Method for generating model for determining degree of growth inhibition, model, and method, system and program for determining degree of growth inhibition using model

Abstract

To determine the degree of bacterial growth inhibition using a machine learning.SOLUTION: Provided is a method comprising: a step of preparing a plurality of pieces of teaching data, each teaching data comprising an image of a plate used to grow a predetermined bacteria and the degree of growth inhibition of the predetermined bacteria related to the image; and a step of generating a model for determining the degree of growth inhibition using at least an image as an input and outputting the degree of growth inhibition, by machine learning using the plurality of pieces of teaching data, the degree of growth inhibition comprised in each teaching data being determined based on at least one or both of the size of the background lawn and the density of the background lawn in the image comprised in the teaching data.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to aiding in the assessment of mutagenicity of a substance. More particularly, the present disclosure relates to using machine learning to determine the degree of bacterial growth inhibition to aid in the assessment of mutagenicity of a substance. [Background technology]

[0002] In order to evaluate the mutagenicity of a substance, a test called the Ames test is carried out. In the Ames test, a specific bacterium (or strain) is exposed to various doses of a substance and then growth is attempted, and the mutagenicity of the substance is evaluated based on the results.

[0003] Furthermore, Patent Document 1 describes a device for counting cells in a specific cell state using machine learning. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2020-166711 A Summary of the Invention [Problem to be solved by the invention]

[0005] In the Ames test, when evaluating the mutagenicity of a substance, the presence or absence of growth inhibition at each dose of the substance is also determined. Conventionally, the degree of growth inhibition has been determined by visual inspection of the bacterial growth results. [Means for solving the problem]

[0006] The present disclosure has been made in view of the above, and its object is to determine the degree of bacterial growth inhibition using machine learning.

[0007] In order to solve the above problem, according to an embodiment of the present disclosure, a method is provided which includes the steps of preparing a plurality of teacher data, each of which includes an image of a plate used to grow a specified bacterium and a degree of growth inhibition of the specified bacterium associated with the image, and generating a growth inhibition degree determination model by machine learning using the plurality of teacher data, the model taking an image as at least an input and outputting a degree of growth inhibition, wherein the degree of growth inhibition included in each teacher data is determined based on at least one or both of the size of the background loan and the density of the background loan in the image included in the teacher data.

[0008] In one embodiment, the method may further include inputting an image of a plate used to grow the specified bacterium into the growth inhibition degree determination model and determining the degree of growth inhibition of the specified bacterium associated with the image based on the output of the model.

[0009] In one embodiment, the degree of growth inhibition contained in each training data may be determined based on the condition of a negative control plate of the same test batch as the plate in the image contained in the training data.

[0010] In one embodiment, the step of preparing a plurality of teacher data includes a step of preparing a plurality of first teacher data, each of which includes an image of a plate used to grow the predetermined bacterium, the state of a negative control plate of the same batch as the plate being a first type, and a degree of growth inhibition of the predetermined bacterium associated with the image; and a step of preparing a plurality of second teacher data, each of which includes an image of a plate used to grow the predetermined bacterium, the state of a negative control plate of the same batch as the plate being a second type, and a degree of growth inhibition of the predetermined bacterium associated with the image. The method includes a step of generating a degree of growth inhibition determination model by machine learning using the plurality of first teacher data, the step of generating a first degree of growth inhibition determination model having at least an image as input and an output of the degree of growth inhibition, and a step of generating a second degree of growth inhibition determination model by machine learning using the plurality of second teacher data, the second degree of growth inhibition determination model having at least an image as input and an output of the degree of growth inhibition, and the type of state of the negative control plate may be determined based at least on one or both of the size of the background loan and the density of the background loan in the negative control plate.

[0011] In one embodiment, the method may further include a step of inputting an image of a plate used to grow the specified bacterium into the first growth inhibition degree determination model if the condition of a negative control plate of the same test batch as the plate is the first type, and into the second growth inhibition degree determination model if the condition of a negative control plate of the same test batch as the plate is the second type, and determining the degree of growth inhibition of the specified bacterium associated with the image based on the output of the model.

[0012] In one embodiment, the method may further include the steps of inputting an image of a negative control plate of the same test batch as the plate used to grow the specified bacterium into the type determination model and determining a type of state of the negative control plate based on the output of the model, and inputting the image of the plate used to grow the specified bacterium into the first growth inhibition degree determination model if the first type is determined as the state of the negative control plate, and into the second growth inhibition degree determination model if the second type is determined as the state of the negative control plate, and obtaining the degree of growth inhibition of the specified bacterium associated with the image based on the output of the model.

[0013] In order to solve the above problem, according to an embodiment of the present disclosure, there is provided one or more growth inhibition degree determination models generated by the above method, which use an image that is to be the input of the model as an input and cause a computer to output the degree of growth inhibition that is to be the output of the model.

[0014] In order to solve the above problem, according to an embodiment of the present disclosure, there is provided a type determination model generated by the above method, which uses an image that is an input of the model as an input, and causes a computer to output a type that is an output of the model.

[0015] In order to solve the above problem, according to an embodiment of the present disclosure, a method is provided that includes the steps of: a computer receiving an image of a plate used to grow a specified bacterium; and the computer inputting the image into a growth inhibition degree determination model generated by machine learning using multiple training data, and determining the degree of growth inhibition of the specified bacterium associated with the image based on the output of the model, wherein each training data includes an image of a plate used to grow the specified bacterium and a degree of growth inhibition of the specified bacterium associated with the image, and the degree of growth inhibition included in each training data is determined based on at least one or both of the size of background loans and the density of background loans in the image included in the training data.

[0016] In one embodiment, the method further includes the steps of: the computer receiving an image of a negative control plate from the same batch as the plate that received the image; and the computer inputting the image of the negative control plate into a type determination model generated by machine learning using a plurality of separate teacher data, and determining a type of state of the negative control plate based on an output of the model, each separate teacher data including an image of the negative control plate and a type of state of the negative control plate associated with the image; and the step of determining a degree of growth inhibition of the predetermined bacterium includes, when a first type is determined as the state of the negative control plate, inputting the image of the plate used to grow the predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining a degree of growth inhibition of the predetermined bacterium based on an output of the model; and, when a second type is determined as the state of the negative control plate, inputting the image of the plate used to grow the predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining a degree of growth inhibition of the predetermined bacterium based on an output of the model. and inputting the image of the plate used to grow a specified bacterium into a second growth inhibition degree determination model generated by machine learning using training data, and determining the degree of growth inhibition of the specified bacterium based on the output of the model, wherein each of the plurality of first training data includes an image of the plate used to grow the specified bacterium, where the state of a negative control plate of the same batch as the plate is a first type, and the degree of growth inhibition of the specified bacterium associated with the image, and each of the plurality of second training data includes an image of the plate used to grow the specified bacterium, where the state of a negative control plate of the same batch as the plate is a second type, and the degree of growth inhibition of the specified bacterium associated with the image, and the type of state of the negative control plate may be determined based on at least one or both of the size of background lawns and the density of background lawns in the negative control plate.

[0017] In order to solve the above problem, according to an embodiment of the present disclosure, a system is provided that is configured to receive an image of a plate used to grow a specified bacterium, input the image into a growth inhibition degree determination model generated by machine learning using multiple training data, and determine the degree of growth inhibition of the specified bacterium associated with the image based on the output of the model, wherein each training data includes an image of a plate used to grow the specified bacterium and the degree of growth inhibition of the specified bacterium associated with the image, and the degree of growth inhibition included in each training data is determined based on at least one or both of the size of the background loan and the density of the background loan in the image included in the training data.

[0018] In one embodiment, the system is further configured to receive an image of a negative control plate from the same batch as the plate that received the image, and input the image of the negative control plate into a type determination model generated by machine learning using a plurality of separate teacher data, and determine a type of state of the negative control plate based on an output of the model, each separate teacher data including an image of the negative control plate and a type of state of the negative control plate associated with the image, and determining the degree of growth inhibition of the predetermined bacterium includes inputting the image of the plate used to grow the predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data when a first type is determined as the state of the negative control plate, and determining the degree of growth inhibition of the predetermined bacterium based on an output of the model, and inputting the image of the plate used to grow the predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data when a second type is determined as the state of the negative control plate. and inputting the image of the plate used to grow a specified bacterium into a second growth inhibition degree determination model generated by learning, and determining the degree of growth inhibition of the specified bacterium based on the output of the model, wherein each of the plurality of first teacher data includes an image of the plate used to grow the specified bacterium, where the state of a negative control plate of the same batch as the plate is of a first type, and the degree of growth inhibition of the specified bacterium associated with the image, and each of the plurality of second teacher data includes an image of the plate used to grow the specified bacterium, where the state of a negative control plate of the same batch as the plate is of a second type, and the degree of growth inhibition of the specified bacterium associated with the image, and the type of state of the negative control plate may be determined based on at least one or both of the size of background lawns and the density of background lawns in the negative control plate.

[0019] In order to solve the above problem, according to an embodiment of the present disclosure, there is provided a program that causes a computer to execute the steps of receiving an image of a plate used to grow a specified bacterium, inputting the image into a growth inhibition degree determination model generated by machine learning using multiple training data, and determining the degree of growth inhibition of the specified bacterium associated with the image based on the output of the model, wherein each training data includes an image of a plate used to grow a specified bacterium and a degree of growth inhibition of the specified bacterium associated with the image, and the degree of growth inhibition included in each training data is determined based on at least one or both of the size of the background loan and the density of the background loan in the image included in the training data.

[0020] In one embodiment, the program further causes the computer to execute the steps of receiving an image of a negative control plate of the same batch as the plate that received the image, and inputting the image of the negative control plate into a type determination model generated by machine learning using a plurality of separate teacher data, and determining a type of the state of the negative control plate based on the output of the model, each separate teacher data including an image of a negative control plate and a type of the state of the negative control plate associated with the image, and the step of determining the degree of growth inhibition of the predetermined bacterium includes, when a first type is determined as the state of the negative control plate, inputting the image of the plate used to grow the predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model, and when a second .... and inputting the image of the plate used to grow the specified bacterium into a second growth inhibition degree determination model generated by machine learning using data, and determining the degree of growth inhibition of the specified bacterium based on the output of the model, wherein each of the plurality of first teacher data includes an image of the plate used to grow the specified bacterium, where the state of a negative control plate of the same batch as the plate is a first type, and the degree of growth inhibition of the specified bacterium associated with the image, and each of the plurality of second teacher data includes an image of the plate used to grow the specified bacterium, where the state of a negative control plate of the same batch as the plate is a second type, and the degree of growth inhibition of the specified bacterium associated with the image, and the type of state of the negative control plate may be determined based on at least one or both of the size of background lawns and the density of background lawns in the negative control plate. Effect of the Invention

[0021] According to embodiments of the present disclosure, the degree of growth inhibition of bacteria can be determined reliably and quickly from images of plates used to grow the bacteria. [Brief description of the drawings]

[0022] [Figure 1] 1 is a flow chart of an example method 100 for generating a stunting degree determination model. [Diagram 2] 2 is a flow chart of a more detailed example process 200 that the step 110 of preparing the basis data may include. [Figure 3A] 13 is an exemplary image of a negative control plate. [Figure 3B] 1 is an exemplary image of a plate corresponding to TO. [Figure 3C] 1 is an exemplary image of a plate corresponding to T1. [Figure 3D] 13 is an exemplary image of a plate corresponding to T2. [Figure 3E] 13 is an exemplary image of a plate corresponding to T3. [Figure 3F] 13 is an exemplary image of a plate corresponding to T4. [Figure 4A] 13 is an exemplary image of a negative control plate. [Figure 4B] 1 is an exemplary image of a plate corresponding to TO. [Figure 4C] 1 is an exemplary image of a plate corresponding to T1. [Figure 4D] 13 is an exemplary image of a plate corresponding to T2. [Figure 4E] 13 is an exemplary image of a plate corresponding to T3. [Figure 4F] 13 is an exemplary image of a plate corresponding to T4. [Diagram 5] This is an example of basic data. [Figure 6] Here is another example of basic data. [Figure 7] 7 is a flowchart of a more detailed example process 700 that may be included in step 120 of preparing a plurality of training data based on basic data. [Figure 8] 8 is a flowchart of a more detailed example process 800 that may be included in step 130 of generating a model through machine learning using multiple training data. [Figure 9] 9 is a flow chart of an example method 900 for determining the degree of stunting. [Figure 10] FIG. 1 is a diagram illustrating an example of a computer. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] 1. Creation of a growth inhibition degree determination model FIG. 1 is a flow diagram of an example method 100 for generating a stunting degree determination model, according to one embodiment of the present invention.

[0024] Step 110 shows a step of preparing basic data. In this embodiment, a plurality of training data are required to generate a growth inhibition degree determination model, and this basic data is used to generate the training data. Step 110 may include a manual operation.

[0025] Step 120 indicates preparing a plurality of training data based on basic data. Step 120 may be executed by a computer.

[0026] Step 130 indicates a step of generating a model by machine learning using a plurality of training data. The generated model includes a growth inhibition degree determination model that receives at least an image as an input and outputs a growth inhibition degree, which will be described later. The generated model may also include a type determination model that receives at least an image as an input and outputs a type, which will be described later. Step 130 may be executed by a computer.

[0027] 2. Preparation of basic data FIG. 2 is a flow chart of a more detailed example process 200 that step 110 may include.

[0028] 210 shows a step of attempting to grow a predetermined bacterium. Step 210 includes a step of attempting to grow a predetermined bacterium multiple times. The predetermined bacterium is preferably the same species of bacterium as the bacterium whose degree of growth inhibition is to be determined using the embodiments of the present disclosure. In each attempt, multiple plates are used, each containing the predetermined bacterium and exposed to different doses of the substance, and a negative control plate containing the predetermined bacterium but not exposed to the substance. In each attempt, preferably, two or more negative control plates are used, and preferably two or more plates are used for each dose of each substance. Hereinafter, one such attempt is referred to as a "test batch."

[0029] In detail, each test batch may include the following steps (1) to (7), but the steps included in each test batch are not limited thereto.

[0030] (1) For a given bacterium, prepare a bacterial suspension grown from a single colony and freeze it in small aliquots (hereinafter referred to as "frozen bacterial solution").

[0031] (2) After thawing the frozen bacterial solution, a certain amount of the solution was inoculated into a nutrient broth culture medium, and the culture was stopped at the beginning of the stationary phase at 37°C under appropriate culture conditions to prevent further growth. This is the "preculture medium." The viable cell count was approximately 1.0 x 10 9 / ml. At this point, the viable cell count is 1.0 x 10 9 If the concentration is significantly less than 1 / ml, the preculture medium must be prepared again.

[0032] (3) Place a solution of the test substance (the substance to be exposed to the specified bacteria) in a sterile test tube.

[0033] (4) Add 0.5 ml of 0.1 M sodium phosphate buffer (pH 7.4) to (3), then add 0.1 ml of preculture solution and mix well (when not using a metabolic activation system). Alternatively, add 0.5 ml of S9mix to (3), then add 0.1 ml of preculture solution and mix well (when using a metabolic activation system). The composition of S9mix per ml is as follows, and it is preferable that it is used within 6 months of manufacture. S9 fraction: 10~30% (0.1~0.3ml) MgCl2: 8 μmol KCl: 33 μmol Glucose 6-phosphate: 5 μmol NADPH: 4 μmol NADH: 4 μmol Sodium phosphate buffer (pH 7.4): 100 μmol

[0034] (5) Preincubate (4) at 37°C for a certain period of time with shaking (in the case of the preincubation method). Alternatively, do not carry out any additional steps on (4) (in the case of the plate method).

[0035] (6) Add 2 ml of top agar to (5) and mix thoroughly. Top agar in this disclosure refers to an agar liquid for forming a soft agar medium containing only small amounts of specific amino acids described below.

[0036] (7) Pour (6) onto a plate (minimal glucose agar plate medium) and spread evenly.

[0037] (8) (7) is incubated at 37°C for at least 48 hours.

[0038] The above-mentioned "plurality of plates each containing a predetermined bacterium and exposed to different doses of the substance" can be prepared by using a plurality of solutions of different concentrations in step (3). The above-mentioned "negative control plate" can be prepared by adding the solvent used in preparing the solution of the test substance in step (3).

[0039] 220 shows a step of acquiring an image of each plate on which bacteria was grown for each test batch (which may include a negative control plate). The acquired image may be an image of each plate captured by a digital camera and subjected to various image processing.

[0040] 230 shows a step of determining the degree of growth inhibition of each plate of each test batch based on the condition of the negative control plate. This step may be a step of a human visually determining the degree of growth inhibition. This step may also be a step of a human visually determining the degree of growth inhibition using a microscope.

[0041] The degree of growth inhibition may be determined based on one or both of the size and density of background lawns on each plate, and may further be determined based on the shape of the background lawns on each plate.

[0042] The degree of growth inhibition is preferably classified into three or more categories, rather than a binary choice of "there is growth inhibition" or "there is no growth inhibition." For example, the degree of growth inhibition may be classified into five categories, such as T0 (same as the negative control plate), T1 (growth inhibition is observed), T2 (growth inhibition is greater than T1), T3 (growth inhibition is greater than T2), and T4 (all bacteria are killed). Figures 3A-3F and 4A-4F show examples of images corresponding to the negative control plate (SC) and T0-T4, respectively, obtained for different test batches. The grains seen in Figures 3A-3E and 4A-4E (images of SC and T0-T3) correspond to background lawns.

[0043] The above five exemplary classifications will be explained in more detail. Generally, bacteria synthesize the amino acids necessary for growth by themselves, but the bacteria used in the Ames test have been genetically modified so that they are unable to synthesize some of the amino acids (hereinafter referred to as "specific amino acids") necessary for growth. When such bacteria (hereinafter referred to as "amino acid auxotrophic bacteria") are grown on a plate containing specific amino acids, they will not grow any more in principle when the specific amino acids contained in the plate are exhausted, resulting in the formation of tiny colonies (background colonies) through several divisions. At this time, if some of the bacteria die during the growth process, the amount of specific amino acids available per bacteria on the plate increases for the remaining bacteria, so the size of the background colonies increases, while the number of background colonies decreases, and therefore the density of the background colonies (which may be defined as the number of background colonies per unit area) decreases. It can be understood that Figures 3B to 3E and Figures 4B to 4E (images T0 to T3) are arranged in order of the probability of bacterial death, i.e., the degree of growth inhibition, and therefore the size of the background colonies increases and the density decreases in order. However, when the probability of bacterial death exceeds a certain threshold, the bacteria are completely killed during the growth process, and the background lawn no longer exists. Figures 3F and 4F (image T4) are images of bacteria completely killed during the growth process, and it can be seen that no background lawn is observed. Note that when bacteria are exposed to a substance having cytotoxicity, the probability of the bacteria dying generally increases with increasing concentration of the substance to which the bacteria are exposed.

[0044] The degree of growth inhibition is determined relatively based on the condition of the negative control plate. For example, the size and density of the background lawn in Figure 3C (T1 image) and Figure 4C (T1 image) are different, but both are determined to be T1 based on the condition of the corresponding negative control plate (Figure 3A and Figure 4A (SC image).

[0045] Returning to FIG. 2, 240 indicates a step of determining the type of state of each negative control plate. Step 240 may be a step of classifying the state of each negative control plate into two or more types based on one or both of the size and density of the background lawns on the plate. Note that step 240 may be a step of classifying the state of each negative control plate into two or more types based further on the shape of the background lawns on the plate. For the sake of explanation, it is assumed in step 240 that the state of each negative control plate is classified into two types, a first type in which the size of the background lawns on the plate is relatively small or the density of the background lawns is relatively large, and a second type in which the size of the background lawns on the plate is relatively small or the density of the background lawns is relatively large, and the second type in which the size of the background lawns on the plate is relatively small or the density of the background lawns is relatively large, but the number of types of the state of the negative control plate (the number of types to be classified) and the classification method are not limited to this.

[0046] Reference numeral 250 denotes a step of generating basic data. The basic data is obtained by organizing the information obtained by the above steps.

[0047] More specifically, the basic data may be data in which the image of the plate (excluding the image of the negative control plate) acquired in step 220 is stored in association with the degree of growth inhibition of the bacteria attempted to grow on the plate determined in step 230 (hereinafter referred to as the "corresponding degree of growth inhibition"). In the basic data, the image of the plate (excluding the image of the negative control plate) acquired in step 220 or the degree of growth inhibition of the bacteria attempted to grow on the plate determined in step 230 may be stored in further association with the type of state of the negative control plate (hereinafter referred to as the "corresponding negative control plate") of the same test batch as the plate acquired in step 240. FIG. 5 shows an example of basic data recorded in the form of a table. The first column in FIG. 5 is a path to the image of the plate on which the growth of the bacteria was attempted, the second column is the corresponding degree of growth inhibition, and the third column is the type of state of the negative control plate of the same test batch as the plate. Note that the third column may be omitted.

[0048] The basic data may include data in which the image of the negative control plate acquired in step 220 is stored in association with the type of state of the negative control plate acquired in step 240 (hereinafter referred to as the "corresponding type"). Fig. 6 shows another example of basic data stored in the form of a table. The first column in Fig. 6 is a path to the image of the negative control plate, and the second column shows the type of state of the negative control plate.

[0049] 3. Preparing training data and generating models 3-1 When not using the type of negative control plate condition Below, an example of preparation of training data and generation of a model will be described in which the type of state of the negative control plate is not used.

[0050] The step 120 of preparing a plurality of teacher data based on the basic data may include a step of extracting, from the basic data, a pair of an image of a plate on which bacterial growth was attempted and a degree of growth inhibition of the bacteria. Thus, each of the plurality of teacher data may include an image of a plate on which bacterial growth was attempted and a degree of growth inhibition of the bacteria.

[0051] In step 130, a model is generated by machine learning using multiple teaching data, and a growth inhibition degree determination model is generated. In step 130, any machine learning method can be used that can generate a model that takes an image as input and outputs a solution to a regression problem or a classification problem.

[0052] An example of a model that outputs the solution to a regression problem is one that outputs a numerical value when an image is input. In machine learning of a model that outputs the solution to a regression problem, learning may be performed so as to reduce the error between the output when an image included in each training data is input and the quantified degree of growth inhibition included in each training data (for example, "T0" can be quantified as "0", "T1" can be quantified as "1", etc.).

[0053] An example of a model that outputs the solution to a classification problem is one that outputs the probability associated with each option among multiple options (or categories or classes) when an image is input. In machine learning that outputs the solution to a classification problem, learning may be performed so that when an image included in each training data is input, the probability associated with the option corresponding to the degree of growth inhibition included in each training data as output increases.

[0054] The growth inhibition degree determination model generated in step 130 includes parameters, weights, and the like specific to the model obtained as a result of training an arbitrary model as described above.

[0055] 3-2 When using the type of condition of the negative control plate Below, an example of preparing training data and generating a model using the state type of the negative control plate will be described.

[0056] FIG. 7 is a flowchart of a more detailed example process 700 that may be included in step 120 of preparing a plurality of training data based on basic data.

[0057] 710 shows a step of preparing a plurality of first teacher data based on basic data. Step 710 may be a step of extracting, as a plurality of first teacher data, a pair of an image of a plate in which the corresponding negative control plate state is of the first type and a corresponding degree of growth inhibition from the basic data. Thus, each of the plurality of first teacher data may include an image of a plate on which bacteria growth has been attempted, in which the corresponding negative control plate state is of the first type, and a degree of growth inhibition of the bacteria.

[0058] 720 shows a step of preparing a plurality of second teacher data based on the basic data. Step 720 may be a step of extracting, as the plurality of second teacher data, a pair of an image of a plate in which the corresponding negative control plate state is the second type and a corresponding degree of growth inhibition from the basic data. Thus, each of the plurality of second teacher data may include an image of a plate on which bacteria growth has been attempted, in which the corresponding negative control plate state is the second type, and a degree of growth inhibition of the bacteria.

[0059] 730 shows a step of preparing a plurality of third teacher data based on the basic data. Step 730 may be a step of extracting a pair of an image of a negative control plate and a corresponding type from the basic data as the plurality of third teacher data. Thus, each of the plurality of third teacher data may include one image of a negative control plate and a type of the state of the negative control plate.

[0060] The order of execution of steps 710 to 730 is not limited to this, and step 730 may be omitted. In addition, in the example process 700, two sets of teacher data (a plurality of first teacher data and a plurality of second teacher data) based on different types of the condition of the negative control plate are prepared, however, one or more further sets of teacher data associated with one or more further types of the condition of the negative control plate may be prepared.

[0061] FIG. 8 is a flowchart of a more detailed example process 800 that may be included in step 130 of generating a model through machine learning using multiple training data.

[0062] 810 shows a step of generating a first growth inhibition degree determination model by machine learning using a plurality of first teacher data. 820 shows a step of generating a second growth inhibition degree determination model by machine learning using a plurality of second teacher data. In steps 810 and 820, any machine learning method capable of generating a model in which an image is input and a solution to a regression problem or a classification problem is output can be used. This is similar to the above-mentioned example in which the type of state of the negative control plate is not used. Note that the first growth inhibition degree determination model is more suitable for determining the growth inhibition degree of bacteria whose corresponding state type of the negative control plate is the first type, and the second growth inhibition degree determination model is more suitable for determining the growth inhibition degree of bacteria whose corresponding state type of the negative control plate is the second type.

[0063] 830 shows a step of generating a type determination model for determining the type of the state of the negative control plate by machine learning using the plurality of third training data.

[0064] Any machine learning technique can be used in step 830 that takes images as input and is capable of generating a model whose output is the solution to a regression or classification problem.

[0065] In machine learning of a model that outputs the solution to a regression problem, learning may be performed so as to minimize the error between the output when an image contained in each training data is input and the numerical representation of the classification contained in each training data (for example, "Type 1" can be quantified as "1" and "Type 2" can be quantified as "2").

[0066] In machine learning of a model that outputs the solution to a classification problem, learning may be performed so that when an image contained in each training data is input, the probability of the option corresponding to the type contained in each training data as the output is increased.

[0067] The generated type determination model includes parameters, weights, etc. specific to the model obtained as a result of training the arbitrary model as described above.

[0068] Furthermore, the order of execution of steps 810 to 830 is not limited to this, and step 830 may be omitted.

[0069] Furthermore, in the example process 800, two growth inhibition degree determination models (a first growth inhibition degree determination model and a second growth inhibition degree determination model) are generated based on different sets of training data, but one or more further growth inhibition degree determination models may be generated based on one or more further sets of training data.

[0070] 4. Determination of the degree of growth inhibition 4-1 When not using the type of condition of the negative control plate An example of determining the degree of growth inhibition without using the type of condition on the negative control plate will be described below.

[0071] An exemplary method for determining the degree of growth inhibition according to one embodiment of the present disclosure includes a computer receiving an image of a plate used to grow a given bacterium, and the computer inputting the received image into the growth inhibition degree determination model and determining the degree of growth inhibition of the given bacterium based on the output of the model.

[0072] In addition, when the degree of growth inhibition is quantified as an integer, the growth inhibition degree determination model may output a real number. In such a case, the quantified integer that is closest to the output real number can be determined as the degree of growth inhibition.

[0073] The growth inhibition degree determination model may output a probability associated with each of a plurality of growth inhibition degrees. In such a case, the degree of growth inhibition with the highest probability may be determined as the growth inhibition degree.

[0074] According to another embodiment of the present disclosure, an exemplary system for determining the degree of growth inhibition is a system configured to receive images of plates used to grow a given bacterium, input the received images into the growth inhibition degree determination model, and determine the degree of growth inhibition of the given bacterium based on the output of the model.

[0075] It will be appreciated that the above-described methods and systems can be realized by a combination of a computer, which is a hardware resource, and software.

[0076] From another perspective, an exemplary program for determining the degree of growth inhibition according to another embodiment of the present disclosure is a program that causes a computer to execute the steps of receiving an image of a plate used to grow a specified bacterium, inputting the received image into the growth inhibition degree determination model, and determining the degree of growth inhibition of the specified bacterium based on the output of the model.

[0077] In addition, training data was prepared based on a portion of the basic data, and the above-mentioned growth inhibition degree determination model was generated using a machine learning method called Extreme Gradient Boosted Trees Regressor in the DataRobot service provided by DataRobot, Inc., and the accuracy rate was evaluated for test data (pairs of plate images and the corresponding degrees of growth inhibition) created based on the remaining basic data, resulting in a result of approximately 75.7%.

[0078] 4-2 When using the type of condition of the negative control plate An example of determining the degree of growth inhibition using the type of condition on the negative control plate will be described below.

[0079] 9 is a flow chart of an exemplary method 900 for determining the degree of stunting, according to one embodiment of the present disclosure, wherein each step of the method 900 is computer-implemented.

[0080] 910 shows the step of receiving an image of the plate used to grow a given bacterium (hereinafter referred to as the "plate image") and an image of a negative control plate from the same batch (hereinafter referred to as the "SC image").

[0081] 920 shows the step of inputting the SC image into the type determination model and determining the type of state of the negative control plate based on the output of the model.

[0082] In addition, when the type of the state of the negative control plate is quantified as an integer, the type determination model may output a real number. In such a case, the type to be determined may be the one whose quantified integer is closest to the output real number.

[0083] The type determination model may also output a probability for each of the multiple types, in which case the type with the highest probability may be the type determined.

[0084] 930 indicates a step of determining whether the type determined in step 920 is the first type. If it is determined to be the first type, the process proceeds to step 940, and if not, the process proceeds to step 950. Note that step 930 is for the case where the number of types of the negative control plate is two (the first type and type). If the number of types of the negative control plate is three or more, step 930 will be a step of determining which of the three or more types the type determined in step 920 is, and branching the process to three or more types.

[0085] 940 indicates a step of inputting a plate image into the first growth inhibition degree determination model, thereby outputting the degree of growth inhibition of the predetermined bacterium. Step 940 is a step of inputting an image of the plate used to grow the predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model, when the first type is determined as the state of the negative control plate.

[0086] 950 indicates a step of inputting a plate image into the second growth inhibition degree determination model, thereby outputting the degree of growth inhibition of the predetermined bacterium. Step 950 is a step of inputting an image of the plate used to grow the predetermined bacterium into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model, when the second type is determined as the state of the negative control plate.

[0087] According to another embodiment of the present disclosure, an exemplary system for determining the degree of growth inhibition is a system configured to receive a plate image and an SC image, input the SC image to the type determination model, and determine the type of the state of the negative control plate based on the output of the model; if a first type is determined as the state of the negative control plate, input an image of the plate used to grow a specified bacterium to a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determine the degree of growth inhibition of the specified bacterium based on the output of the model; if a second type is determined as the state of the negative control plate, input an image of the plate used to grow a specified bacterium to a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and determine the degree of growth inhibition of the specified bacterium based on the output of the model.

[0088] It will be appreciated that the above-described methods and systems can be realized by a combination of a computer, which is a hardware resource, and software.

[0089] From another perspective, an exemplary program for determining the degree of growth inhibition according to another embodiment of the present disclosure is a program that causes a computer to execute the following steps: receiving a plate image and an SC image; inputting the SC image into the type determination model and determining the type of state of the negative control plate based on the output of the model; if the first type is determined as the state of the negative control plate, inputting an image of the plate used to grow a specified bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data and determining the degree of growth inhibition of the specified bacterium based on the output of the model; and if the second type is determined as the state of the negative control plate, inputting an image of the plate used to grow a specified bacterium into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data and determining the degree of growth inhibition of the specified bacterium based on the output of the model.

[0090] In addition, training data was prepared based on a portion of the basic data, and the above-mentioned type determination model, first growth inhibition degree determination model, and second growth inhibition degree determination model were generated using a machine learning method called Extreme Gradient Boosted Trees Regressor in a service called DataRobot provided by DataRobot, Inc., and the accuracy rate was evaluated for test data (pairs of plate images and corresponding degrees of growth inhibition) created based on the remaining basic data.The results were 91.8% for plate images where the corresponding negative control plate state was the first type, and 100% for plate images where the corresponding negative control plate state was the second type.

[0091] 5. Computers An example of a hardware configuration of a computer that can be used to implement one embodiment of the present invention will be described below. The computer that can be used to implement one embodiment of the present invention may be any computer, such as a personal computer or a computer on a cloud. The computer that can be used to implement one embodiment of the present invention may be a GPU or may include a GPU.

[0092] Fig. 10 shows an example of a computer hardware configuration. As shown in the figure, the computer 1000 mainly comprises, as hardware resources, a processor 1010, a main memory device 1020, an auxiliary memory device 1030, an input / output interface 1040, and a communication interface 1050, which are connected to each other via a bus line 1060 including an address bus, a data bus, a control bus, etc. Note that an interface circuit (not shown) may be interposed between the bus line 1060 and each hardware resource as appropriate.

[0093] The processor 1010 is a device that controls the entire computer or at least a part of the computer, such as a CPU or a microprocessor. Note that one computer may include multiple processors 1010. In such a case, the term "processor" may be a general term for the multiple processors 1010.

[0094] The main memory 1020 is a volatile memory such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), which provides a working area for the processor 1010.

[0095] The auxiliary storage device 1030 is a non-volatile memory such as an HDD, SSD, or flash memory that stores software programs and data. The programs and data are loaded from the auxiliary storage device 1030 to the main storage device 1020 via a bus line 1060 at any time. The auxiliary storage device 1030 may be referred to as a computer-readable storage medium, a non-transitory computer-readable storage medium, or a computer-readable storage medium. The programs include instructions that cause the processor to execute desired processes.

[0096] The input / output interface 1040 presents information and / or receives input of information, and may be a digital camera, a keyboard, a mouse, a display, a touch panel display, a microphone, a speaker, various sensors, or the like.

[0097] The communication interface 1050 is connected to a network 1055 consisting of one or more of the Internet, a local area network (LAN), etc., and transmits and receives data via the network 1055. The communication interface 1050 and the network 1055 can be connected by wire or wirelessly. The communication interface 1050 may also obtain information related to the network, such as information related to Wi-Fi access points and information related to the base stations of a communication carrier.

[0098] It will be apparent to those skilled in the art that the above-exemplified hardware resources and software can cooperate to cause the computer 1000 to function as desired means, execute desired steps, and achieve desired functions.

[0099] 6. Conclusion Although some examples of the embodiments of the present invention have been described above, it should be understood that these are merely illustrative and do not limit the technical scope of the present invention. It should be understood that modifications, additions, improvements, etc. of the embodiments can be appropriately made without departing from the spirit and scope of the present disclosure. The technical scope of the present invention should not be limited by any of the above-described embodiments, but should be defined only by the claims and their equivalents.

[0100] Finally, some of the features of the present disclosure are described below.

[0101] [Feature 1] A step of preparing a plurality of teacher data, each of the teacher data being: Images of plates used to grow a given bacterium; The degree of growth inhibition of the specified bacteria related to the image; and generating a growth inhibition degree determination model that uses at least an image as an input and outputs a degree of growth inhibition by machine learning using the plurality of training data; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of the background loan and the density of the background loan in the image included in the training data. method.

[0102] [Feature 2] The method according to feature 1, inputting an image of the plate used to grow the selected bacterium into the model for determining the degree of growth inhibition and determining the degree of growth inhibition of the selected bacterium associated with the image based on the output of the model; The method further comprises:

[0103] [Feature 3] 3. The method according to claim 1 or 2, The degree of growth inhibition contained in each teaching data was determined based on the state of the negative control plate of the same test batch as the plate of the image contained in the teaching data. method.

[0104] [Feature 4] 4. The method according to any one of features 1 to 3, further comprising: The step of preparing a plurality of training data includes: A step of preparing a plurality of first teacher data, each of the first teacher data being: an image of a plate used to grow the bacteria of interest, the image showing a negative control plate of the same batch as the plate in a first type of condition; The degree of growth inhibition of the specified bacteria related to the image; and A step of preparing a plurality of second teacher data, each of the second teacher data being: an image of the plate used to grow the selected bacteria, the image showing a second type of condition of a negative control plate from the same batch as the plate; The degree of growth inhibition of the specified bacteria related to the image; Steps Including, The step of generating a growth inhibition degree determination model includes: A step of generating a first growth inhibition degree determination model that uses at least an image as an input and outputs a degree of growth inhibition by machine learning using the plurality of first teacher data; A step of generating a second growth inhibition degree determination model that uses an image as at least an input and outputs a degree of growth inhibition by machine learning using the plurality of second teacher data. Including, The type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate. method.

[0105] [Feature 5] The method according to feature 4, inputting an image of a plate used to grow the predetermined bacterium into the first growth inhibition degree determination model if the state of a negative control plate of the same test batch as the plate is the first type, and into the second growth inhibition degree determination model if the state of a negative control plate of the same test batch as the plate is the second type, and determining the degree of growth inhibition of the predetermined bacterium related to the image based on the output of the model; The method further comprises:

[0106] [Feature 6] 6. The method according to claim 4 or 5, A step of preparing a plurality of third teacher data, each of the third teacher data being: Images of the negative control plate; The type of condition of the negative control plate associated with the image; and and generating a type determination model that uses at least an image as an input and a type as an output by machine learning using the plurality of third teacher data; The method further comprises:

[0107] [Feature 7] Feature 6. The method according to feature 6, inputting an image of a negative control plate of the same test batch as the plate used to grow the selected bacterium into the typing model and determining a type of state of the negative control plate based on the output of the model; inputting an image of the plate used to grow the predetermined bacterium into the first growth inhibition degree determination model when the first type is determined as the state of the negative control plate, and into the second growth inhibition degree determination model when the second type is determined as the state of the negative control plate, and acquiring the degree of growth inhibition of the predetermined bacterium related to the image based on the output of the model; The method further comprises:

[0108] [Feature 8] One or more growth inhibition degree determination models generated by the method described in feature 1, 3 or 4, which use an image that is to be an input of the model as an input, and cause a computer to output a degree of growth inhibition that is to be an output of the model.

[0109] [Feature 9] A type determination model generated by the method according to feature 6, the model being configured to output a type to a computer by using an image to be an input of the model as an input.

[0110] [Feature 10] receiving by a computer an image of a plate used to grow a given bacterium; a step of inputting the image into a growth inhibition degree determination model generated by machine learning using a plurality of training data by the computer, and determining a degree of growth inhibition of the predetermined bacterium related to the image based on an output of the model; A method comprising: Each teacher data is an image of the plate used to grow the selected bacteria; The degree of growth inhibition of the specified bacteria related to the image; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of the background loan and the density of the background loan in the image included in the training data. method.

[0111] [Feature 11] 11. The method according to claim 10, further comprising: said computer receiving an image of a negative control plate of the same batch as said plate that has been imaged; The computer inputs the image of the negative control plate into a type determination model generated by machine learning using a plurality of different training data, and determines the type of the state of the negative control plate based on the output of the model. Further comprising: Each teacher data is Images of the negative control plate; The type of condition of the negative control plate associated with the image; and Including, The step of determining the degree of growth inhibition of the predetermined bacteria comprises: When the state of the negative control plate is determined to be a first type, inputting the image of the plate used to grow a predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model; When the second type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model. Including, Each of the plurality of first teacher data is an image of a plate used to grow the bacteria of interest, the image showing a negative control plate of the same batch as the plate in a first type of condition; The degree of growth inhibition of the specified bacteria related to the image; Including, Each of the plurality of second teacher data is an image of a plate used to grow the selected bacteria, the image showing a second type of condition of a negative control plate from the same batch as the plate; The degree of growth inhibition of the specified bacteria related to the image; Including, The type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate. method.

[0112] [Feature 12] Receive images of plates used to grow a given bacterium, The image is input to a growth inhibition degree determination model generated by machine learning using multiple training data, and the degree of growth inhibition of the predetermined bacterium related to the image is determined based on the output of the model. A system configured as follows: Each teacher data is an image of the plate used to grow the selected bacteria; The degree of growth inhibition of the specified bacteria related to the image; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of the background loan and the density of the background loan in the image included in the training data. system.

[0113] [Feature 13] 13. The system according to claim 12, receiving an image of a negative control plate from the same batch as the plate that has been imaged; The image of the negative control plate is input to a type determination model generated by machine learning using a plurality of different training data, and the type of the state of the negative control plate is determined based on the output of the model. The method further comprises: Each teacher data is Images of the negative control plate; The type of condition of the negative control plate associated with the image; and Including, Determining the degree of growth inhibition of the predetermined bacteria comprises: When the state of the negative control plate is determined to be a first type, inputting the image of the plate used to grow a predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model; When the second type is determined as the state of the negative control plate, the image of the plate used to grow a predetermined bacterium is inputted into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and a degree of growth inhibition of the predetermined bacterium is determined based on an output of the model. Including, Each of the plurality of first teacher data is an image of a plate used to grow the bacteria of interest, the image showing a negative control plate of the same batch as the plate in a first type of condition; The degree of growth inhibition of the specified bacteria related to the image; Including, Each of the plurality of second teacher data is an image of a plate used to grow the selected bacteria, the image showing a second type of condition of a negative control plate from the same batch as the plate; The degree of growth inhibition of the specified bacteria related to the image; Including, The type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate. system.

[0114] [Feature 14] On the computer, receiving an image of a plate used to grow a given bacterium; inputting the image into a growth inhibition degree determination model generated by machine learning using a plurality of training data, and determining the degree of growth inhibition of the predetermined bacterium related to the image based on the output of the model; A program for executing Each teacher data is Images of plates used to grow a given bacterium; The degree of growth inhibition of the specified bacteria related to the image; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of the background loan and the density of the background loan in the image included in the training data. program.

[0115] [Feature 15] 15. The program according to claim 14, further comprising: receiving an image of a negative control plate from the same batch as the plate that has been imaged; inputting the image of the negative control plate into a type determination model generated by machine learning using a plurality of different training data, and determining the type of the state of the negative control plate based on the output of the model; Then, Each teacher data is Images of the negative control plate; The type of condition of the negative control plate associated with the image; and Including, The step of determining the degree of growth inhibition of the predetermined bacteria comprises: When the state of the negative control plate is determined to be a first type, inputting the image of the plate used to grow a predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model; When the second type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model. Including, Each of the plurality of first teacher data is an image of a plate used to grow the bacteria of interest, the image showing a negative control plate of the same batch as the plate in a first type of condition; The degree of growth inhibition of the specified bacteria related to the image; Including, Each of the plurality of second teacher data is an image of a plate used to grow the selected bacteria, the image showing a second type of condition of a negative control plate from the same batch as the plate; The degree of growth inhibition of the specified bacteria related to the image; Including, The type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate. program. [Explanation of symbols]

[0116] 100...Exemplary method for generating a growth inhibition degree determination model 200...More detailed example processes that the step of preparing basic data may include 700...More detailed example processing that may be included in the step of preparing multiple teaching data based on basic data 800...More detailed example processing that may include a step of generating a model by machine learning using multiple training data 900...Exemplary methods for determining the degree of growth inhibition 1055…Network 1060…Bus line

Claims

1. A step of preparing a plurality of training data, each of which includes: Images of plates used to grow a given bacterium; the degree of growth inhibition of the predetermined bacteria in the image; and generating a growth inhibition degree determination model that receives at least an image as an input and outputs the degree of growth inhibition by machine learning using the plurality of training data; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of background loans and the density of background loans in the image included in the training data. method.

2. 10. The method of claim 1, inputting an image of the plate used to grow the predetermined bacterium into the growth inhibition degree determination model, and determining the degree of growth inhibition of the predetermined bacterium associated with the image based on the output of the model; The method further comprises:

3. 10. The method of claim 1, The degree of growth inhibition contained in each training data is determined based on the state of the negative control plate of the same test batch as the plate in the image contained in the training data. method.

4. 10. The method of claim 1, The step of preparing a plurality of training data includes: A step of preparing a plurality of first teacher data, each of the first teacher data being: an image of a plate used to grow the selected bacteria, the image being of a negative control plate from the same batch as the plate, in a first type of condition; the degree of growth inhibition of the predetermined bacteria in the image; and A step of preparing a plurality of second teacher data, each of the second teacher data being: an image of a plate used to grow the selected bacteria, wherein the negative control plate from the same batch as the plate is in a second type of condition; and the degree of growth inhibition of the predetermined bacteria in the image; Including steps and Including, The step of generating a growth stunt degree determination model comprises: A step of generating a first growth inhibition degree determination model that uses an image as at least an input and outputs a degree of growth inhibition by machine learning using the plurality of first teacher data; generating a second growth inhibition degree determination model that receives at least an image as an input and outputs a degree of growth inhibition by machine learning using the plurality of second teacher data; Including, the type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate; method.

5. 5. The method of claim 4, inputting an image of a plate used to grow the predetermined bacterium into the first growth inhibition degree determination model if the state of a negative control plate of the same test batch as the plate is the first type, and into the second growth inhibition degree determination model if the state of a negative control plate of the same test batch as the plate is the second type, and determining the degree of growth inhibition of the predetermined bacterium related to the image based on the output of the model; The method further comprises:

6. 5. The method of claim 4, A step of preparing a plurality of third teacher data, each of the third teacher data being: Images of the negative control plate; The type of condition of the negative control plate associated with the image; and and generating a type determination model that receives at least an image as an input and outputs a type by machine learning using the plurality of third teacher data; The method further comprises:

7. 7. The method of claim 6, inputting an image of a negative control plate from the same test batch as the plate used to grow the selected bacterium into the typing model and determining the type of condition of the negative control plate based on the output of the model; inputting an image of the plate used to grow the predetermined bacterium into the first growth inhibition degree determination model if the first type is determined as the state of the negative control plate, and into the second growth inhibition degree determination model if the second type is determined as the state of the negative control plate, and obtaining the degree of growth inhibition of the predetermined bacterium related to the image based on the output of the model; The method further comprises:

8. One or more growth inhibition degree determination models generated by the method described in claim 1, 3 or 4, which use an image that is to be the input of the model as input and output the degree of growth inhibition that is to be the output of the model to a computer.

9. 7. A type determination model generated by the method of claim 6, which causes a computer to output a type that is an output of the model by using an image that is an input of the model as an input.

10. receiving, by a computer, an image of a plate used to grow a predetermined bacterium; a step in which the computer inputs the image into a growth inhibition degree determination model generated by machine learning using a plurality of training data, and determines the degree of growth inhibition of the predetermined bacterium related to the image based on the output of the model; A method comprising: Each teacher data is an image of the plate used to grow the selected bacteria; the degree of growth inhibition of the predetermined bacteria in the image; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of background loans and the density of background loans in the image included in the training data. method.

11. 11. The method of claim 10, receiving by the computer an image of a negative control plate from the same batch as the plate that has been imaged; the computer inputs the image of the negative control plate into a type determination model generated by machine learning using a plurality of different training data, and determines the type of state of the negative control plate based on the output of the model; Further comprising: Each teacher data is Images of the negative control plate; The type of condition of the negative control plate associated with the image; and Including, The step of determining the degree of growth inhibition of the predetermined bacteria comprises: When the first type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model; When the second type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model. Including, Each of the plurality of first teacher data is an image of a plate used to grow the selected bacteria, the image being of a negative control plate from the same batch as the plate, in a first type of condition; the degree of growth inhibition of the predetermined bacteria in the image; Including, Each of the plurality of second teacher data is an image of a plate used to grow the selected bacteria, wherein the negative control plate from the same batch as the plate is in a second type of condition; and the degree of growth inhibition of the predetermined bacteria in the image; Including, the type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate; method.

12. Receive images of plates used to grow a given bacterium, The image is input into a growth inhibition degree determination model generated by machine learning using multiple training data, and the degree of growth inhibition of the predetermined bacterium associated with the image is determined based on the output of the model. A system configured as follows: Each teacher data is an image of the plate used to grow the selected bacteria; the degree of growth inhibition of the predetermined bacteria in the image; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of background loans and the density of background loans in the image included in the training data. system.

13. 13. The system of claim 12, receiving an image of a negative control plate from the same batch as the imaged plate; The image of the negative control plate is input into a type determination model generated by machine learning using a plurality of different training data, and the type of the state of the negative control plate is determined based on the output of the model. further configured as follows: Each teacher data is Images of the negative control plate; The type of condition of the negative control plate associated with the image; and Including, Determining the degree of growth inhibition of the predetermined bacteria comprises: When the first type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model; When the second type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model. Including, Each of the plurality of first teacher data is an image of a plate used to grow the selected bacteria, the image being of a negative control plate from the same batch as the plate, in a first type of condition; the degree of growth inhibition of the predetermined bacteria in the image; Including, Each of the plurality of second teacher data is an image of a plate used to grow the selected bacteria, wherein the negative control plate from the same batch as the plate is in a second type of condition; and the degree of growth inhibition of the predetermined bacteria in the image; Including, the type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate; system.

14. On the computer, receiving an image of a plate used to grow a given bacterium; inputting the image into a growth inhibition degree determination model generated by machine learning using a plurality of training data, and determining the degree of growth inhibition of the predetermined bacterium related to the image based on the output of the model; A program for executing Each teacher data is Images of plates used to grow a given bacterium; the degree of growth inhibition of the predetermined bacteria in the image; Including, The degree of growth inhibition included in each training data is determined based on at least one or both of the size of background loans and the density of background loans in the image included in the training data. program.

15. 15. The program according to claim 14, wherein the computer: receiving an image of a negative control plate from the same batch as the imaged plate; inputting the image of the negative control plate into a type determination model generated by machine learning using a plurality of different training data, and determining the type of state of the negative control plate based on the output of the model; Then run Each teacher data is Images of the negative control plate; The type of condition of the negative control plate associated with the image; and Including, The step of determining the degree of growth inhibition of the predetermined bacteria comprises: When the first type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a first growth inhibition degree determination model generated by machine learning using a plurality of first teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model; When the second type is determined as the state of the negative control plate, inputting the image of the plate used to grow a predetermined bacterium into a second growth inhibition degree determination model generated by machine learning using a plurality of second teacher data, and determining the degree of growth inhibition of the predetermined bacterium based on the output of the model. Including, Each of the plurality of first teacher data is an image of a plate used to grow the selected bacteria, the image being of a negative control plate from the same batch as the plate, in a first type of condition; the degree of growth inhibition of the predetermined bacteria in the image; Including, Each of the plurality of second teacher data is an image of a plate used to grow the selected bacteria, wherein the negative control plate from the same batch as the plate is in a second type of condition; and the degree of growth inhibition of the predetermined bacteria in the image; Including, the type of condition of the negative control plate is determined based on at least one or both of the size of background lawns and the density of background lawns on the negative control plate; program.

Images