate, so eventually all the cells in the colony are undifferentiated. Transition to a deviating state. Therefore, the observer checks every day whether cells that have deviated from the undifferentiated state (cells that have already differentiated or are likely to differentiate, hereinafter referred to as “undifferentiated deviant cells”) have occurred in the cultured cells, If undifferentiated cells are found, they need to be removed quickly.

Whether or not a pluripotent stem cell maintains an undifferentiated state can be reliably determined by staining with an undifferentiated marker. However, since the stained cells are killed, undifferentiated marker staining cannot be performed for determination of pluripotent stem cells for regenerative medicine. In view of this, in the current field of cell culture for regenerative medicine, it is determined whether or not the observer is an undifferentiated cell based on the morphological observation of the cell using a phase contrast microscope. The phase contrast microscope is used because cells are generally transparent and difficult to observe with a normal optical microscope. However, skill is required to make an accurate determination using such a method. In addition, since it is based on human judgment, it is inevitable that the judgment will vary. Therefore, such conventional methods are not suitable for industrially mass-producing pluripotent stem cells.

In order to solve such a problem, a technique for evaluating the state of a cell by performing image processing on an observation image of the cell has been conventionally proposed.
For example, in Patent Document 1, a texture feature amount of a cell internal structure is calculated from a plurality of cell observation images acquired at predetermined time intervals, and a difference or correlation value of the texture feature amount with respect to the plurality of cell observation images is calculated. A method for discriminating cell activity based on the time series change is described. In this method, for example, when the difference value of the texture feature amount with time elapses, it can be determined that the activity of the cell is decreasing.

Further, Patent Document 2 describes a method of predicting cell quality such as a proliferation rate by performing fuzzy neural network (FNN) analysis using a plurality of index values acquired from a cell observation image. This document also describes that a texture feature amount obtained by image processing on a cell observation image is used as an index value.

The texture feature amount used in the above-described conventional cell evaluation method is calculated for each cell colony that is an aggregate of cells in the cell observation image, or calculated for each specific region in the cell observation image. It has been done. However, when the texture feature amount is calculated for a certain area in the image, even if a pattern different from the other part appears only in a part of the area, it is difficult to appear in the calculated value.

As described above, when pluripotent stem cells are cultured in large quantities, if it is found that some of the colonies are undifferentiated cells, it is necessary to remove them or to stop the culture itself . For that purpose, it is required to recognize an undifferentiated departure cell as a cell in a defective state and to detect an event that a part of the colony is in a defective state in an observation image of a stem cell colony. However, if the region in the poor state, that is, the region of undifferentiated departure cells is a small part of the region for calculating the texture feature amount, it is difficult to detect the defective state from the calculated value of the feature amount. is there.

In addition, in the above-described conventional method, morphological information of cells or colonies may be used when calculating feature quantities for cell evaluation. In order to obtain cell shape information, it is first necessary to accurately recognize the cell region in the cell observation image. If the cell region is not correctly recognized (for example, two cell regions are actually converted into one cell). The feature quantity based on the form information becomes inaccurate (for example, when it is erroneously recognized as an area), and may cause a large misjudgment. Further, when colonies grow, adjacent colonies may be combined, but it is difficult to treat such a combined colony as a single colony. For this reason, it is necessary to perform image processing after separating the combined colonies into single colonies, but inaccuracy of such separation processing and changes in form information due to separation can also be a cause of erroneous determination.

Japanese Patent No. 4968595 Japanese Patent No. 5181385 International Patent Publication No. 2016/084420 JP-A-10-268740 International Patent Publication No. 2013/097772

The present invention has been made in order to solve the above-described problems, and the object of the present invention is to perform non-invasive and accurate determination of a cell state such as whether or not an undifferentiated departure cell exists in a part of a stem cell colony. It is providing the cell state determination method and cell state determination apparatus which can be implemented in this.

The cell state determination method according to the present invention made to solve the above problems is a cell state for determining the state of a cell or a cell in a colony based on an observation image of a cell or a colony that is an aggregate of cells. A determination method comprising:
a) calculating a texture feature amount by performing texture analysis for each of a plurality of first small regions set on the observed image, and creating a texture feature image indicating a two-dimensional distribution of the texture feature amount Texture feature image creation step;
b) For one or a plurality of second small regions on the texture feature image, a plurality of types of index values related to the feature amount distribution in the small regions are calculated based on the texture feature amounts respectively included in the second small regions. An index value calculating step,
c) a determination step of determining a state of a cell included in the second small area based on a plurality of types of index values calculated for each second small area in the index value calculating step;
It is characterized by carrying out.

Moreover, the cell state determination device according to the present invention made to solve the above-mentioned problems is a device for carrying out the above-described cell state determination method, and is based on an observation image of a colony that is a cell or an aggregate of cells. A cell state determination device for determining the state of a cell or a cell in a colony,
a) calculating a texture feature amount by performing texture analysis for each of a plurality of first small regions set on the observed image, and creating a texture feature image indicating a two-dimensional distribution of the texture feature amount A texture feature image creation unit;
b) For one or a plurality of second small regions on the texture feature image, a plurality of types of index values related to the feature amount distribution in the small regions are calculated based on the texture feature amounts respectively included in the second small regions. An index value calculation unit to
c) a determination execution unit that determines a state of a cell included in the second small region based on a plurality of types of index values calculated for each second small region by the index value calculation unit;
It is characterized by having.

In the present invention, cells to be determined are typically pluripotent stem cells such as iPS cells and ES cells. In addition, the determination of the cell state performed in the determination step is typically performed by determining whether a single cell or a cell in a colony is an undifferentiated cell, an undifferentiated departure cell, or a differentiated cell. is there.

In the cell state determination method embodied by the cell state determination apparatus according to the present invention, in the texture feature image creation step, texture analysis is performed for each of the plurality of first small regions on the observation image of the cell or colony. A texture feature amount is calculated. Then, a texture feature image showing a two-dimensional distribution of texture feature quantities is created. Since the size of the first small region is desirably a size that can easily detect a change in the acquired pattern information, it may be appropriately determined in advance according to the size of the cell to be measured.

As is well known, texture analysis is roughly divided into structural texture analysis and statistical texture analysis. In general, the former is suitable for extracting structural or regular features of the pattern or pattern of the entire image, and the latter is suitable for extracting local features of the pattern or pattern in the image. Here, since it is desired to grasp the characteristics in a narrow range in the image corresponding to some cells in the colony, statistical texture analysis is suitable as texture analysis. Note that general descriptions of image texture analysis, texture characteristics, and the like are described in detail in Non-Patent Document 1 and the like.

There are various methods for statistical texture analysis, such as a density histogram method using a density histogram, a density level difference method using a difference statistic, a spatial density level dependency method using a co-occurrence matrix, and a method using a density co-occurrence matrix. . For this reason, an appropriate method may be selected. However, according to an experimental study by the present inventor, it has been confirmed that a sufficiently reliable determination can be made by a method using a concentration co-occurrence matrix. Non-Patent Document 2 proposes 14 types of texture feature amounts in texture analysis using a density co-occurrence matrix, and the properties represented by the feature amounts proposed in Reference 2 are described in Non-Patent Document 3. Explained. In this way, by using not only one type but also a plurality of types of texture feature amounts, it is possible to improve the accuracy of determination.

Next, in the index value calculation step, the small region for one second small region or each of the plurality of second small regions set in one or a plurality of texture feature images created in the texture feature image creation step A plurality of predetermined index values are calculated based on a plurality of texture feature amounts included in the. For example, when determining the state of the cells in the colony, the entire colony may be used as the second small area, or a plurality of second small areas having a size that includes one or more appropriate numbers of cells. An area may be defined. That is, the second region can be a cell colony or a region corresponding to one or a plurality of cells. Further, as the index value, statistics such as an average value, standard deviation, variance, maximum value, minimum value, median value, and mode value of a plurality of feature amounts in the second small region can be used.

In the determination step, based on a plurality of types of index values for one or a plurality of texture feature amounts calculated as described above, the state of the cells included in the second small region from which the index value is calculated, For example, it is determined whether an undifferentiated departure cell or a differentiated cell is contained in the colony. As a simple determination method, the state of the cell may be determined by comparing a plurality of types of index values with predetermined threshold values. Further, in order to make a more accurate determination, the state of the cell may be determined by machine learning based on a plurality of types of index values. Further, various multivariate analysis methods other than machine learning may be used.

For example, in human iPS cells, a colony is formed densely in an undifferentiated state, but often changes thinly when deviating from the undifferentiated state. Therefore, the size of the first small region is set to about the size of the thinly spread cell, a texture feature image showing the distribution of the texture feature amount for each small region is created, and the index value is obtained from the texture feature image. When it is calculated, the presence of thinly spread cells is easily reflected in the index value. As a result, it is possible to detect a region where undifferentiated departure cells exist with high accuracy.

In the present invention, the observation image may be an observation image obtained by a general optical microscope. However, since the cells are usually transparent, it is difficult to clearly observe the boundary between the cells and the background with a general optical microscope. . Therefore, a phase image obtained with a phase contrast microscope may be used as the observation image. In recent years, digital holographic microscopes using digital holography technology have been put into practical use, and phase images are obtained by performing data processing such as phase recovery and image reconstruction on hologram data obtained with digital holographic microscopes. Has been established (see Patent Documents 3 and 4). Usually, in such a phase image, cell pattern information appears more clearly than a light intensity image, a visible image or the like, and therefore a phase image is suitable as the observation image in the present invention.

In addition, since it is necessary to perform focusing when taking a microscopic image in a general phase contrast microscope, it takes time for measurement when acquiring microscopic images of each small area obtained by finely dividing a wide observation target area. It takes. In contrast, in a digital holographic microscope, phase information at an arbitrary distance can be calculated at the stage of arithmetic processing after acquiring a hologram, so that it is not necessary to perform focusing every time during shooting, and the measurement time is shortened. be able to. For these reasons, it is more preferable to use an image created by processing based on hologram data acquired by a digital holographic microscope as the observation image in the present invention.

Therefore, the cell state determination apparatus according to the present invention preferably further includes a digital holographic microscope and a phase image creation unit that creates a phase image by processing based on hologram data acquired by the digital holographic microscope. The phase image may be provided as the observation image to the texture feature image creation unit.

According to the present invention, for example, in a site where pluripotent stem cells such as iPS cells and ES cells are cultured, whether the cells being cultured are in an undifferentiated state or undifferentiated state, or in a colony It is possible to determine whether or not there are undifferentiated cells in a non-invasive manner with high accuracy. Thereby, the quality control of the cells in culture becomes easy, and the productivity in cell culture can be improved.

The schematic block diagram of the cell state determination apparatus by one Example of this invention. The flowchart which shows the procedure of the cell state determination process in the apparatus of a present Example. Explanatory drawing of the texture analysis by the density | concentration co-occurrence matrix used with the apparatus of a present Example. The figure which shows the kind of texture feature-value calculated by the texture analysis in the apparatus of a present Example, and the property of the image estimated that it represents. The figure which shows the comparative example of the various texture characteristic images calculated | required from the phase image obtained by an in-line type holographic microscope, and this phase image. The figure which shows the example of calculation of the index value about the colony area | region in a texture characteristic image. Explanatory drawing of the index value calculation process for every small area | region on a texture characteristic image.

Hereinafter, an embodiment of a cell state determination method and a cell state determination device according to the present invention will be described with reference to the accompanying drawings.
FIG. 1 is a schematic configuration diagram of an embodiment of a cell state determination apparatus for carrying out the cell state determination method according to the present invention.

The cell state determination apparatus of the present embodiment includes a microscopic observation unit 1, a control / processing unit 2, an input unit 3 and a display unit 4 which are user interfaces.
The microscopic observation unit 1 is an in-line holographic microscope (IHM), and includes a light source unit 10 including a laser diode or the like and an image sensor 11, and between the light source unit 10 and the image sensor 11, A culture plate 12 including a colony (or a single cell) 13 to be determined is arranged. The control / processing unit 2 controls the operation of the microscopic observation unit 1 and processes data acquired by the microscopic observation unit 1, and includes an imaging control unit 20, a data storage unit 21, and a phase information calculation unit 22. And a phase image creation unit 23 and a cell state evaluation unit 24 are provided as functional blocks. In addition, the cell state evaluation unit 24 includes a texture feature amount calculation unit 241, a texture feature image creation unit 242, a cell region extraction unit 243, an index value calculation unit 244, and a state determination processing unit 245 as lower functional blocks. Include as.

The entity of the control / processing unit 2 is a personal computer or a higher-performance workstation, and the functions of the above-described functional blocks are realized by operating dedicated control / processing software installed in the computer on the computer. Can be configured. Further, the function of the control / processing unit 2 may be shared by a plurality of computers connected via a communication network, as will be described later, instead of a single computer.

First, in the cell state determination apparatus of the present embodiment, operations and processes until a phase image that is an observation image used when determining a cell state will be described. In the following example, it is determined whether undifferentiated departure cells or differentiated cells are present in the colony.
When the operator sets the culture plate 12 including the colony 13 at a predetermined position and performs a predetermined operation with the input unit 3, the imaging control unit 20 controls the microscopic observation unit 1 to acquire hologram data as follows.

In other words, the light source unit 10 irradiates a predetermined region of the culture plate 12 with coherent light having a small angle spread of about 10 °. The coherent light (object light 15) that has passed through the culture plate 12 and the colony 13 reaches the image sensor 11 while interfering with the light (reference light 14) that has passed through a region close to the colony 13 on the culture plate 12. The object light 15 is light whose phase has changed when passing through the colony 13. On the other hand, the reference light 14 is light which does not pass through the colony 13 and thus does not undergo phase change due to the colony 13. Accordingly, on the detection surface (image surface) of the image sensor 11, an interference image (hologram) between the object light 15 whose phase has been changed by the colony 13 and the reference light 14 whose phase has not changed is formed.

The light source unit 10 and the image sensor 11 are sequentially moved in the X axis-Y axis direction (in a plane perpendicular to the paper surface of FIG. 1) by a moving mechanism (not shown). Thereby, the irradiation area | region (observation area | region) of the coherent light emitted from the light source part 10 can be moved on the culture plate 12, and the hologram over a wide two-dimensional area | region can be acquired.

As described above, hologram data obtained by the microscopic observation unit 1 (two-dimensional light intensity distribution data of a hologram formed on the detection surface of the image sensor 11) is sequentially sent to the control / processing unit 2 for data storage. Stored in the unit 21. In the control / processing unit 2, the phase information calculation unit 22 reads the hologram data from the data storage unit 21, and calculates phase information for the entire observation region by executing a predetermined calculation process for phase recovery. The phase image creation unit 23 creates a phase image of the entire observation region based on the calculated phase information.

In addition, when calculating such phase information or creating a phase image, a known algorithm disclosed in

Patent Documents

3 and 4 may be used. In addition, intensity information, pseudo phase information, and the like may be calculated based on the hologram data, and the phase image creation unit 23 may create a reproduced image based on these. In the following description, the cell state is determined using the phase image, but the cell state may be determined using the intensity image, the pseudo phase image, or the like.

When a phase image in which a colony that is a determination target is reflected as an image is obtained as described above, the cell state evaluation unit 24 executes a cell state determination process according to the procedure of the flowchart shown in FIG.

First, the texture feature quantity calculation unit 241 obtains data constituting the phase image created by the phase image creation unit 23, that is, intensity value data for each pixel of the phase image (step S1). Then, a large number of first small areas obtained by dividing the phase image composed of the intensity value data, for example, for each predetermined size are set, and predetermined statistical texture analysis is executed for each first small area. As a result of the texture analysis, a plurality of types of texture feature amounts are obtained for each first small area. Then, the texture feature image creation unit 242 creates a texture feature image indicating a two-dimensional distribution of the texture feature value for each type of the texture feature value (step S2).

The first small area may be a first small area by grouping a plurality of adjacent pixels. In this case, the texture feature amount of the reference pixel is set as the texture feature amount of the pixel. The reference pixel may be set at the center of the first small area. The size of the first small region is desirably a size that is easy to detect, such as a change in the pattern of cells or cell colonies on the phase image, and classifies undifferentiated cells, undifferentiated cells to be described later, etc. according to the size. The accuracy of is affected. Therefore, the size of the first small region may be appropriately determined in advance according to the type of cell to be measured, the cell, the size of the cell colony, and the like. Further, the phase image to be processed in step S1 does not have to be an image of the entire observation area, and may be an image limited to a range designated by the user (worker), for example.

As is well known, there are various methods for statistical texture analysis. In the apparatus of this embodiment, texture analysis using a density co-occurrence matrix widely used for image analysis is used. The density co-occurrence matrix is a typical texture analysis method proposed in Non-Patent Document 2 by Haralick et al. FIG. 3 is a conceptual diagram for explaining the density co-occurrence matrix.

Now, in an image f (x, y) having pixel position information on two axes perpendicular to x and y, two pixels P ₁ (x ₁ , y ₁ ) whose relative positional relationship is (d, θ). ), P ₂ (x ₂ , y ₂ ). The density co-occurrence matrix P (i, j; d, θ) represents the frequency at which the density becomes (i, j) in the pixel pair of the pixels P ₁ and P ₂ . The distance d between the pixels P ₁ and P ₂ is d = max (| x ₁ −x ₂ |, | y ₁ −y ₂ |), and a straight line passing through the two pixels P ₁ and P ₂ and a horizontal line are formed. The angle θ is four directions of 0 °, 45 °, 90 °, and 135 °. Therefore, when the number of gradations of the image is g, the density co-occurrence matrix P (i, j; d, θ) is a matrix having a size of g × g in one direction.

In Non-Patent Document 2, as shown in the following equation (1), elements in all directions of the concentration co-occurrence matrix P (i, j; d, θ) are added to obtain P (i, j), and ( As shown in the equation (2), the elements of P (i, j) are normalized and expressed.

In this document, Entropy, Angular Second Moment, Contrast, Correlation, Sum of Squares, Inverse Difference Moment, Sum Average, Sum Variance, Sum Entropy, as texture feature quantities using this density co-occurrence matrix P (i, j) 14 types of Difference Variance, Difference Entropy, 2 types of Information Measures of Correlation, Maximal Correlation Coefficient have been proposed. In the apparatus of this embodiment, 6 types (Entropy, Angular Second Moment, Contrast, Correlation, Sum of, Squares and Inverse Difference Moment) are selected and used. FIG. 4 shows the six types of texture feature quantities used in the apparatus of this embodiment, their calculation formulas, and the properties of the image estimated to represent the feature quantities. The nature of this image is described in Non-Patent Document 3.

It should be noted that the above six types of texture feature values are representatively selected from those having relatively different image properties that are estimated to be represented, and are not necessarily optimal combinations. Therefore, an appropriate type of texture feature amount may be added or reduced.

FIG. 5 shows IHM phase images of colonies containing only undifferentiated cells (undifferentiated colonies) and colonies including undifferentiated deviation cells (undifferentiated deviation colonies), and three types of textures created based on the phase images. It is a comparative example of a feature image. Although the undifferentiated departure area can be visually recognized even in the IHM phase image shown at the left end, the clarity is not necessarily high. On the other hand, in the three types of texture feature images on the right side, the difference between the undifferentiated departure area and the surrounding undifferentiated areas is clear, and the undifferentiated departure area is very easy to visually recognize.

In the process of step S2, texture feature images are obtained for each of the six types of texture feature amounts, so that six texture feature images are obtained from one phase image. Next, the cell region extraction unit 243 extracts a region where cells or colonies are estimated to exist based on the texture feature image, and obtains data indicating the outline of the cell region or colony region (step S3). For this region extraction processing, for example, a well-known image processing algorithm described in Patent Document 5 can be used. When the outline of the cell or colony region is obtained, a mask for limiting the small region for calculating the index value on the texture feature image is generated.

The index value calculation unit 244 first selects a colony region (or cell region or the like) by applying the mask created in step S3 to the six types of texture feature images. Then, based on the texture feature value for each small region included in the colony region in each texture feature image, a plurality of index values that are statistics are calculated (step S4). Generally, the average value, standard deviation, variance, maximum value, minimum value, median value, mode value, etc. are known as statistics, but in the apparatus of this embodiment, the average value, standard deviation, maximum value are known. The four types of minimum values are used as index values. Since there are six types of texture feature images and four types of index values, a total of 6 × 4 = 24 index values are obtained for one colony region.

FIG. 6 shows an example of calculating the index value for the colony region in the texture feature image. This is an example of an undifferentiated departure colony, and FIG. 6 selectively shows only three types of texture feature images. As shown in FIG. 4, the different types of texture feature amounts reflect different properties of the images, and therefore, the four types of index values tend to be different from the texture feature amounts.

Next, the state determination processing unit 245 determines whether or not there is an undifferentiated departure area in the colony region based on the plurality of index values, that is, whether or not there are undifferentiated departure cells in the colony region. The apparatus of this embodiment uses supervised machine learning for the determination, specifically, a support vector machine (SVM), which is a typical two-class classification method as a machine learning method (step S5). In other words, a learning model was constructed using index value data acquired based on phase images prepared in advance for each of undifferentiated colonies and undifferentiated deviation colonies, and index value data acquired based on unknown phase images was input. As a result, the classification result to an undifferentiated colony or an undifferentiated departure colony is output. The state determination processing unit 245 displays such a determination result on the screen of the display unit 4.

Of course, the method for determining the state of a cell based on a plurality of index values is not limited to machine learning. As a simple method, whether or not undifferentiated deviation cells (or differentiated cells) are included in a colony depending on a combination of conditions in which a threshold value is set in advance for a plurality of index values and the index value is greater than or less than the threshold value. It may be determined. Further, a method other than the support vector machine may be used as the machine learning, or a multivariate analysis method that is not included in the machine learning in a narrow sense may be used.

In the above example, the index value is calculated only for one colony region determined by the mask created in step S3. However, as will be described later, a colony, a cell, or a specific part in the cell is displayed on the texture feature image. Set appropriate subregions for cells such as cell nuclei, etc., calculate index values for each of the subregions, and independently determine the cell state for each subregion Also good. As a result, it is possible to identify at which position the undifferentiated departure cell exists in a state where a large number of cells exist or in a colony.

[Experimental Example 1]
In order to confirm the effectiveness of the method according to the above-described procedure, the present inventors conducted an experiment for discriminating undifferentiated colonies from undifferentiated colonies. Specifically, when 66 colonies of human iPS cells cultured on feeder cells (SNL) were analyzed by the above method, an accuracy rate of 84.8% was obtained by cross-validation evaluation.

[Experiment 2]
The presence of undifferentiated cells was also determined for human iPS cells cultured by a culture method that does not use feeder cells. However, in this experiment, in order to determine in which part of the colony region the undifferentiated departure cells exist, a rectangular small region is gridded on the IHM phase image for the colony as shown in FIG. The index value was calculated for each small region including a part of the colony region. The size of one small region is a size that includes about a dozen cells. The determination of the cell state based on the index value is divided into two types: one in which undifferentiated cells occupy and one in which undifferentiated deviation cells are mixed (as much as 10% or more in actual area is estimated). Classify into classes. Was also implemented. In this experiment, as in the case of the culture on the feeder cells shown in Experimental Example 1, the evaluation by cross-validation gave a result of 80% or more in all of the accuracy rate, sensitivity, and specificity. This confirmed the effectiveness of the method.

In the configuration of the embodiment shown in FIG. 1, all processing is performed in the control / processing unit 2. However, in general, calculation of phase information based on hologram data and imaging of the calculation result are enormous. A quantity calculation is required. For this reason, a normally used personal computer takes a long time for calculation, and an efficient analysis work is difficult. Therefore, a personal computer connected to the microscopic observation unit 1 may be used as a terminal device, and a computer system in which this terminal device and a server that is a high-performance computer are connected via a communication network such as the Internet or an intranet may be used. In this case, complicated processing such as calculation of phase information based on hologram data and creation of a phase image is performed on the server side, and the terminal device receives the image data created thereby, and the phase configured by this image data The processing on the image, that is, the processing by the cell state evaluation unit 24 described above may be performed on the terminal device side. In such a configuration, the functional blocks of the control / processing unit 2 shown in FIG. 1 are separated into the terminal device side and the server side. As described above, the functions of the control / processing unit 2 may be shared by a plurality of computers.

In the cell state determination apparatus of the above embodiment, an in-line holographic microscope is used as the microscopic observation unit 1, but other microscopes such as an off-axis (off-axis) type and a phase shift type may be used as long as the microscope can obtain a hologram. Naturally, it can be replaced with a holographic microscope of the above type. Moreover, you may implement the process as shown in FIG. 2 based on the phase image obtained with a general phase contrast microscope instead of the phase image obtained with a holographic microscope. Here, the type of observation image is not limited as long as the target cell or colony is an image that can be observed relatively clearly.

Further, the above-described embodiments and the above-described various modifications are examples of the present invention, and it is apparent that even if appropriate modifications, corrections, and additions are made within the scope of the present invention, they are included in the scope of the claims of the present application. is there.

DESCRIPTION OF SYMBOLS 1 ... Microscopic observation part 10 ... Light source part 11 ... Image sensor 12 ... Culture plate 13 ... Colony 14 ... Reference light 15 ... Object light 2 ... Control and processing part 20 ... Imaging control part 21 ... Data storage part 22 ... Phase information calculation part 23 ... Phase image creation unit 24 ... Cell state evaluation unit 241 ... Texture feature amount calculation unit 242 ... Texture feature image creation unit 243 ... Cell region extraction unit 244 ... Index value calculation unit 245 ... State determination processing unit 3 ... Input unit 4 ... Display section

Claims

A cell state determination method for determining a state of a cell or a cell in a colony based on an observation image about a colony which is a cell or a collection of cells,
a) calculating a texture feature amount by performing texture analysis for each of a plurality of first small regions set on the observed image, and creating a texture feature image indicating a two-dimensional distribution of the texture feature amount Texture feature image creation step;
b) For one or a plurality of second small regions on the texture feature image, a plurality of types of index values related to the feature amount distribution in the small regions are calculated based on the texture feature amounts respectively included in the second small regions. An index value calculating step,
c) a determination step of determining a state of a cell included in the second small area based on a plurality of types of index values calculated for each second small area in the index value calculating step;
The cell state determination method characterized by implementing.
The cell state determination method according to claim 1,
The cell state determination method, wherein the second region is a cell colony or a region corresponding to one cell.
The cell state determination method according to claim 1,
The texture analysis is statistical texture analysis, and the type of texture feature used is one type or a plurality of types.
The cell state determination method according to claim 3,
The cell state determination method, wherein the texture analysis is texture analysis using a density co-occurrence matrix.
The cell state determination method according to claim 1,
The cell state determination method, wherein the observation image is an image created by a process based on hologram data acquired by a digital holographic microscope.
The cell state determination method according to claim 1,
The cell to be measured is a pluripotent stem cell including a human iPS cell, and in the determination step, it is determined whether the cell is an undifferentiated cell, an undifferentiated departure cell, or a differentiated cell. Cell state determination method.
The cell state determination method according to claim 1,
In the determination step, the cell state determination method includes determining a cell state by machine learning based on the plurality of types of index values.
The cell state determination method according to claim 1,
In the determination step, the cell state is determined by comparing each of the plurality of types of index values with a predetermined threshold value.
A cell state determination device for determining the state of a cell or a cell in a colony based on an observation image of a cell or a colony that is an aggregate of cells,
a) calculating a texture feature amount by performing texture analysis for each of a plurality of first small regions set on the observed image, and creating a texture feature image indicating a two-dimensional distribution of the texture feature amount A texture feature image creation unit;
b) For one or a plurality of second small regions on the texture feature image, a plurality of types of index values related to the feature amount distribution in the small regions are calculated based on the texture feature amounts respectively included in the second small regions. An index value calculation unit to
c) a determination execution unit that determines a state of a cell included in the second small region based on a plurality of types of index values calculated for each second small region by the index value calculation unit;
A cell state determination device comprising:
The cell state determination device according to claim 9,
A digital holographic microscope,
A phase image creating unit that creates a phase image by processing based on hologram data acquired by the digital holographic microscope;
The cell state determination apparatus, further comprising: providing the phase image as the observation image to the texture feature image creation unit.