JP2017026482A - Data processor, determination tree generation method, identification device, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data processor for generating an identification rule for enabling a doctor or the like to easily understand a basis of a determination result which can be obtained from the featured value of a cell image.SOLUTION: A data processor includes: an input part; and a determination tree generation part. The input part inputs learning data consisting of a label given to a cell image and a plurality of featured values extracted from the cell image as one set. The determination tree generation part generates a determination tree for identifying information which is equivalent to the label of a sample on the basis of the learning data.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a data processing device, a decision tree generation method, an identification device, and a program. In particular, the present invention relates to a data processing apparatus, a decision tree generation method, an identification apparatus, and a program that process a feature vector generated from a cell image.

  Pathological diagnosis is performed in which a part of an organ is collected from a patient as a specimen, and a cross section obtained by thinly cutting the specimen is observed with a microscope. For example, a doctor may confirm a pathological image (liver pathological image) obtained by collecting hepatocytes from a patient suspected of having liver cancer and photographing the hepatocytes, and determine the malignancy (grade) of cancer. Has been done.

  However, the amount of work for the pathological diagnosis itself as described above is enormous, which places a heavy burden on the doctor. For this reason, technical developments such as image processing technology and information processing technology for reducing the burden on doctors have been carried out.

  For example, in Patent Document 1, by extracting a sub-image centered on cell nuclei, vacancies, cytoplasm, stroma and the like from a pathological image, simultaneously extracting color information of cell nuclei and storing both as feature candidates, A technique for determining the presence or absence of a tumor and the benign / malignant tumor with higher accuracy is disclosed. Further, Non-Patent Document 1 discloses details regarding a decision tree generation method.

JP 2006-153742 A

Roman Timofeev, “Classification and Regression Trees (CART) Theory and Applications”, December 20, 2004, [online], [Search June 26, 2015], Internet <URL: http: //edoc.hu- berlin.de/master/timofeev-roman-2004-12-20/PDF/timofeev.pdf>

  Each disclosure of the above prior art document is incorporated herein by reference. The following analysis was made by the present inventors.

  For example, as disclosed in Patent Document 1, there is a pathological image analysis technique that supports pathological diagnosis by a doctor. However, the above technique is limited to the determination of whether or not the patient is afflicted with cancer (determination of cancer or non-cancer), and the reason why the above determination is made, and the basis for actively disclosing the reason is not. That is, a doctor or the like cannot easily understand what reason and process the determination result is calculated from various features appearing in the pathological image (cell image). When the relationship between the characteristics appearing in the cell image and the judgment result is difficult for a doctor or the like to understand, it is difficult to understand how much the judgment result can be trusted, and to reduce the burden on the doctor etc. It is difficult to say that the technology is fully utilized.

  An object of the present invention is to provide a data processing device, a decision tree generation method, an identification device, and a program for generating an identification rule that can easily understand the basis of a determination result obtained from a feature amount of a cell image.

  According to a first aspect of the present invention, an input unit for inputting learning data including a label given to a cell image and a plurality of feature amounts extracted from the cell image as a set, and the sample of the sample There is provided a data processing apparatus comprising: a decision tree generation unit that generates a decision tree for identifying information corresponding to a label based on the learning data.

  According to a second aspect of the present invention, a step of inputting learning data including a label given to a cell image and a plurality of feature amounts extracted from the cell image as a set, and the label of the sample Generating a decision tree for identifying information corresponding to the learning data based on the learning data.

  According to a third aspect of the present invention, there is provided an identification device for identifying a sample using the decision tree generated by the above-described decision tree generation method.

According to a fourth aspect of the present invention, a process of inputting learning data including a label given to a cell image and a plurality of feature amounts extracted from the cell image, and the label of the sample There is provided a program for causing a computer mounted on a data processing apparatus to execute a process for generating a decision tree for identifying information corresponding to the above-described learning data based on the learning data.
This program can be recorded on a computer-readable storage medium. The storage medium may be non-transient such as a semiconductor memory, a hard disk, a magnetic recording medium, an optical recording medium, or the like. The present invention can also be embodied as a computer program product.

  According to each aspect of the present invention, there is provided a data processing device, a decision tree generation method, an identification device, and a program that contribute to generating an identification rule that can easily understand the basis of a determination result obtained from a feature amount of a cell image. Provided.

It is a figure for demonstrating the outline | summary of one Embodiment. It is a figure which shows an example of a structure of the pathological image processing system which concerns on 1st Embodiment. It is a figure which shows an example of an internal structure of a learning data generation apparatus. It is a figure which shows an example of gaze area | region image data, and an example of label information. It is a figure for demonstrating the feature-value which a feature-value vector production | generation part produces | generates. It is a figure which shows an example of a cell nucleus area | region. It is a figure for demonstrating the statistical process with respect to the feature-value by a feature-value vector production | generation part. It is a figure which shows an example of the learning data which a learning data generation apparatus produces | generates. It is a figure which shows an example of an internal structure of a data processor. It is a figure which shows an example of the 1st selection policy which a feature-value selection part refers. It is a figure which shows an example of the learning data of the result of performing the 1st selection process. It is a figure which shows an example of the 2nd selection policy which a feature-value selection part refers. It is a flowchart which shows an example of the 2nd selection process of a feature-value selection part. It is a figure which shows an example of the decision tree which a feature-value selection part produces | generates. It is a figure for demonstrating the quality of a feature-value. It is a figure for demonstrating the narrowing-down of the feature-value by a feature-value selection part. It is a figure which shows an example of the learning data of the result of performing the 2nd selection process. It is a flowchart which shows an example of operation | movement of a data processor. It is a figure which shows an example of the learning data used for the production | generation of a decision tree. It is a figure which shows an example of the decision tree obtained from the learning data shown in FIG. It is a figure which shows an example of the grading result by a decision tree. It is a figure which shows an example of a structure of the pathological image processing system which concerns on 2nd Embodiment. It is a figure which shows an example of accompanying information. It is a figure which shows an example of the internal structure of the data processor which concerns on 2nd Embodiment. It is a figure which shows an example of the classification result by a decision tree. It is a figure which shows an example of the analysis result by the analysis part of 2nd Embodiment. It is a figure for demonstrating the effectiveness of the anticancer agent for every classification result by a decision tree. It is a figure which shows another example of accompanying information. It is the figure which plotted the analysis result of the analysis part by 2nd Embodiment. It is a figure which shows an example of another internal structure of a data processor. It is a figure which shows an example of label information in the case of using the cancer recurrence information of the patient corresponding to gaze area ID as a label.

  First, an outline of one embodiment will be described. Note that the reference numerals of the drawings attached to the outline are attached to the respective elements for convenience as an example for facilitating understanding, and the description of the outline is not intended to be any limitation.

  The data processing apparatus 100 according to an embodiment includes an input unit 101 and a decision tree generation unit 102. The input unit 101 inputs learning data including a label given to the cell image and a plurality of feature amounts extracted from the cell image as a set. The decision tree generation unit 102 generates a decision tree for identifying information corresponding to the sample label based on the learning data.

  The data processing apparatus 100 accepts a plurality of feature amounts (feature amount vectors) that add feature amounts to cell images. The data processing apparatus 100 generates a decision tree for identifying a label (for example, cancer cell grade) given to the cell image using the feature amount. The decision tree has a tree structure in which leaves represent classifications (class labels) and branches indicate the basis for reaching the classification. Therefore, by generating an identification rule used for grading a cell image or the like using a decision tree, a doctor or the like can easily understand the basis of the determination result and the prediction result based on the identification rule.

  Hereinafter, specific embodiments will be described in more detail with reference to the drawings. In addition, in each embodiment, the same code | symbol is attached | subjected to the same component and the description is abbreviate | omitted.

[First Embodiment]
The first embodiment will be described in more detail with reference to the drawings.

  FIG. 2 is a diagram illustrating an example of the configuration of the pathological image processing system according to the first embodiment. Referring to FIG. 2, the pathological image processing system includes a learning data generation device 10, a data processing device 20, and an identification device 30.

  In the first embodiment, a cell image acquired from cells collected from a patient's liver will be described as a system target. However, it is not intended to limit the cells and organs, but of course, cells collected from other organs may be used.

  The learning data generation device 10 includes label data including image data (hereinafter referred to as gaze area image data) related to a gaze area (ROI) extracted from a cell image, and a grade corresponding to the gaze area. And enter.

  An image obtained by imaging a part of a cell image acquired by a doctor or the like with a CCD (Charge Coupled Device) camera mounted on a microscope is gaze area image data.

  The grade of the gaze area image data included in the label information (label given to the gaze area image data) is checked by the doctor, and based on the knowledge of the doctor, the grade 0 (G0) to the grade 4 (G4 ). In the first embodiment, the case where the grade is an integer value will be described as an example. However, the grade is not necessarily an integer, and may be grade 2.5, for example. In this case, the classification can be executed in the same procedure as in the case of an integer by using a regression decision tree. Alternatively, when the grade by the doctor is not an integer, the learning data generation device 10 may change the grade to an integer by processing such as rounding up, rounding down, or rounding off after the decimal point.

  The learning data generation device 10 inputs a plurality of gaze area image data and label information corresponding to each of the plurality of gaze area image data.

  The learning data generation device 10 generates learning data based on the gaze area image data and the label information, and outputs the learning data to the data processing device 20.

  The data processing device 20 generates an identification rule (identification model, identification rule or identification function) for grading (rating) hepatocytes based on the input learning data. More specifically, the data processing device 20 generates a decision tree based on the input learning data. The decision tree (identification rule) generated by the data processing device 20 is provided to the identification device 30.

  The identification device 30 inputs a feature amount (feature amount vector) of a sample that has not been graded. The identification device 30 uses the decision tree provided from the data processing device 20 as a prediction model and outputs a response (class label attached to the leaf of the decision tree) to the input feature amount (outputs the identification result). That is, the data processing device 20 generates a decision tree for identifying the grade of the sample (information corresponding to the label) based on the learning data. Further, the identification device 30 performs grading of sample data using the decision tree generated by the data processing device 20.

  FIG. 3 is a diagram illustrating an example of an internal configuration of the learning data generation device 10. Referring to FIG. 3, the learning data generation apparatus 10 includes an input unit 11, a feature vector generation unit 12, a learning data output unit 13, and a storage unit 14 including an HDD (Hard Disk Drive). Note that illustrations of operation devices (keyboard, mouse, etc.) and display devices for operating the learning data generation apparatus 10 are omitted. Each unit including the input unit 11 is configured to access the storage unit 14 and write and read data.

  The input unit 11 is means for inputting the above-described gaze area image data and label information. Each gaze area image data is provided with an identifier (ID), and the input unit 11 inputs gaze area image data and an identifier for identifying the image data (hereinafter referred to as gaze area ID). For example, the input unit 11 inputs a plurality of gaze area image data as shown in FIG. The gaze area image data input by the input unit 11 may be a grayscale image or a color image, and there is no limitation on the image format (image gradation, color format, etc.).

  The label information is input as table information in which the gaze area ID and the grade determined by a doctor or the like are one set. For example, the input unit 11 inputs label information including a grade associated with a gaze area ID corresponding to each of a plurality of gaze area image data as illustrated in FIG.

  The input unit 11 delivers the label information corresponding to the plurality of input gaze area image data to the feature vector generation unit 12.

  The feature quantity vector generation unit 12 calculates a feature quantity vector that characterizes the gaze area image data. Note that the feature quantity vector generation unit 12 generates a plurality of types of feature quantities from a single gaze area image data, and generates a feature quantity vector including a plurality of feature quantities by statistical processing for each type of feature quantity. . In the first embodiment, it is assumed that the feature quantity vector generation unit 12 generates 12 types of feature quantities as shown in FIG.

  First, the feature vector generation unit 12 extracts a cell nucleus region (hereinafter referred to as a cell nucleus region) included in the input gaze region image data. For example, referring to FIG. 4, the feature vector generation unit 12 sequentially extracts regions such as cell nucleus regions 201 and 202. At this time, the feature vector generation unit 12 extracts a cell nucleus region using a luminance difference (contrast) between the cell nucleus region and other regions.

  Next, the feature quantity vector generation unit 12 calculates various feature quantities by performing a feature quantity calculation process on the extracted cell nucleus region. Here, for example, a cell nucleus region as shown in FIG. 6 is extracted. In this case, when calculating the size of the cell nucleus (the area of the cell nucleus; the feature amount F1, see FIG. 5), the feature amount vector generation unit 12 is a pixel constituting the gray region (cell nucleus region) shown in FIG. Count the number of Thereafter, the feature vector generation unit 12 multiplies the count value of the pixel by a predetermined constant (the size of the cell corresponding to the area of one pixel), and sets the result as the feature F1. Alternatively, the feature quantity vector generation unit 12 may set the number of pixels (number of pixels) constituting the cell nucleus region as the feature quantity F1.

  Further, the feature vector generation unit 12 counts the number of pixels (pixels on the boundary line 211 shown in FIG. 6) that form the boundary of the cell nucleus region, and calculates the perimeter of the cell nucleus (feature amount F2) based on the result. To do.

When the size (area) of the cell nucleus and the circumference thereof are obtained, the feature vector generation unit 12 can calculate the circularity (feature F3) of the cell nucleus by the following equation (1).


Where S is the area of the cell nucleus and L is the circumference of the cell nucleus.

  The feature vector generation unit 12 treats the cell nucleus region as an elliptical shape, counts the number of pixels forming the major axis (for example, the major axis 212 shown in FIG. 6), and based on the result, calculates the elliptical major axis length (feature amount) of the cell nucleus. F4) can be calculated. Further, the feature quantity vector generation unit 12 counts the number of pixels forming the elliptical short axis (short axis 213 shown in FIG. 6), and can calculate the elliptic short axis length (feature quantity F5) of the cell nucleus from the result. Further, the feature quantity vector generation unit 12 calculates the feature quantity F6 by calculating the ratio of the elliptical minor axis length to the elliptical major axis length of the cell nucleus.

  The feature quantity vector generation unit 12 calculates feature quantities F7 to F11 using pixel values (density and luminance values) around the cell nucleus region. For example, when the cell nucleus is stained, the feature vector generation unit 12 calculates a threshold that can most efficiently separate the fluorescent region and the non-fluorescent region in the cell nucleus region, and calculates the threshold as the feature F7.

  Further, the feature vector generation unit 12 calculates a gray level co-occurrence matrix (GLCM) from the pixel value of the cell nucleus region, and uses the angular moment of the cell nucleus region from the GLCM value (GLCM: Gray Level Co-occurrence Matrix). ASM: Angular Second Moment, feature quantity F8), contrast (feature quantity F9), uniformity (feature quantity F10), entropy (ENT; Entropy, feature quantity F11), and other feature quantities can be calculated. Further, the feature vector generation unit 12 can calculate the feature (F12) by calculating the nuclear density (NDens; Nuclear Density) of the cell nucleus region.

  The feature quantity vector generation unit 12 includes at least the size of the cell nucleus (feature quantity F1), the roundness of the cell nucleus (feature quantity F3), the contrast of the cell nucleus (feature quantity F9), the cell nucleus as the feature quantity characterizing the gaze area image data. Is generated (feature value F10).

  The feature quantity vector generation unit 12 calculates feature quantities F1 to F12 that characterize the gaze area image data for all cell nuclei (cell nucleus areas) included in the gaze area image data. As a result, for example, if 100 cell nucleus regions are included in one gaze region image data, 100 feature amounts are calculated for each of the feature amounts F1 to F12.

  The feature vector generation unit 12 performs statistical processing on each of a plurality of feature amounts calculated from a single gaze area image data, thereby calculating a plurality of indexes representing the feature amounts. In the following description, a statistical value (index) representing a specific feature amount F is described using a hyphen and a number. For example, referring to FIG. 5, taking the feature amount F1 related to the size of the cell nucleus as an example, the size of the cell nucleus is represented by F1-1 to F1-5. A plurality of statistical values calculated from each feature amount are also referred to as feature amounts because there is no difference in values that characterize the features of cell nuclei. For example, the five feature amounts F1-1 to F1-5 are statistical values representing the feature amount F1.

  For example, the feature vector generation unit 12 generates a frequency distribution (histogram) relating to the feature F1 calculated as described above. Here, for example, it is assumed that a frequency distribution is obtained as shown in FIG. Next, the feature vector generation unit 12 generates a cumulative distribution (see FIG. 7B) from the generated frequency distribution, and calculates the percentile value obtained from the cumulative distribution, whereby the feature relating to the size of the cell nucleus. The amounts F1-1 to F1-5 are calculated.

  Also for the other feature amounts F2 to F12, after calculating individual feature amounts, a frequency distribution and a cumulative distribution of the feature amounts are generated, thereby generating a plurality of feature amounts representing each feature amount. The feature quantity vector generation unit 12 calculates 60 (12 × 5) feature quantities from one gaze area image data by repeating the above processing. That is, the feature vector generation unit 12 calculates a feature vector that characterizes each gaze area image data.

  The feature vector generation unit 12 delivers the label information acquired from the input unit 11 and a plurality of feature amounts calculated for each gaze area image data to the learning data output unit 13.

  The learning data output unit 13 generates learning data based on the information acquired from the feature vector generation unit 12. Specifically, the learning data output unit 13 learns information obtained by combining a gaze area ID, label information (gaze area image data grade), and feature quantity vectors (60 feature quantities). Data is generated (see FIG. 8). That is, the learning data output unit 13 is extracted from the identifier (gaze area ID) for identifying the gaze area image data, the label (grade of the cell image) given to each gaze area image data, and the gaze area image data. Learning data including a plurality of feature quantities (feature quantity vectors) as a set is generated and output.

  The learning data output unit 13 outputs the generated learning data to the data processing device 20. Note that learning data may be input / output from the learning data generation device 10 to the data processing device 20 using an external storage device such as a USB (Universal Serial Bus) memory, a network, a database server, or the like. .

  FIG. 9 is a diagram illustrating an example of an internal configuration of the data processing device 20. Referring to FIG. 9, the data processing device 20 includes an input unit 21, a feature amount selection unit 22, a decision tree generation unit 23, an output unit 24, and a storage unit 25 including an HDD or the like. Note that illustrations of operation devices (keyboard, mouse, etc.) and display devices for operating the data processing device 20 are omitted. Each unit including the input unit 21 is configured to access the storage unit 25 and write and read data.

  The input unit 21 is means for inputting learning data output from the learning data generation device 10. The input unit 21 passes the acquired learning data to the feature amount selection unit 22.

  The feature quantity selection unit 22 uses the feature quantity vectors (a plurality of feature quantities; 60 feature quantities in the above example) included in the acquired learning data to be used for generation of a decision tree by the decision tree generation unit 23. Is a means for selecting. Specifically, the feature quantity selection unit 22 executes a first selection process and a second selection process, and finally narrows down the feature quantities used by the decision tree generation unit 23.

  The feature quantity selection unit 22 executes the first selection process while referring to the first selection policy stored in the storage unit 25. For example, information as shown in FIG. 10 is stored in the storage unit 25 as the first selection policy.

  Referring to FIG. 10, the first selection policy indicates that the types of feature quantities to be used are feature quantities F1 to F11. Therefore, the feature quantity selection unit 22 is included in the feature quantity vector of the learning data. Among the feature amounts F1 to F12 to be selected, the feature amounts F1 to F11 excluding the feature amount F12 are selected.

  Further, the first selection policy includes information on which one of the plurality of statistical values representing the feature value is adopted, and the information is described as “median”, The feature amount selection unit 22 selects a feature amount corresponding to the median value. Specifically, referring to FIG. 8, among the feature amounts F1-1 to F1-5 representing the feature amount F1 related to the size of the cell nucleus, the feature amount F1-3 corresponds to the median value (FIG. 7B). Therefore, the feature amount selection unit 22 selects the feature amount F1-3 as the feature amount representing the feature amount F1 related to the size of the cell nucleus. As described above, the feature quantity selection unit 22 selects one feature quantity from a plurality of feature quantities (statistical values) representing each feature quantity in accordance with the first selection policy.

  The feature quantity selection unit 22 performs the first selection process on the learning data shown in FIG. 8, and the result is shown in FIG. As illustrated in FIG. 11, the feature amount selection unit 22 narrows down from 60 feature amounts to 11 feature amounts by executing the first selection process.

  The feature quantity selection unit 22 executes the second selection process on the learning data for which the first selection process has been completed, according to the second selection policy stored in the storage unit 25.

  For example, information as illustrated in FIG. 12 is stored in the storage unit 25 as the second selection policy. The “analysis model = decision tree” of the second selection policy has a label (grade) as an objective variable (explained variable) and a feature amount as an explanatory variable for learning data (for example, learning data shown in FIG. 11). It means to generate an analysis model by decision tree. Then, using the analysis model based on the decision tree, the importance of the feature quantity (explanatory variable) is evaluated, two feature quantities with low importance are deleted, and finally the feature quantity is narrowed down to four. The second selection policy is shown.

  Here, the second selection process by the feature amount selection unit 22 will be described with reference to FIG.

  First, the feature quantity selection unit 22 refers to the second selection policy stored in the storage unit 25 (step S101).

  Next, the feature quantity selection unit 22 generates a decision tree based on the learning data for which the first selection process has been completed in accordance with “analysis model = decision tree” described in the second selection policy (step S102). . It should be noted that an algorithm such as CART (Classification And Regression Trees) algorithm or ID (Iterative Dichotomiser) 3 can be used for generation of a decision tree by the feature quantity selection unit 22 or a decision tree generation unit 23 described later. In addition, a decision tree branch condition can be generated by calculating the Gini coefficient and entropy.

  Here, for example, it is assumed that a decision tree as shown in FIG. 14 is obtained. In the decision tree branch conditions 301 to 307 shown in FIG. 14, the variables X, Y, and Z are any one of the feature amounts F1-3 to F11-3.

Next, the feature quantity selection unit 22 calculates the quality of each explanatory variable (feature quantity) included in the branch condition (branch conditions 301 to 307 in the example of FIG. 14) of the generated decision tree. (Step S103). Specifically, the feature quantity selection unit 22 calculates the Gini coefficient G of each explanatory variable using the following equation (2).


In the equation (2), Pi represents the probability of class i.

  The feature quantity selection unit 22 that has calculated the Gini coefficient G determines the value of the variable X that minimizes the Gini coefficient G as the best split point (Best Split Point) Xs. Alternatively, for example, if the Gini coefficient is used when generating a decision tree as illustrated in FIG. 14, the feature amount selection unit 22 may specify the division point Xs using the calculated Gini coefficient.

The feature quantity selection unit 22 calculates the quality Q (X, Xs) of the variable X at the dividing point Xs using the following equation (3).

... (3)

In Expression (3), N represents the total number of variables, N _L represents the number of variables classified into the left child node, and N _R represents the number of variables classified into the right child node. I _{{A} in} Equation (3) represents an indicator function, and is a function that outputs “1” when the condition A is satisfied, and outputs “0” when the condition A is not satisfied (the condition A is not satisfied). .

I _{{Ci = Neg}} shown in Expression (3) is an instruction function that outputs “1” when the class (category) Ci is negative, and outputs “0” otherwise. I _{{Xi> Xs}} is an instruction function that outputs “1” when Xi is larger than Xs, and outputs “0” otherwise. In the expression (3), I _{A} I _{B} means that “1” is output when the conditions A and B are satisfied simultaneously, and “0” is output otherwise. . Therefore, the total sum of I _{{Ci = Neg}} I _{{Xi ≦ Xs}} in Expression (3) is a sum related to data belonging to the Negative class and having a feature quantity equal to or less than Xs. The product of other indicator functions shown in Equation (3) has the same meaning. For example, the total sum of I _{{Ci = Pos}} I _{{Xi ≦ Xs}} in Expression (3) is a sum relating to data belonging to the Positive class and having a feature quantity equal to or less than Xs.

For example, as shown in FIG. 15, it is assumed that the variable X is optimally divided by the division point Xs. In this case, since N = 10, NL = 5, and NR = 5, the quality Q (X, Xs) of the variable X at the dividing point Xs is obtained by applying Q (X, Xs) = ( 4 ² +1 ² ) / 5 + (0 ² +5 ² ) /5=8.4.

  In the example illustrated in FIG. 14, the quality of variables (X, Y, Z) used for the branch conditions 301 to 307 is calculated. In FIG. 14, the quality Q under each branch condition is shown in the branch condition by the variable used in the branch condition and its sign. For example, since the variable X is used in the branch condition 301, the quality Q in the branch condition 301 is expressed as Q (X, 301).

Next, the feature quantity selection unit 22 calculates the importance (Importance) of each feature quantity whose quality has been calculated (step S104). Specifically, the importance of the feature quantity is calculated from the ratio of the quality of each variable (feature quantity) to the sum of the quality of each branch condition of the decision tree. For example, in the example shown in FIG. 14, the importance of the variable X can be calculated by Expression (4), the importance of the variable Y can be calculated by Expression (5), and the importance of the variable Z can be calculated by Expression (6).

  Next, the feature quantity selection unit 22 narrows down the feature quantity according to the “squeezing method” included in the second selection policy (step S105). For example, the feature quantity selection unit 22 creates a decision tree for the learning data shown in FIG. 11 and arranges the importance of each variable in descending order (in descending order of importance) as shown in FIG. If there is, the lower two feature quantities F2-3 and F11-3 are deleted. In FIG. 16, rows colored in gray are rows that are deleted by narrowing down by the feature amount selection unit 22. As described above, the feature amount selection unit 22 generates new learning data by deleting a predetermined number of feature amounts from a plurality of feature amounts included in the learning data based on the importance calculated in the previous step. To do.

  Next, the feature amount selection unit 22 determines whether or not the new learning data matches the “end condition” included in the second selection policy (step S106). Here, since “end condition = number of feature quantities is four”, the feature quantity selection unit 22 determines whether the number of feature quantities has been narrowed down to four.

  If the new learning data does not satisfy the termination condition (No at Step S106), the feature amount selection unit 22 returns to Step S102 and continues the processing. That is, the feature quantity selection unit 22 creates a decision tree again using new learning data in which the feature quantity has been narrowed down, calculates the quality and importance of the variables that constitute the branch condition of the decision tree, Delete low feature quantities.

  If the new learning data satisfies the end condition (step S106, Yes branch), the feature amount selection unit 22 ends the process.

  As a result of the narrowing down as described above, the feature quantity shown in FIG. 16A is narrowed down as shown in FIGS. 16B and 16C, and finally the feature quantity shown in FIG. 16D. (Four feature values from the top).

  The learning data obtained as a result of narrowing down the two-stage feature quantity by the feature quantity selection unit 22 is as shown in FIG. The feature quantity selection unit 22 hands over the learning data (for example, the learning data shown in FIG. 17) whose feature quantity has been narrowed down by performing the first selection process and the second selection process to the decision tree generation unit 23.

  The decision tree generator 23 generates an identification rule based on the acquired learning data. Specifically, the decision tree generator 23 generates a decision tree based on the learning data shown in FIG. When the decision tree is generated, the decision tree generation unit 23 selects variables to be divided until the impurity becomes “0” or until the branch of the decision tree reaches a predetermined depth. Repeat the division of the subset. The decision tree generation unit 23 delivers the generated decision tree to the output unit 24.

  For example, the output unit 24 outputs the acquired decision tree to an external device (for example, the identification device 30) in the form of “If-Then”. Alternatively, the output unit 24 may visualize the “If-Then” format as shown in FIG. 14, for example, and output the image data.

  The first selection policy and the second selection policy are preferably configured so that the user can arbitrarily change the contents thereof. If the feature quantity used when the decision tree is generated by the decision tree generator 23 is different, the branching conditions (grading grounds and reasons) and the classification results (identification results, grading) also differ. Therefore, even if learning data including feature amounts extracted from the same cell image (for example, learning data including 60 feature amounts) is input to the data processing device 20, the feature amount used for generation of the decision tree is changed. By doing this, it becomes possible to realize multi-faceted and multi-faceted research and analysis for the sample that is the basis of the learning data (sample from which the gaze area image data is extracted).

  The operations of the data processing apparatus 20 described above are summarized as shown in FIG.

  In step S <b> 01, the data processing device 20 inputs learning data from the learning data generation device 10.

  In step S02, the data processing device 20 narrows down the feature amount included in the input learning data by executing the first and second selection processes.

  In step S03, the data processing device 20 generates a decision tree using the learning data including the narrowed feature amount.

  In step S04, the data processing device 20 outputs the decision tree to the outside.

[Application example]
Next, an example where the decision tree generation method described in the first embodiment is applied will be described. Here, a feature vector is generated from gaze area image data (part of a cell image) generated from hepatocytes of 1105 patients, and finally from learning data (see FIG. 19) narrowed down to four feature values. A case where a decision tree is generated will be described. In FIG. 19, the feature quantity F1-3 regarding the size of the cell nucleus uses the number of pixels forming the cell nucleus region.

  FIG. 20 is a diagram illustrating an example of a decision tree obtained from the learning data illustrated in FIG. In calculating the decision tree, the depth of the decision tree is “4”. Further, in the decision trees shown in FIG. 20 and subsequent figures, when the branch condition is satisfied, the branch is made on the left side, and when not satisfied, the branch is made on the right side. Furthermore, since even the same grade may be assigned to different classification results, alphabets are added for the purpose of distinguishing class labels (grades G0 to G4) of the classification results. For example, even the same grade G2 can be classified into the classification results 401 to 405, so that G2a to G2e are written in the classification results in order to distinguish them.

  Referring to FIG. 20, it can be seen that whether or not the grade is less than G3 largely depends on the circularity (feature amount F3-3) of the cell nucleus used in the first branch condition from the root node. Further, as described above, it can be seen that even the same grade G2 is distributed to five types of classification results. In other words, it can be said that gaze area image data belonging to different classification results have different characteristics even with the same grade.

  In this way, the identification rule indicated by the decision tree is expressed in the “If-Then” format, so that visualization as shown in FIG. 20 is easy. Therefore, the reason and basis of grading can be easily understood by referring to the decision tree visualized by a doctor or the like. For example, doctors in contact with FIG. 20 can obtain the grounds and reasons for grading that a high grade is given because the grade is high because the degree of circularity is high, and a high grade is given because the cell nucleus is large. Alternatively, by confirming the branch condition from the leaf (class label) node toward the root node (confirming the flow of the decision tree), the doctor or the like can grasp the characteristics of each class label.

  As the depth of the decision tree generated by the decision tree generator 23 increases, the classification accuracy increases. FIG. 21 shows a grading result when the depth of the decision tree is allowed up to 20 from the learning data shown in FIG. 19 (FIG. 21A) and a grading result when the depth of the decision tree is allowed up to 4 ( It is a figure which shows (b) of FIG. Grades G0t to G4t shown in FIG. 21 indicate judgments by a doctor (label; true value, True), and grades G0p to G4p indicate grade predicted values (Prediction) obtained by applying the generated decision tree.

  A portion where the vertical and horizontal grading in FIG. 21 intersect (gray portion in the figure) indicates that the determination by the doctor and the prediction by the decision tree match, and the more the number included in the intersection, the more the determination. This indicates that the accuracy of grading with trees is high. Specifically, when the depth of the decision tree is set to “20”, the accuracy is 96.2%. On the other hand, when the depth of the decision tree is set to “4”, the accuracy is 56.7%.

  In this way, the grading accuracy improves as the depth of the decision tree increases, but as the depth of the generated decision tree increases, the basis for grading by the decision tree is difficult for doctors to understand. It becomes. In other words, there is a trade-off relationship between the accuracy of grading by a decision tree and the ease of understanding the grounds and reasons for grading by a decision tree. Therefore, it is desirable to generate a decision tree with a depth that optimizes the relationship between accuracy and ease of understanding.

  As described above, the data processing apparatus 20 according to the first embodiment uses the decision tree for the purpose of grasping the influence (importance) of the feature quantity used for generating the identification rule. In addition, the data processing device 20 narrows down the feature amount to be finally used by deleting the feature amount having a small influence while leaving the feature amount having a large influence on the grading result among the plurality of feature amounts. . By narrowing down the feature amount, the size of the decision tree generated by the decision tree generator 23 is reduced, and the ease of understanding the grounds and reasons for grading is enhanced.

  In addition, the data processing apparatus 20 does not perform the procedures of decision tree generation, feature amount evaluation, and feature amount narrowing down only once, but narrows down feature amounts by performing the same procedure multiple times. . The reason for performing such a plurality of times of narrowing down is to eliminate the influence of a complicated relationship existing between feature quantities as much as possible and to improve the accuracy of grading. For example, when the feature quantity A and the feature quantity B represent the same feature of the cell nucleus, it is less necessary to use these feature quantities at the same time for generating a decision tree. For example, if the feature amount A is used preferentially, the influence on the result of the feature amount B is low, and even if the feature amount B is deleted, the influence is small. On the other hand, when the feature amount A and the feature amount B are used at the same time, the classification accuracy may be increased. In this case, when the feature amount A is used, the influence amount of the feature amount B is high, but when the feature amount A is not used, the utility value (the influence degree on the result) of the feature amount B is also low. As described above, since the importance of the feature quantity depends on the presence of other feature quantities, the importance of each feature quantity changes for each combination of feature quantities. For example, referring to FIG. 16A, the importance of the feature amount F3-3 is fifth. On the other hand, in FIG. 16D as a result of sequentially narrowing down the feature amount, the importance amount of the feature amount F3-3 is the first. That is, it can be said that the influence of the feature amount F3-3 is large when the number of feature amounts to be used is small. When four highly important feature amounts are selected in the stage of FIG. 16A, the feature amount F3-3 is excluded, and the feature amount F3-3 having a high influence level is not used with a small number of feature amounts. Occurs. In order to avoid such inconvenience, the data processing apparatus 20 repeats the procedures of generating a decision tree, evaluating feature amounts, and narrowing down feature amounts.

  The decision tree has the advantage that the objective variable can be separated non-linearly, and high accuracy can be obtained if the depth of the decision tree is sufficiently large. In addition, the identification rule based on the decision tree can be easily visualized, and there is an advantage that it is easy to understand the basis and reason for the classification result. These advantages are absent or sparse in other analytical models, learning models (eg, Support Vector Machine (SVM)). The data processing device 20 according to the first embodiment generates a decision tree as an identification rule based on the provided learning data, thereby achieving both classification accuracy and ease of understanding.

[Second Embodiment]
Next, a second embodiment will be described in detail with reference to the drawings.

  In the first embodiment, generation of a decision tree from learning data has been described. In the second embodiment, further utilization of the decision tree will be described.

  FIG. 22 is a diagram illustrating an example of a configuration of a pathological image processing system according to the second embodiment. Referring to FIG. 22, the learning data generation device 10a inputs the accompanying information of the gaze area image data associated with the gaze area ID in addition to the gaze area image data and the label information. The learning data generation device 10a generates learning data by the method described in the first embodiment and outputs the learning data to the data processing device 20a. The accompanying information acquired by the learning data generation device 10a is provided to the data processing device 20a together with the learning data.

  The data processing device 20a generates a decision tree by the method described in the first embodiment. The data processing device 20a has a function of analyzing the characteristics of each classification result based on the generated decision tree based on the accompanying information. Specifically, the data processing device 20a generates and outputs various analysis data and analysis images as analysis results using the decision tree, its classification result, and accompanying information.

  FIG. 23 is a diagram illustrating an example of accompanying information. In FIG. 23, labels are also shown for easy understanding. The accompanying information shown in FIG. 23 includes the anticancer drug administered to the patient who collected the liver pathological image that is the basis of the gaze area image data associated with the gaze area ID, and the effect of the anticancer drug (+ is effective, − is ineffective ). The learning data output unit 13 of the learning data generation device 10a adds the accompanying information to the learning data and outputs it to the data processing device 20a. Note that the anticancer agents A to D and the effects thereof shown in FIG. 23 and subsequent figures are virtual cases (data) for explaining the operation of the data processing device 20a.

  FIG. 24 is a diagram illustrating an example of an internal configuration of the data processing device 20a according to the second embodiment. Differences between the data processing device 20 and the data processing device 20a according to the first embodiment are that each unit of the data processing device 20a is configured to handle incidental information, a point having an analysis unit 26, and generation The classification result by the determined decision tree is handed over to the analysis unit 26.

  FIG. 25 is a diagram illustrating an example of a classification result based on a decision tree generated by the decision tree generation unit 23 of the data processing device 20a. As illustrated in FIG. 25, the decision tree generation unit 23 delivers a list of gaze area IDs belonging to each classification result (each class label) based on the generated decision tree to the analysis unit 26 as a classification result. Referring to FIG. 23 and FIG. 25, the gaze area image data belonging to each classification result by the gaze area ID is related to the effect of the anticancer agent administered to the patient who provided the gaze area image data. For example, gaze area image data acquired from a patient corresponding to gaze area ID = 1 is classified into a grade of “G2a”, and at least anticancer drugs A, B, and D are effective among anticancer drugs administered to the patient. This can be understood from FIGS. 23 and 25.

  The analysis unit 26 is a means for analyzing the characteristics of each classification result based on the decision tree based on the above information (decision tree, classification result, and accompanying information). For example, the analysis unit 26 analyzes the effectiveness of various anticancer agents for the patient corresponding to the gaze area ID assigned to each classification result. Specifically, the analysis unit 26 analyzes the effectiveness of the anticancer agent by the following procedure.

  First, the analysis unit 26 acquires a gaze area ID included in each classification result. Next, the analysis unit 26 refers to the accompanying information and acquires the effect of each anticancer agent for each acquired gaze area ID. Thereafter, the analysis unit 26 calculates a ratio of data indicating that the anticancer drug is effective for each classification result (grade; class label) and anticancer drug, and the ratio is a threshold value (for example, 50% or more; majority decision). For example, it is determined that the anticancer drug is effective.

  For example, taking the grade G2a shown in the classification result of FIG. 25 as an example, the grade includes at least gaze area image data specified by the gaze area ID = 1, 2, and 3. Next, referring to the accompanying information shown in FIG. 23, the results of anticancer drug administration for gaze area ID = 1, 2, 3 are obtained. For example, taking anticancer agent A as an example, it is effective for two patients out of three patients (patients corresponding to gaze area ID = 1 to 3) (two + exist), so anticancer agent A is effective. The ratio of data indicating that is 66.6%. Therefore, it is determined that the anticancer agent A is effective for the patient specified from the gaze area ID belonging to the grade G2a.

  Note that the threshold (50% in the above example) when the analysis unit 26 determines the effectiveness of the anticancer agent may be common to all anticancer agents, or may be set individually. . For example, when it is desired to carefully determine the effectiveness of anticancer drug A, the threshold value may be set higher (for example, 80%). Alternatively, when the analysis unit 26 calculates the effectiveness of the anticancer agent for each grade, the data indicating that the anticancer agent is effective is within a predetermined range (for example, a range of 40% to 60%, etc.). The effect of the anticancer agent may be “unknown”.

  The analysis unit 26 performs the determination as described above for each grade and anticancer agent of the classification result, and obtains an analysis result as shown in FIG. The analysis unit 26 passes the analysis result and the decision tree to the output unit 24.

  The output unit 24 uses the decision tree and the analysis result to generate data indicating the effectiveness of the anticancer agent for the patients (a group of patients associated with the gaze area ID) assigned to each classification result (grade) based on the decision tree. Output to an external device or display device.

  For example, the output unit 24 generates image data as shown in FIG. 27 and outputs it to the outside. In FIG. 27, for the sake of easy understanding, only the effectiveness of anticancer agents related to grade G2a and grade G2b is illustrated. Referring to FIG. 27, it can be confirmed that there is a significant difference in the effectiveness of anticancer agents between G2a and G2b even in patients who are assigned to the same grade G2. Doctors who have contacted the information as shown in FIG. 27 have found that anticancer drug B is not effective for patients assigned to G2a grade, and that anticancer drug D is effective for patients assigned to G2b. Can be obtained.

  As described above, when the accompanying information provided to the data processing device 20a is a result regarding the effectiveness of the anticancer agent for the patient corresponding to the gaze region ID, the data processing device 20a includes the gaze region included in each classification result. An analysis result indicating the effectiveness of the anticancer agent for the patient corresponding to the ID can be output.

  The data analysis by the data processing device 20a is not limited to the effectiveness of the anticancer agent. Other analyzes can be performed by changing the content of the accompanying information. For example, it is assumed that the information shown in FIG. 28 is input to the data processing device 20a as the accompanying information. The accompanying information shown in FIG. 28 includes the number of days that cancer has recurred in the patient associated with the gaze area ID.

  The analysis unit 26 calculates the probability of cancer recurrence over time for each grade of the classification result, and calculates it as an analysis result. Specifically, the analysis unit 26 calculates data relating to a graph as illustrated in FIG. 29 as an analysis result, and passes it to the output unit 24.

  Referring to FIG. 29, it can be seen that patients assigned to the grade of “G1a” have no possibility of cancer recurrence within the range of collected samples. Moreover, it turns out that the tendency of the cancer recurrence of the patients classified into the grade of "G2a" and "G2b" is different. Specifically, if the number of days is less than 1000 days, there is no significant difference in the cancer recurrence rate of patients assigned to grades G2a and G2b, but if the number of days exceeds 1000 days, the cancer recurrence rate between the two A marked difference is observed.

  As described above, when the accompanying information provided to the data processing device 20a is information on a period until the patient corresponding to the gaze area ID recurs cancer, it corresponds to the gaze area ID included in each classification result. The tendency regarding the recurrence of cancer in the patient can be output as the analysis result.

  As described above, in the pathological image processing system according to the second embodiment, by analyzing the accompanying information, it is possible to provide a doctor or the like with information that shows remarkable features in each classification result (grade) based on the decision tree.

  The configuration of the pathological image processing system described in the above embodiment is an example, and is not intended to limit the configuration of the system. For example, some of the functions of the data processing device 20 may be incorporated in the learning data generation device 10. For example, the learning data generation apparatus 10 may execute all or part of the feature amount narrowing down in the data processing apparatus 20 described in the first embodiment. Alternatively, in place of the learning data generation device 10, a device that extracts a feature vector from gaze area image data is prepared, and label information is directly input to the data processing device 20, and learning data is stored inside the data processing device 20. May be generated. Alternatively, the function of the identification device 30 may be included in the data processing device 20. In this case, as illustrated in FIG. 30, the data processing device 20b includes an identification unit 27 that performs prediction of sample data using the decision tree generated by the decision tree generation unit 23. The input unit 21 inputs sample data, and the output unit 24 outputs the identification result.

  In the above embodiment, the grade of the gaze area image data is used as the label information, but the label is not limited to the grade of the gaze area image data (cell image). For example, information regarding cancer recurrence of a patient may be used as a label. For example, as shown in FIG. 31, the cancer recurrence information (no long-term recurrence, early recurrence) of the patient corresponding to the gaze area ID may be used as a label. In this case, the features (for example, the size of the cell nucleus, the circularity, etc.) possessed by the cell nucleus of the gaze region image data are obtained by extracting the feature amount, narrowing down the feature amount, and creating the decision tree described in the first embodiment. A decision tree (prediction model) relating to cancer recurrence as a branching condition can be obtained (a decision tree corresponding to FIG. 20 can be obtained). In addition, according to the same procedure as the method described in the second embodiment, by using the label and the number of days until the patient relapses as incidental information, the cancer recurrence of the patient included in each classification result of the decision tree Can be obtained (a graph corresponding to FIG. 29 can be obtained).

  In the above embodiment, the system configuration (FIGS. 2 and 22) including the learning data generation device that calculates the feature amount vector from the gaze area image data and generates the learning data has been described. The present invention is not limited to a generation device (information processing device, computer). For example, a feature amount (feature amount vector) calculated by a doctor or the like may be used, or a feature amount calculated by the apparatus and a feature amount calculated by a doctor or the like may be combined. That is, as long as the learning data provided to the data processing device 20 includes a feature vector that characterizes each of a plurality of samples, any method for generating the feature vector may be used.

  Further, the decision tree (identification rule) used in the identification device 30 is not limited to the decision tree generated by the data processing device 20. That is, as long as the decision tree is generated by the method and procedure described in the above embodiment, the generation subject is not limited to the information processing apparatus (computer), and may be anything. That is, it is possible to perform sample grading by preparing learning data (data including cell image identifiers, labels, and feature amounts) and using a decision tree generated from the learning data.

  In the above embodiment, the case where twelve types of feature values are calculated has been described. However, the type of feature values to be calculated is not limited. For example, contrast (feature amount F9) and uniformity (feature amount F10) have been shown as feature amounts indicating the texture of the cell nucleus region, but feature amounts obtained by texture analysis such as Fourier transform or wavelet transform may also be used. Good.

  In the above-described embodiment, the case has been described in which the feature vector generation unit 12 calculates the percentile value obtained from the cumulative distribution of a plurality of feature amounts as a statistical value representing the feature amount. However, it will be appreciated that other statistics can be used. For example, statistical values such as variance values and mode values obtained from a plurality of feature amounts may be used. Further, although the case has been described in which the feature amount selection unit 22 of the data processing device 20 selects one feature amount (statistical value) from the same type of feature amount, a plurality of feature amounts are selected from the same type of feature amount. May be. For example, an intermediate value (feature value F1-3) of the feature amount related to the size of the cell nucleus and a variance value of the size of the cell nucleus may be selected.

  Further, the processing performed by each unit such as the feature vector generation unit 12 of the learning data generation device 10, the feature amount selection unit 22 of the data processing device 20, and the decision tree generation unit 23 is performed by these devices (the learning data generation device 10, The present invention can be realized by a computer program that causes a computer mounted on the data processing device 20) to execute the above-described processes using its hardware. That is, it is sufficient if there is a means for executing the functions performed by the above-described units with some hardware and / or software.

  Furthermore, by installing a computer program in the storage unit of the computer, the computer can function as the learning data generation device 10, the data processing device 20, and the identification device 30. Furthermore, by causing a computer to execute the above-described computer program, a learning data generation method, a decision tree generation method, a prediction method using a decision tree, and the like can be executed by the computer. The program can be downloaded via a network or updated using a storage medium storing the program.

  A part or all of the above embodiments can be described as in the following supplementary notes, but is not limited thereto.

[Appendix 1]
The data processing apparatus according to the first aspect described above.
[Appendix 2]
The data processing apparatus according to appendix 1, further comprising a feature quantity selection unit that selects a feature quantity used for generation of a decision tree by the decision tree generation unit from among the plurality of feature quantities.
[Appendix 3]
The feature amount selection unit includes:
Generation of a decision tree based on the learning data;
Calculating the quality of each feature quantity included in the branch condition of the generated decision tree;
Calculating the importance of each feature quantity for which the quality has been calculated;
Deleting a predetermined number of feature amounts from a plurality of feature amounts included in the learning data based on the calculated importance, and generating new learning data;
Determining whether or not the feature amount included in the new learning data satisfies a predetermined condition;
The data processing apparatus according to appendix 2, wherein the feature quantity used for generation of the decision tree by the decision tree generation unit is selected by repeating the above.
[Appendix 4]
The input unit inputs accompanying information associated with the cell image by an identifier for identifying the cell image together with the learning data,
The data processing apparatus according to any one of appendices 1 to 3, further comprising an analysis unit that analyzes the characteristics of each of the classification results obtained by the decision tree by the decision tree generation unit based on the accompanying information.
[Appendix 5]
The analysis unit
When the accompanying information is a result regarding the effectiveness of the anticancer agent for the patient corresponding to the identifier of the cell image, the effectiveness of the anticancer agent for the patient corresponding to the identifier of the cell image included in each of the classification results is determined. The data processing apparatus according to appendix 4, which is output as an analysis result.
[Appendix 6]
The analysis unit
When the accompanying information is information regarding a period until the patient corresponding to the identifier of the cell image recurs cancer, it relates to cancer recurrence of the patient corresponding to the identifier of the cell image included in each of the classification results. The data processing apparatus according to appendix 4, which outputs a trend as an analysis result.
[Appendix 7]
Any one of appendices 1 to 6, wherein the decision tree generated by the decision tree generation unit includes at least one of the size, circularity, uniformity, and contrast of a cell nucleus included in the cell image as a branching condition. A data processing apparatus according to one.
[Appendix 8]
The data processing apparatus according to any one of appendices 1 to 7, wherein the decision tree generated by the decision tree generation unit includes a circularity of a cell nucleus included in the cell image in a first branch condition from a root node.
[Appendix 9]
The cell image is an image obtained from hepatocytes, and the label given to the cell image is information regarding the grade of cancer of the hepatocytes or cancer recurrence of the patient, according to any one of appendices 1 to 8. Data processing equipment.
[Appendix 10]
This is the same as the decision tree generation method according to the second viewpoint described above.
[Appendix 11]
This is as the identification device according to the third viewpoint described above.
[Appendix 12]
It is as the program which concerns on the above-mentioned 4th viewpoint.
Note that the form of Supplementary Notes 10 to 12 can be expanded to the form of Supplementary Note 2 to the form of Supplementary Note 9 as with the form of Supplementary Note 1.

  Each disclosure of the cited patent documents and the like cited above is incorporated herein by reference. Within the scope of the entire disclosure (including claims) of the present invention, the embodiments and examples can be changed and adjusted based on the basic technical concept. In addition, various combinations or selections of various disclosed elements (including each element in each claim, each element in each embodiment or example, each element in each drawing, etc.) within the scope of the entire disclosure of the present invention. Is possible. That is, the present invention of course includes various variations and modifications that could be made by those skilled in the art according to the entire disclosure including the claims and the technical idea. In particular, with respect to the numerical ranges described in this document, any numerical value or small range included in the range should be construed as being specifically described even if there is no specific description.

10, 10a Learning data generation device 11, 21, 101 Input unit 12 Feature amount vector generation unit 13 Learning data output unit 14, 25, 34 Storage unit 20, 20a, 20b Data processing device 22 Feature amount selection unit 23, 102 Decision tree Generation unit 24 Output unit 26 Analysis unit 27 Identification unit 30 Identification device 100 Data processing device 201, 202 Cell nucleus region 211 Boundary line 212 Long axis length 213 Short axis length 301-307 Branching conditions 401-405 Classification result

Claims (12)

An input unit for inputting learning data including a label given to the cell image and a plurality of feature amounts extracted from the cell image;
A decision tree generating unit for generating a decision tree for identifying information corresponding to the label of the sample based on the learning data;
A data processing apparatus.   The data processing apparatus according to claim 1, further comprising a feature quantity selection unit that selects a feature quantity used for generation of a decision tree by the decision tree generation unit from the plurality of feature quantities. The feature amount selection unit includes:
Generation of a decision tree based on the learning data;
Calculating the quality of each feature quantity included in the branch condition of the generated decision tree;
Calculating the importance of each feature quantity for which the quality has been calculated;
Deleting a predetermined number of feature amounts from a plurality of feature amounts included in the learning data based on the calculated importance, and generating new learning data;
Determining whether or not the feature amount included in the new learning data satisfies a predetermined condition;
The data processing apparatus according to claim 2, wherein the feature amount used for generation of the decision tree by the decision tree generation unit is selected by repeating the above. The input unit inputs accompanying information associated with the cell image by an identifier for identifying the cell image together with the learning data,
4. The data processing apparatus according to claim 1, further comprising an analysis unit that analyzes characteristics of each classification result obtained by the decision tree by the decision tree generation unit based on the accompanying information. 5. The analysis unit
When the accompanying information is a result regarding the effectiveness of the anticancer agent for the patient corresponding to the identifier of the cell image, the effectiveness of the anticancer agent for the patient corresponding to the identifier of the cell image included in each of the classification results is determined. The data processing device according to claim 4, wherein the data processing device outputs the result as an analysis result. The analysis unit
When the accompanying information is information regarding a period until the patient corresponding to the identifier of the cell image recurs cancer, it relates to cancer recurrence of the patient corresponding to the identifier of the cell image included in each of the classification results. The data processing apparatus according to claim 4, wherein the tendency is output as an analysis result.   7. The decision tree generated by the decision tree generation unit includes at least one of the size, circularity, uniformity, and contrast of cell nuclei included in the cell image as a branching condition. A data processing apparatus according to claim 1.   The data processing according to any one of claims 1 to 7, wherein the decision tree generated by the decision tree generation unit includes a circularity of a cell nucleus included in the cell image in a first branch condition from a root node. apparatus.   The cell image is an image obtained from a hepatocyte, and a label given to the cell image is a grade relating to cancer of the hepatocyte or information relating to cancer recurrence of a patient. The data processing apparatus described in 1. Inputting learning data including one set of a label given to the cell image and a plurality of feature amounts extracted from the cell image;
Generating a decision tree for identifying information corresponding to the label of a sample based on the learning data;
A decision tree generation method including:   An identification apparatus that identifies a sample using the decision tree generated by the decision tree generation method according to claim 10. A process of inputting learning data including a label given to the cell image and a plurality of feature amounts extracted from the cell image;
Processing for generating a decision tree for identifying information corresponding to the label of the sample based on the learning data;
A program for causing a computer mounted on a data processing apparatus to execute the program.