Drug development support device, method for operating drug development support device, and program for operating drug development support device

Abstract

The present invention comprises a processor. The processor: acquires a first segment image in which a tissue segment of at least one subject among a plurality of subjects constituting an administration group to which a candidate substance for a drug has been administered is image-captured; extracts a first feature amount from the first segment image; acquires a plurality of second segment images obtained by image-capturing tissue segments of a plurality of subjects constituting a control group to which the candidate substance for the drug has not been administered, the plurality of subjects having the same attributes as the subjects constituting the administration group and being placed in the same breeding environment; extracts a second feature amount from each of the plurality of second section images; acquires third feature amount information pertaining to a third feature amount extracted from each of the plurality of third segment images obtained by image-capturing tissue segments of a plurality of subjects constituting a control group to which a candidate substance has not been administered in a past evaluation test; and determines whether or not there is an abnormality in the tissue segment shown in the first segment image on the basis of the first feature amount, the second feature amount, and the third feature amount information.

Description

Drug discovery support device, drug discovery support device operation method, and drug discovery support device operation program

[0001] The technology of the present disclosure relates to a drug discovery support device, an operating method for the drug discovery support device, and an operating program for the drug discovery support device.

[0002] In the field of drug discovery, drug candidate substances are administered to test subjects such as rats to evaluate the efficacy and toxicity of the candidate substances. These evaluation tests use images of tissue sections (brain, liver, heart, etc.) collected by autopsy. Traditionally, pathologists performed the evaluation by observing the section images. However, recent advances in image analysis technology have led to the development of automated evaluation techniques that can perform the evaluation without the need for pathologists.

[0003] For example, JP 2021-508373 A describes an evaluation using a machine learning model trained using slice images of tissue slices free of abnormalities such as tumors and inflammation (referred to in JP 2021-508373 A as "tissue samples substantially free of abnormalities"). Specifically, in JP 2021-508373 A, a slice image of a subject administered with a candidate substance is first input into the machine learning model. Next, the difference between the extracted feature values ​​and the feature values ​​extracted from a training slice image of a tissue slice free of abnormalities (referred to in JP 2021-508373 A as "deviation from normal score") is calculated. Then, based on the calculated difference, it is determined whether or not an abnormality has occurred in the tissue slice of the subject administered with the candidate substance.

[0004] Subjects are divided into a treatment group administered with a candidate substance and a control group not administered with the candidate substance for comparison with the treatment group. Some subjects in the control group develop abnormalities despite not being administered with the candidate substance. The occurrence of abnormalities among subjects in the control group is thought to be due to attributes such as age and / or rearing environment such as diet. There are quite a few subjects in the treatment group who develop abnormalities due to attributes and / or rearing environment. Therefore, when abnormalities occur in tissue sections of subjects in the treatment group, it is necessary to determine whether the candidate substance is truly the cause of the abnormality. However, JP 2021-508373 A does not describe abnormalities caused by attributes and / or rearing environment, and naturally, it does not describe a method for determining whether the candidate substance is the cause of the abnormality.

[0005] One embodiment of the technology disclosed herein provides a drug discovery support device, an operating method for the drug discovery support device, and an operating program for the drug discovery support device that are capable of determining whether abnormalities caused by a drug candidate substance have occurred in subjects in an administration group.

[0006] The drug discovery support device of the present disclosure includes a processor, which acquires a first slice image of a tissue slice of at least one of a plurality of subjects constituting an administration group to which a drug candidate substance has been administered, extracts a first feature from the first slice image, acquires a plurality of second slice images of tissue slices of a plurality of subjects constituting a control group to which the drug candidate substance has not been administered, the plurality of subjects having the same attributes as the subjects constituting the administration group and being kept in the same rearing environment, extracts a second feature from each of the second slice images, acquires third feature information regarding the third feature extracted from each of a plurality of third slice images of tissue slices of a plurality of subjects constituting a control group to which the drug candidate substance has not been administered in a past evaluation test, and determines whether or not an abnormality exists in the tissue slice depicted in the first slice image based on the first feature, the second feature, and the third feature information.

[0007] It is preferable that the processor extracts the first feature amount and the second feature amount using a machine learning model trained using a plurality of third slice images.

[0008] It is preferable that the first feature, the second feature, and the third feature have the same number of dimensions and can be compared in the same feature space, the third feature information is information on representative positions of the multiple third feature in the feature space, and the processor calculates a first distance between the representative position in the feature space and the position of the first feature, calculates second distances between the representative position in the feature space and the positions of each of the multiple second feature, and makes a judgment based on the first distance and the second distance.

[0009] It is preferable that the processor extracts a first feature amount for each of a plurality of first patch images obtained by subdividing the first slice image and calculates a first distance, extracts a second feature amount for each of a plurality of second patch images obtained by subdividing the second slice image and calculates a second distance, and makes a determination based on a first area ratio of the first slice image of the first patch image for which the first distance is outside a set range and a second area ratio of the second slice image of the second patch image for which the second distance is outside a set range.

[0010] Preferably, the processor determines that an abnormality exists in the tissue slice shown in the first slice image when there is a statistically significant difference between the first area ratio and the second area ratio.

[0011] Preferably, the processor determines that an abnormality exists in a tissue slice captured in the first slice image in which the first area ratio exceeds a threshold value set based on the second area ratio.

[0012] It is preferable that the first feature, the second feature, and the third feature have the same number of dimensions and can be compared in the same feature space, the third feature information is information on representative positions of the plurality of third feature values ​​in the feature space, and the processor calculates a first distance between the representative positions in the feature space and the position of the first feature value, calculates a third distance between the representative positions of the plurality of second feature values ​​in the feature space and the position of the first feature value, and makes a judgment based on the first distance and the third distance.

[0013] Preferably, the processor determines that an abnormality exists in the tissue slice captured in the first slice image when the weighted sum of the first distance and the third distance is outside a set range.

[0014] The processor preferably receives an instruction to set a weighting factor by which the first distance and the third distance are multiplied in the weighted sum.

[0015] The first and second slice images are images obtained by photographing a slide specimen on which tissue slices of multiple types of organs are placed, and it is preferable that the processor identify the tissue slices of each organ from the first and second slice images, extract first and second feature amounts for each of the identified tissue slices of each organ, and determine whether or not an abnormality exists in each of the tissue slices of each organ shown in the first slice image based on the first, second, and third feature amount information.

[0016] A method for operating the drug discovery support device disclosed herein includes acquiring a first slice image of a tissue slice of at least one of a plurality of subjects constituting an administration group to which a drug candidate substance has been administered, extracting a first feature from the first slice image, acquiring a plurality of second slice images of tissue slices of a plurality of subjects constituting a control group to which the drug candidate substance has not been administered, the plurality of subjects having the same attributes as the subjects constituting the administration group and being kept in the same rearing environment, extracting the second feature from each of the plurality of second slice images, acquiring third feature information regarding the third feature extracted from each of a plurality of third slice images of tissue slices of a plurality of subjects constituting a control group to which the drug candidate substance has not been administered in a past evaluation test, and determining whether or not an abnormality exists in the tissue slice depicted in the first slice image based on the first feature, the second feature, and the third feature information.

[0017] The operating program for the drug discovery support device disclosed herein causes a computer to execute processes including: acquiring a first slice image of a tissue slice of at least one of a plurality of subjects constituting an administration group to which a drug candidate substance was administered; extracting a first feature from the first slice image; acquiring a plurality of second slice images of tissue slices of a plurality of subjects constituting a control group to which the drug candidate substance was not administered, the plurality of subjects having the same attributes as the subjects constituting the administration group and being kept in the same rearing environment; extracting the second feature from each of the plurality of second slice images; acquiring third feature information regarding the third feature extracted from each of a plurality of third slice images of tissue slices of a plurality of subjects constituting a control group to which the drug candidate substance was not administered in a previous evaluation test; and determining whether or not an abnormality exists in the tissue slice depicted in the first slice image based on the first feature, the second feature, and the third feature information.

[0018] According to the technology disclosed herein, it is possible to provide a drug discovery support device, an operating method for the drug discovery support device, and an operating program for the drug discovery support device that are capable of determining whether abnormalities caused by candidate drug substances have occurred in subjects in an administration group.

[0019] 1 is a diagram illustrating the steps of an evaluation test, slice images, and a drug discovery support apparatus. It is a diagram illustrating a current administration group and a current control group. It is a block diagram illustrating a computer constituting the drug discovery support apparatus. It is a block diagram illustrating a processing unit of a CPU of the drug discovery support apparatus. It is a diagram illustrating a first patch image obtained by dividing a first slice image. It is a diagram illustrating how a first feature is extracted from a first patch image by a feature extractor. It is a diagram illustrating how a second feature is extracted from a second patch image by a feature extractor. It is a diagram illustrating the structure of a feature extractor. It is a diagram illustrating processing in the learning phase of an autoencoder. It is a diagram illustrating a past control group. It is a diagram illustrating how a third feature is extracted from a third patch image by a feature extractor. It is a diagram illustrating a graph in which a third feature is plotted in a feature space, and third feature information. It is a diagram illustrating a first distance and a second distance. It is a diagram illustrating how a determination unit outputs a comparison result between the first distance and a set range. It is a diagram illustrating how a determination unit outputs a comparison result between the second distance and a set range. It is a diagram illustrating a first area ratio. It is a diagram illustrating how multiple second area ratios are calculated from multiple second slice images. 18A shows a case where the p-value in a statistical hypothesis test using the first area ratio and the second area ratio is equal to or greater than the significance level, and FIG. 18B shows a case where the p-value is less than the significance level. FIG. 18B shows a case where the p-value is less than the significance level. FIG. 18C shows a drug discovery support screen before an abnormality determination instruction is given. FIG. 18D shows a drug discovery support screen after an abnormality determination instruction is given. FIG. 18E shows a flowchart showing the processing procedure of the drug discovery support device. FIG. 18F shows another example of the processing of the determination unit. FIG. 18G shows the processing of the determination unit of the 2_1 embodiment. FIG. 18H shows a 2_2 embodiment in which a coefficient setting instruction is accepted. FIG. 18I shows a third embodiment in which slice images obtained by photographing a slide specimen on which tissue slices of multiple types of organs are placed are handled. FIG. 18I shows how a feature extractor for heart slices extracts a first feature from a first patch image obtained by dividing a heart slice. FIG. 18I shows another example of the drug discovery support screen.

[0020] First Embodiment As shown in Fig. 1 as an example, a drug discovery support device 10 of the present disclosure is used to evaluate the efficacy and toxicity of a drug candidate substance 26 (see Fig. 2). The drug discovery support device 10 is, for example, a desktop personal computer, and includes a display 11 that displays various screens, and an input device 12 that includes a keyboard, a mouse, a touch panel, and / or a microphone for voice input. The drug discovery support device 10 is operated by an operator, such as a staff member at a pharmaceutical facility involved in drug development.

[0021] A slice image 15 is input to the drug discovery support device 10. The slice image 15 is generated, for example, by the following procedure. First, a subject S, such as a rat, prepared for evaluation of a candidate substance 26 is autopsied, and multiple tissue slices (hereinafter referred to as liver slices) LVS of the subject S's organs, in this case, the liver LV, are collected. Next, the collected liver slices LVS are attached one by one to glass slides 16, and then stained, in this case, with hematoxylin-eosin dye. Next, the stained liver slices LVS are covered with a cover glass 17 to complete a slide specimen 18. The slide specimen 18 is then placed in an imaging device 19, such as a digital optical microscope, and the slice image 15 is captured by the imaging device 19. The slice image 15 thus obtained is assigned a subject ID (Identification Data) for uniquely identifying the subject S, an image ID for uniquely identifying the slice image 15, and the date and time of imaging, etc. Note that the staining may be performed using hematoxylin dye alone, nuclear fast red dye, or the like.

[0022] As an example, as shown in FIG. 2 , the slice images 15 input to the drug discovery support apparatus 10 include a first slice image 151 and a second slice image 152. The first slice image 151 is an image of a liver slice LVS of a subject S in a current administration group 25. The current administration group 25 is composed of multiple subjects S to whom a drug candidate substance 26 has been administered. The number of subjects S constituting the current administration group 25 is, for example, approximately 5 to 10. Since multiple first slice images 151 are obtained from one subject S, the number of first slice images 151 obtained from the current administration group 25 is calculated by multiplying the number obtained from one subject S by the number of subjects S. For example, if the number obtained from one subject S is 10 and the number of subjects S constituting the current administration group 25 is 10, 10 × 10 = 100 first slice images 151 are obtained.

[0023] The second slice image 152 is an image of a liver slice LVS of a subject S in the current control group 27. Unlike the current administration group 25, the current control group 27 is composed of multiple subjects S who were not administered the drug candidate substance 26. The subjects S constituting the current control group 27 have the same attributes as the subjects S constituting the current administration group 25 and are placed in the same rearing environment. The same attributes include, for example, the same age in weeks and / or the same gender. The same attributes also include the same age composition ratio and / or the same gender composition ratio (e.g., five males and five females). The same rearing environment includes, for example, the same food, the same temperature and humidity of the rearing space, and / or the same size of the rearing space. The term "same" in the term "same rearing environment" refers not only to "completely the same" but also to "same" in the sense of including a margin of error generally acceptable in the technical field to which the technology of the present disclosure pertains and that does not contradict the spirit of the technology of the present disclosure. The number of subjects S constituting the current control group 27 is also, for example, about 5 to 10, similar to the current administration group 25. Similar to the first slice image 151, multiple second slice images 152 are obtained from one subject S, and therefore the number of second slice images 152 obtained from the current control group 27 is equal to the number obtained from one subject S multiplied by the number of subjects S.

[0024] The current administration group 25 includes a plurality of groups with different doses of the candidate substance 26. For example, the dose of the candidate substance 26 is differentiated into three levels, such as a high dose group, a medium dose group, and a low dose group. This allows the influence of the dose of the candidate substance 26 on the subject S to be ascertained.

[0025] 3, the computer constituting the drug discovery support apparatus 10 includes, in addition to the display 11 and input device 12, a storage 30, a memory 31, a CPU (Central Processing Unit) 32, and a communication unit 33. These are interconnected via a bus line 34.

[0026] The storage 30 is a hard disk drive built into the computer constituting the drug discovery support apparatus 10 or connected via a cable or network. Alternatively, the storage 30 is a disk array with multiple hard disk drives connected in series. The storage 30 stores control programs such as an operating system, various application programs, and various data associated with these programs. Note that a solid state drive may be used instead of a hard disk drive.

[0027] The memory 31 is a work memory for the CPU 32 to execute processing. The CPU 32 loads programs stored in the storage 30 into the memory 31 and executes processing in accordance with the programs. In this way, the CPU 32 comprehensively controls each part of the computer. The CPU 32 is an example of a "processor" according to the technology of the present disclosure. Note that the memory 31 may be built into the CPU 32. The communication unit 33 controls the transmission of various information to and from external devices such as the imaging device 19.

[0028] As an example, as shown in FIG. 4 , an operating program 40 is stored in the storage 30 of the drug discovery support apparatus 10. The operating program 40 is an application program for causing a computer to function as the drug discovery support apparatus 10. In other words, the operating program 40 is an example of an "operating program for a drug discovery support apparatus" according to the technology of the present disclosure. The storage 30 also stores a feature extractor 41, third feature information 42, setting range information 43, and the like. The feature extractor 41 is an example of a "machine learning model" according to the technology of the present disclosure.

[0029] When the operating program 40 is started, the CPU 32 of the computer constituting the drug discovery support device 10 cooperates with the memory 31 and the like to function as a read / write (hereinafter abbreviated as RW) control unit 50, a first feature extraction unit 511, a second feature extraction unit 512, a determination unit 52, and a display control unit 53.

[0030] The RW control unit 50 controls the storage of various data in the storage 30 and the reading of various data from the storage 30. For example, the RW control unit 50 stores a first slice image 151 and a second slice image 152 from the imaging device 19 in the storage 30.

[0031] The RW control unit 50 acquires the first slice image 151 by reading it from the storage 30. The RW control unit 50 outputs the first slice image 151 to the first feature amount extraction unit 511 and the display control unit 53. The first slice image 151 output from the RW control unit 50 to the first feature amount extraction unit 511 is a target for determining whether or not an abnormality exists in the liver slice LVS. Hereinafter, this first slice image 151 for determining whether or not an abnormality exists in the liver slice LVS will be referred to as a target first slice image 151T (see FIG. 5 , etc.).

[0032] The RW control unit 50 acquires a plurality of second slice images 152 by reading them from the storage 30 , and outputs all of the second slice images 152 to the second feature amount extraction unit 512 .

[0033] The RW control unit 50 reads the feature extractor 41 from the storage 30 and outputs the feature extractor 41 to the first feature extractor 511 and the second feature extractor 512. The RW control unit 50 also obtains third feature information 42 by reading it from the storage 30 and outputs the third feature information 42 to the determination unit 52. The RW control unit 50 also reads setting range information 43 from the storage 30 and outputs the setting range information 43 to the determination unit 52.

[0034] The first feature amount extraction unit 511 extracts a first feature amount 601 from the target first slice image 151T using the feature amount extractor 41. The first feature amount extraction unit 511 outputs the first feature amount 601 to the determination unit 52.

[0035] The second feature amount extraction unit 512 extracts second feature amounts 602 from all of the second slice images 152 using the feature amount extractor 41. The second feature amount extraction unit 512 outputs the second feature amounts 602 to the determination unit 52.

[0036] The determination unit 52 determines whether or not an abnormality caused by the candidate substance 26 is present in the liver slice LVS depicted in the target first slice image 151T, based on the first feature amount 601, the second feature amount 602, the third feature amount information 42, and the set range information 43. An abnormality is a lesion that is not seen in a normal liver slice LVS, such as a tumor, inflammation, a cyst, or infiltration. The determination unit 52 outputs a determination result 61 indicating whether or not an abnormality caused by the candidate substance 26 is present in the liver slice LVS depicted in the target first slice image 151T to the display control unit 53.

[0037] The display control unit 53 controls the display of various screens on the display 11. The various screens include a drug discovery support screen 100 (see FIGS. 19 and 20) that displays a first slice image 151 and an assessment result 61. In addition to these processing units 50 to 53, the CPU 32 also includes an instruction receiving unit 110 (see FIG. 24) that receives various operation instructions from the input device 12.

[0038] As an example, as shown in FIG. 5 , the first feature extraction unit 511 recognizes the liver slice LVS depicted in the target first slice image 151T using a well-known image recognition technique and subdivides the recognized liver slice LVS into a plurality of first patch images 651. The first patch images 651 have a predetermined size that can be handled by the feature extractor 41. Although not shown in the figure, the second feature extraction unit 512 similarly subdivides the liver slice LVS depicted in the second slice image 152 into a plurality of second patch images 652 (see FIG. 7 ). Note that, although the first patch image 651 does not have an area overlapping with other first patch images 651 in FIG. 5 , the first patch image 651 may partially overlap with other first patch images 651. The same applies to the second patch images 652.

[0039] 6 as an example, the first feature amount extraction unit 511 extracts a first feature amount 601 for each of a plurality of first patch images 651 obtained by subdividing the target first slice image 151T using the feature amount extractor 41. Also, as an example, as shown in FIG. 7 as an example, the second feature amount extraction unit 512 extracts a second feature amount 602 for each of a plurality of second patch images 652 obtained by subdividing all of the second slice images 152 using the feature amount extractor 41.

[0040] As an example, as shown in FIG. 8 , the feature extractor 41 is configured by reusing an encoder unit 71 of an autoencoder 70. The autoencoder 70 includes a decoder unit 72 in addition to the encoder unit 71. Patch images 65, such as a first patch image 651 and a second patch image 652, are input to the encoder unit 71. The encoder unit 71 converts the patch images 65 into feature amounts 60. The encoder unit 71 passes the feature amounts 60 to the decoder unit 72. The decoder unit 72 generates a restored image 73 of the patch images 65 from the feature amounts 60.

[0041] As is well known, the encoder unit 71 includes a convolution layer that performs convolution processing using a filter, a pooling layer that performs pooling processing such as maximum value pooling, and the like. The decoder unit 72 also includes a convolution layer that performs convolution processing using a filter, a pooling layer that performs pooling processing such as maximum value pooling, and the like. The encoder unit 71 extracts feature quantities 60 by repeating convolution processing using the convolution layer and pooling processing using the pooling layer multiple times on the input patch image 65. The extracted feature quantities 60 represent the shape and texture characteristics of the liver slice LVS depicted in the patch image 65.

[0042] The feature 60 is a set of multiple numerical values. In other words, the feature 60 is multidimensional data. The number of dimensions of the feature 60 is, for example, 512, 1024, or 2048. The first feature 601, the second feature 602, and the third feature 603 (described later) (see FIG. 11) have the same number of dimensions and can be compared in the same feature space 80 (see FIG. 12, etc.).

[0043] As an example, as shown in FIG. 9 , in a learning phase before the encoder unit 71 is converted into the feature extractor 41, the autoencoder 70 receives a third training patch image 653L as input and is trained. The autoencoder 70 outputs a reconstructed training image 73L for the third training patch image 653L. A loss calculation is performed on the autoencoder 70 using a loss function based on the third training patch image 653L and the reconstructed training image 73L. Then, update settings are made for various coefficients of the autoencoder 70 (such as the filter coefficients of the convolution layer) according to the results of the loss calculation, and the autoencoder 70 is updated according to the update settings.

[0044] During the learning phase of the autoencoder 70, the above-described series of processes, including input of the third training patch image 653L to the autoencoder 70, output of the training restored image 73L from the autoencoder 70, loss calculation, update setting, and update of the autoencoder 70, are repeatedly performed while the third training patch image 653L is being replaced. The repetition of the above-described series of processes is terminated when the accuracy of restoration from the third training patch image 653L to the training restored image 73L reaches a predetermined set level. The encoder unit 71 of the autoencoder 70 whose restoration accuracy has reached the set level is stored in the storage 30 of the drug discovery support apparatus 10 as the feature extractor 41. Note that learning may be terminated when the above-described series of processes has been repeated a set number of times, regardless of the accuracy of restoration from the third training patch image 653L to the training restored image 73L.

[0045] The learning of the autoencoder 70 may be performed by the drug discovery support device 10, or may be performed by a device separate from the drug discovery support device 10. In the latter case, the feature extractor 41 is transmitted from the separate device to the drug discovery support device 10, and the feature extractor 41 is stored in the storage 30 by the RW control unit 50.

[0046] As an example, as shown in FIG. 10 , the third learning patch image 653L is supplied from a plurality of third patch images 653 obtained by subdividing the liver slice LVS depicted in the third slice image 153. The third slice image 153 is an image of the liver slice LVS of a subject S in the past control group 75. The past control group 75 is composed of a plurality of subjects S who were not administered a drug candidate substance in past evaluation tests. For this reason, the number of subjects S comprising the past control group 75 is significantly greater than the number of subjects S comprising the current administration group 25 and the current control group 27, for example, on the order of several hundred to several thousand. Like the first slice image 151, etc., multiple third slice images 153 are obtained from one subject S. Therefore, the number of third slice images 153 obtained from the past control group 75 is calculated by multiplying the number obtained from one subject S by the number of subjects S.

[0047] Next, the structure of the third feature amount information 42 will be described. First, as an example, as shown in Fig. 11 , the feature amount extractor 41 extracts a plurality of third feature amounts 603 from each of a plurality of third patch images 653 based on a plurality of third section images 153 obtained from all of the subjects S constituting the past control group 75.

[0048] FIG. 12 is an example of a graph in which multiple third feature quantities 603 extracted in FIG. 11 are plotted in a feature quantity space 80. The third feature quantity information 42 includes coordinates 81 in the feature quantity space 80 of representative positions of the third feature quantities 603, indicated by crosses (hereinafter referred to as representative position coordinates). The representative position is, for example, the center point or average point of a distribution 82 of the third feature quantities 603. Note that in FIG. 12, for convenience of explanation, the dimension of the feature quantity space 80 is shown as two-dimensional having a D1 axis and a D2 axis, but the actual dimension of the feature quantity space 80 is, as described above, 512 dimensions or the like. Similarly, in the following FIG. 13 and the like, for convenience of explanation, the dimension of the feature quantity space 80 is shown as two-dimensional.

[0049] As with the learning of the autoencoder 70, the third feature amount information 42 may be created by the drug discovery support apparatus 10, or may be created by a device other than the drug discovery support apparatus 10. In the latter case, the third feature amount information 42 is transmitted from the other device to the drug discovery support apparatus 10, and the RW control unit 50 stores the third feature amount information 42 in the storage 30. The third feature amount information 42 may also be a plurality of third feature amounts 603 themselves. In this case, the representative position coordinates 81 are derived by the drug discovery support apparatus 10.

[0050] 13 , the dashed-dotted line enclosure and the reference numeral 85 indicate the distribution of the first feature amount 601 extracted for each of the plurality of first patch images 651 obtained by subdividing one target first section image 151T. The dashed-dotted line enclosure and the reference numeral 86 indicate the distribution of the second feature amount 602 extracted for each of the plurality of second patch images 652 obtained by subdividing all of the second section images 152.

[0051] The determination unit 52 calculates a first distance D1 in the feature space 80 between the representative position of the third feature 603 represented by the representative position coordinates 81 of the third feature information 42 and the position of the first feature 601. The determination unit 52 calculates the first distance D1 between the multiple first feature 601 extracted for each of the multiple first patch images 651 obtained by subdividing one target first slice image 151T. The determination unit 52 also calculates a second distance D2 in the feature space 80 between the representative position of the third feature 603 and the position of the second feature 602. The determination unit 52 calculates the second distance D2 between the multiple second feature 602 extracted for each of the multiple second patch images 652 obtained by subdividing all of the second slice images 152. The first distance D1 and the second distance D2 are Mahalanobis distances.

[0052] 14 , the determination unit 52 compares the calculated first distance D1 with the setting range of the setting range information 43. The setting range information 43 includes a threshold value indicating the setting range. If the first distance D1 is less than the threshold value of the setting range information 43, the determination unit 52 outputs a comparison result 90 indicating that the first distance D1 is within the setting range. On the other hand, if the first distance D1 is equal to or greater than the threshold value of the setting range information 43, the determination unit 52 outputs a comparison result 90 indicating that the first distance D1 is outside the setting range.

[0053] 15 as an example, the determination unit 52 compares the calculated second distance D2 with the setting range of the setting range information 43. If the second distance D2 is less than the threshold of the setting range information 43, the determination unit 52 outputs a comparison result 90 indicating that the second distance D2 is within the setting range. On the other hand, if the second distance D2 is equal to or greater than the threshold of the setting range information 43, the determination unit 52 outputs a comparison result 90 indicating that the second distance D2 is outside the setting range.

[0054] 16 as an example, the determination unit 52 calculates a first area ratio 951 of the first patch images 651 (indicated by hatching) whose first distance D1 is outside the set range to the target first section image 151T. Specifically, the determination unit 52 calculates the first area ratio 951 by dividing the number of first patch images 651 whose first distance D1 is outside the set range by the total number of first patch images 651. FIG. 16 illustrates an example in which the first area ratio 951 is 11.9%.

[0055] The determination unit 52 calculates a second area ratio 952 of the second patch image 652 in which the second distance D2 is outside the set range, in the second segment image 152, in the same manner as in the case of the first area ratio 951. However, as described above, the second distance D2 is calculated based on all the second segment images 152, and therefore the determination unit 52 calculates the second area ratio 952 for each of all the second segment images 152, as shown in FIG.

[0056] 18 , the determination unit 52 determines whether or not an abnormality caused by the candidate substance 26 is present in the liver slice LVS captured in the target first slice image 151T, based on a p-value in a statistical hypothesis test based on the null hypothesis that there is no difference between the first area ratio 951 and the second area ratio 952. As shown in FIG. 18A , if the p-value is equal to or greater than the significance level (0.05 is used as an example here), the determination unit 52 determines that there is no statistically significant difference and outputs a determination result 61 indicating that there is no abnormality caused by the candidate substance 26 in the liver slice LVS captured in the target first slice image 151T. On the other hand, as shown in FIG. 18B , if the p-value is less than the significance level, the determination unit 52 determines that there is a statistically significant difference and outputs a determination result 61 indicating that there is an abnormality caused by the candidate substance 26 in the liver slice LVS captured in the target first slice image 151T.

[0057] As an example, as shown in FIG. 19 , the display control unit 53 controls the display of a drug discovery support screen 100 on the display 11. The display control unit 53 displays the drug discovery support screen 100 when an operator issues an instruction to display a first slice image 151 via the input device 12 and the instruction is accepted by the instruction accepting unit 110. A plurality of first slice images 151 are displayed side by side on the drug discovery support screen 100. These first slice images 151 are first slice images 151 obtained from one of the plurality of subjects S constituting the current administration group 25. FIG. 19 shows an example in which ten first slice images 151 with image IDs "SP001" to "SP010" obtained from the subject S with subject ID "R001" are displayed side by side. The subject ID is displayed as a pull-down menu 101, making it possible to switch the subject S for which the first slice image 151 is displayed on the drug discovery support screen 100. The pull-down menu 101 also has an option for specifying all of the multiple subjects S that make up the current administration group 25 .

[0058] An abnormality determination button 102 is provided at the bottom of the drug discovery support screen 100. When the abnormality determination button 102 is selected by the operator, an abnormality determination instruction is received by the instruction receiving unit 110. The abnormality determination instruction is an instruction to cause the first feature amount extraction unit 511, the second feature amount extraction unit 512, and the determination unit 52 to extract the first feature amount 601 and the second feature amount 602, and to determine whether or not an abnormality exists in the liver slice LVS shown in the first slice image 151 displayed on the drug discovery support screen 100. In other words, the first slice image 151 displayed on the drug discovery support screen 100 corresponds to the target first slice image 151T.

[0059] When the abnormality determination button 102 is selected and the determination unit 52 determines whether or not an abnormality exists in the liver slice LVS captured in the target first slice image 151T, the display control unit 53 displays a determination result 61 below each first slice image 151, as shown in Fig. 20 as an example. Fig. 20 illustrates a case in which it has been determined that an abnormality exists in the liver slice LVS captured in three first slice images 151 with image IDs "SP007" to "SP009." Note that a frame indicating a first patch image 651 whose comparison result 90 is outside the set range may be superimposed on the first slice image 151.

[0060] Next, the operation of the above configuration will be described with reference to the flowchart shown in Fig. 21. First, when the operating program 40 is started in the drug discovery support apparatus 10, the CPU 32 of the drug discovery support apparatus 10 functions as the RW control unit 50, the first feature amount extraction unit 511, the second feature amount extraction unit 512, the determination unit 52, and the display control unit 53, as shown in Fig. 4.

[0061] First, the imaging device 19 captures a first slice image 151 of the liver slice LVS of the subject S in the current administration group 25, and a second slice image 152 of the liver slice LVS of the subject S in the current control group 27. The first slice image 151 and the second slice image 152 are output from the imaging device 19 to the drug discovery support device 10. In the drug discovery support device 10, the first slice image 151 and the second slice image 152 from the imaging device 19 are stored in the storage 30 by the RW control unit 50 (step ST100).

[0062] When the operator issues an instruction to display the first section image 151 via the input device 12 (YES in step ST110), the RW control unit 50 reads out the first section image 151 specified in the display instruction from the storage 30 (step ST120). The first section image 151 is output from the RW control unit 50 to the display control unit 53. As shown in Fig. 19 , the first section image 151 is displayed on the display 11 through the drug discovery support screen 100 under the control of the display control unit 53 (step ST130).

[0063] When the operator selects the abnormality determination button 102 on the drug discovery support screen 100 and issues an abnormality determination instruction (YES in step ST140), the RW control unit 50 reads out the first slice image 151 currently displayed on the drug discovery support screen 100 from the storage 30 as a target first slice image 151T, thereby acquiring the image (step ST150). The RW control unit 50 also reads out all second slice images 152 from the storage 30, thereby acquiring the image (step ST150). The target first slice image 151T is output from the RW control unit 50 to the first feature extraction unit 511. The second slice images 152 are also output from the RW control unit 50 to the second feature extraction unit 512.

[0064] The RW control unit 50 reads the feature extractor 41 from the storage 30 (step ST150) and outputs it to the first feature extraction unit 511 and the second feature extraction unit 512. The RW control unit 50 also reads and acquires third feature information 42 from the storage 30 (step ST150) and outputs it to the determination unit 52. The RW control unit 50 also reads setting range information 43 from the storage 30 (step ST150) and outputs it to the determination unit 52.

[0065] 5 , the first feature extraction unit 511 subdivides the first slice image 151 into a plurality of first patch images 651 (step ST160). Similarly, the second feature extraction unit 512 subdivides the second slice image 152 into a plurality of second patch images 652 (step ST160).

[0066] 6 , in the first feature amount extraction unit 511, the feature amount extractor 41 extracts a first feature amount 601 from the first patch image 651 (step ST170). The first feature amount 601 is output from the first feature amount extraction unit 511 to the determination unit 52. Also, as shown in FIG. 7 , in the second feature amount extraction unit 512, the feature amount extractor 41 extracts a second feature amount 602 from the second patch image 652 (step ST170). The second feature amount 602 is output from the second feature amount extraction unit 512 to the determination unit 52.

[0067] 13 , the determination unit 52 calculates a first distance D1 between the representative position of the third feature amount 603 and the position of the first feature amount 601 (step ST180), and also calculates a second distance D2 between the representative position of the third feature amount 603 and the position of the second feature amount 602 (step ST180).

[0068] As shown in Fig. 14 , the determination unit 52 compares the first distance D1 with the set range in the set range information 43, and outputs a comparison result 90. Also, as shown in Fig. 15 , the second distance D2 is compared with the set range in the set range information 43, and outputs a comparison result 90. Then, as shown in Fig. 16 , a first area ratio 951 of the first patch image 651, in which the first distance D1 is outside the set range, to the first segment image 151 is calculated (step ST190). Also, as shown in Fig. 17 , a second area ratio 952 of the second patch image 652, in which the second distance D2 is outside the set range, to the second segment image 152 is calculated (step ST190).

[0069] 18 , the determination unit 52 determines whether or not an abnormality caused by the candidate substance 26 is present in the liver slice LVS depicted in the target first slice image 151T based on the p-value in the statistical hypothesis test using the first area ratio 951 and the second area ratio 952 (step ST200). The determination result 61 is output from the determination unit 52 to the display control unit 53. As shown in FIG. 20 , the determination result 61 is displayed on the display 11 via the drug discovery support screen 100 under the control of the display control unit 53 (step ST210). The operator views the determination result 61 and uses the determination result 61 as a foothold to evaluate the candidate substance 26, for example, by closely observing the first slice image 151 determined to be abnormal.

[0070] As described above, the CPU 32 of the drug discovery support apparatus 10 includes the RW control unit 50, the first feature extraction unit 511, the second feature extraction unit 512, and the determination unit 52. The RW control unit 50 acquires the first slice image 151, the second slice image 152, and the third feature information 42 by reading them from the storage 30. The first slice image 151 is an image of a liver slice LVS of a subject S constituting the current administration group 25, who has been administered the drug candidate substance 26. The second slice image 152 is a plurality of images of liver slice LVS of a plurality of subjects S constituting the current control group 27, who have the same attributes as the subjects S constituting the current administration group 25 and are placed in the same rearing environment, but who have not been administered the drug candidate substance 26. The third feature information 42 is information regarding the third feature 603 extracted from the third slice image 153. The third slice images 153 are multiple images of liver slices LVS of multiple subjects S constituting a past control group 75 to which no candidate substance was administered in a past evaluation test.

[0071] The first feature extraction unit 511 extracts a first feature 601 from the target first slice image 151T. The second feature extraction unit 512 extracts a second feature 602 from each of the second slice images 152. The determination unit 52 determines whether or not an abnormality exists in the liver slice LVS depicted in the target first slice image 151T based on the first feature 601, the second feature 602, and the third feature information 42.

[0072] The current control group 27 includes subjects S whose abnormalities have occurred due to their attributes and / or rearing environment. The same applies to the past control group 75. The technology disclosed herein compares the second feature 602 and third feature information 42 based on the current control group 27 and the past control group 75 with the first feature 601 based on the current administration group 25 to determine whether or not an abnormality exists in the liver slice LVS of the subjects S constituting the current administration group 25. Therefore, if the determination result 61 output by such a determination indicates an abnormality, it is determined that the cause of the abnormality is not the attributes and / or rearing environment, in other words, that the cause of the abnormality is the candidate substance 26. Therefore, the technology disclosed herein makes it possible to determine whether or not an abnormality caused by the drug candidate substance 26 has occurred in the subjects S in the current administration group 25.

[0073] The subjects S constituting the current control group 27 have the same attributes and have been placed in the same rearing environment as the subjects S constituting the current administration group 25, but their small number makes it unreliable to make a judgment based on them alone. Therefore, in the technology disclosed herein, the small number of subjects S constituting the current control group 27 is compensated for by the past control group 75. On the other hand, the subjects S constituting the past control group 75, while numerous, may have different attributes and rearing environments from the subjects S constituting the current administration group 25. For this reason, making a judgment based on them alone may result in a high number of erroneous judgments. Therefore, in the technology disclosed herein, the current control group 27 compensates for the differences in attributes and rearing environments of the subjects S constituting the past control group 75. In other words, the current control group 27 and the past control group 75 are in a complementary relationship. In the technology disclosed herein, by appropriately combining the current control group 27 and the past control group 75, which are in such a complementary relationship, it is possible to determine whether the candidate substance 26 is the cause of the abnormality occurring in the subject S.

[0074] The first feature extraction unit 511 and the second feature extraction unit 512 extract the first feature 601 and the second feature 602 by the feature extractor 41, which is a repurposed version of the encoder unit 71 of the autoencoder 70 that has been trained using a plurality of third slice images 153. This makes it possible to easily extract the first feature 601 and the second feature 602 that effectively represent the shape and texture features of the liver slice LVS.

[0075] The first feature amount 601, the second feature amount 602, and the third feature amount 603 have the same number of dimensions and can be compared in the same feature amount space 80. The third feature amount information 42 is information on representative positions of the plurality of third feature amounts 603 in the feature amount space 80. The determination unit 52 calculates a first distance D1 between the representative position of the third feature amount 603 and the position of the first feature amount 601 in the feature amount space 80. The determination unit 52 also calculates a second distance D2 between the representative position of the third feature amount 603 and the positions of each of the plurality of second feature amounts 602 in the feature amount space 80. The determination unit 52 makes a determination based on the first distance D1 and the second distance D2.

[0076] The first distance D1 indicates the degree of deviation between the target first section image 151T and the third section image 153, and the second distance D2 indicates the degree of deviation between the second section image 152 and the third section image 153. The degree of deviation between the target first section image 151T and the third section image 153 indicates the degree of difference between the abnormality occurring in the current administration group 25 and the abnormality occurring in the past control group 75. In addition, the degree of deviation between the second section image 152 and the third section image 153 indicates the degree of difference between the abnormality occurring in the current control group 27 and the abnormality occurring in the past control group 75. Therefore, by making a judgment based on the first distance D1 and the second distance D2, it is possible to make a judgment that takes into account both the degree of difference between the abnormalities that occurred in the current administration group 25 and the abnormalities that occurred in the past control group 75, and the degree of difference between the abnormalities that occurred in the current control group 27 and the abnormalities that occurred in the past control group 75, thereby improving the accuracy of the judgment.

[0077] The first feature amount extraction unit 511 extracts a first feature amount 601 for each of a plurality of first patch images 651 obtained by subdividing the target first slice image 151T. The second feature amount extraction unit 512 extracts a second feature amount 602 for each of a plurality of second patch images 652 obtained by subdividing the second slice image 152. The determination unit 52 calculates a first distance D1 for each of the plurality of first feature amounts 601 extracted from the plurality of first patch images 651. The determination unit 52 calculates a second distance D2 for each of the plurality of second feature amounts 602 extracted from the plurality of second patch images 652. The determination unit 52 makes a determination based on a first area ratio 951 of the first patch image 651, in which the first distance D1 is outside the set range, to the target first slice image 151T, and a second area ratio 952 of the second patch image 652, in which the second distance D2 is outside the set range, to the second slice image 152. Compared to a case where a determination is made without subdividing the target first slice image 151T and the second slice image 152 into the first patch image 651 and the second patch image 652, the amount of information to be used as a reference for the determination is significantly increased, and therefore the determination accuracy can be further improved.

[0078] The first area ratio 951 increases as the difference between the abnormality occurring in the current administration group 25 and the abnormality occurring in the past control group 75 increases (the more the abnormality occurring in the current administration group 25 is caused by the candidate substance 26). On the other hand, the second area ratio 952 decreases as the difference between the abnormality occurring in the current control group 27 and the abnormality occurring in the past control group 75 decreases (the more similar the abnormality occurring in the current control group 27 is to the abnormality occurring in the past control group 75). Therefore, if a statistically significant difference is considered to exist between the first area ratio 951 and the second area ratio 952, it can be said that there is a high possibility that an abnormality caused by the candidate substance 26 exists in the liver slice LVS depicted in the target first slice image 151T. Therefore, if a statistically significant difference exists between the first area ratio 951 and the second area ratio 952, it can be determined that an abnormality exists in the liver slice LVS depicted in the target first slice image 151T, thereby further improving the accuracy of the determination.

[0079] As an example, the determination criterion may be a threshold value TH set based on the second area ratio 952, as shown in FIG. 22 . In FIG. 22 , the threshold value TH is the maximum value of multiple second area ratios 952 calculated based on multiple second section images 152. In this case, the determination unit 52 determines that an abnormality caused by the candidate substance 26 exists in the liver section LVS depicted in the target first section image 151T whose first area ratio 951 exceeds the threshold value TH. This aspect eliminates the need to derive a p-value or compare the p-value with a significance level, making the determination easier. The threshold value TH may be the average value or mode of multiple second area ratios 952, for example.

[0080] 23 as an example, in the 2_1 embodiment, the determination unit 52 calculates a third distance D3 instead of the second distance D2. The third distance D3 is the distance between a representative position (indicated by a + sign) of a plurality of second feature amounts 602 in the feature amount space 80 and the position of the first feature amount 601. The third distance D3 is also a Mahalanobis distance. The representative position of the second feature amount 602 is the center point, the average point, or the like of the distribution 86 of the second feature amounts 602 shown in FIG. 13.

[0081] The determination unit 52 calculates a weighted sum DS of the first distance D1 and the third distance D3, DS = W1 × D1 + W3 × D3. Here, W1 is a weighting coefficient by which the first distance D1 is multiplied, and W3 is a weighting coefficient by which the third distance D3 is multiplied. 0≦W1≦1, 0≦W3≦1, and W1+W3=1. The determination unit 52 compares the weighted sum DS with the setting range in the setting range information 105. The setting range information 105 is stored in the storage 30 and includes a threshold value indicating the setting range. If the weighted sum DS is less than the threshold value of the setting range, the determination unit 52 outputs a comparison result 106 indicating that the weighted sum DS is within the setting range. On the other hand, if the weighted sum DS is equal to or greater than the threshold value of the setting range, the determination unit 52 outputs a comparison result 106 indicating that the weighted sum DS is outside the setting range.

[0082] 13 , the determination unit 52 calculates the first distance D1 and the third distance D3, and the weighted sum DS, for each of the first feature amounts 601 constituting the distribution 85 shown in Fig. 13 , and outputs the comparison result 106. If the comparison result 106 indicates that even one of the weighted sums DS calculated from the multiple first feature amounts 601 constituting the distribution 85 is outside the set range, the determination unit 52 outputs the determination result 61 indicating that an abnormality caused by the candidate substance 26 is present in the liver slice LVS captured in the target first slice image 151T.

[0083] As described above, in the second embodiment, the determination unit 52 calculates the first distance D1 between the representative position of the third feature amount 603 and the position of the first feature amount 601 in the feature amount space 80, and calculates the third distance D3 between the representative positions of the plurality of second feature amounts 602 and the position of the first feature amount 601 in the feature amount space 80. Then, the determination unit 52 makes a determination based on the first distance D1 and the third distance D3.

[0084] As described above, the first distance D1 indicates the degree of deviation between the target first section image 151T and the third section image 153, and the degree of deviation between the target first section image 151T and the third section image 153 indicates the degree of difference between the abnormality that occurred in the current administration group 25 and the abnormality that occurred in the past control group 75. The third distance D3 indicates the degree of deviation between the target first section image 151T and the second section image 152, and the degree of deviation between the target first section image 151T and the second section image 152 indicates the degree of difference between the abnormality that occurred in the current administration group 25 and the abnormality that occurred in the current control group 27. Therefore, if the determination is made based on the first distance D1 and the third distance D3, it is possible to make a determination that takes into account both the degree of difference between the abnormality that occurred in the current administration group 25 and the abnormality that occurred in the past control group 75, and the degree of difference between the abnormality that occurred in the current administration group 25 and the abnormality that occurred in the current control group 27, thereby improving the determination accuracy. However, in the second_1 embodiment, it is necessary to derive a representative position of the second feature amount 602. Therefore, the first embodiment, which does not require this, requires less processing effort.

[0085] When the weighted sum DS of the first distance D1 and the third distance D3 is outside the set range, the determination unit 52 determines that an abnormality exists in the liver slice LVS captured in the target first slice image 151T. This allows for a determination that takes into account both the degree of difference between the abnormality that occurred in the current administration group 25 and the abnormality that occurred in the past control group 75, and the degree of difference between the abnormality that occurred in the current administration group 25 and the abnormality that occurred in the current control group 27, thereby improving the accuracy of the determination.

[0086] Note that the determination may be made based on a simple sum D1+D3 of the first distance D1 and the third distance D3, instead of the weighted sum DS. Furthermore, if the area ratio of the first patch image 651, for which the comparison result 106 is outside the set range, to the first slice image 151 is equal to or greater than a threshold, it may be determined that an abnormality exists in the liver slice LVS depicted in the first slice image 151.

[0087] [Embodiment 2_2] As shown in FIG. 24 as an example, in embodiment 2_2, the instruction receiving unit 110 receives a coefficient setting instruction 111 from an operator via the input device 12. The coefficient setting instruction 111 is an instruction for setting weighting factors W1 and W3 and includes the weighting factors W1 and W3 set by the operator. When the instruction receiving unit 110 receives the coefficient setting instruction 111, it outputs coefficient setting information 112 to the determination unit 52. The coefficient setting information 112 is information that is a direct copy of the weighting factors W1 and W3 included in the coefficient setting instruction 111. The determination unit 52 calculates the weighted sum DS using the weighting factors W1 and W3 of the coefficient setting information 112. FIG. 24 shows an example in which the weighting factor W1 is set to 0.2 and the weighting factor W3 is set to 0.8.

[0088] When the current evaluation test for the current administration group 25 and the current control group 27 and the past evaluation test for the past control group 75 are similar in content, the operator sets the weighting factor W1 to a value greater than the weighting factor W3. On the other hand, when the current evaluation test for the current administration group 25 and the current control group 27 and the past evaluation test for the past control group 75 are different in content, the operator sets the weighting factor W3 to a value greater than the weighting factor W1.

[0089] Thus, in the second embodiment, the instruction receiving unit 110 receives a coefficient setting instruction 111, which is an instruction to set the weighting coefficients W1 and W3 by which the first distance D1 and the third distance D3 are multiplied in the weighted sum DS. This allows the operator to set the weighting coefficients W1 and W3 appropriate for the similarity between the current evaluation test for the current administration group 25 and the current control group 27 and the past evaluation test for the past control group 75. As a result, the reliability of the determination result 61 can be improved.

[0090] In addition, the similarity between the current evaluation test for the current administration group 25 and the current control group 27 and the past evaluation test for the past control group 75 may be derived by comparing the test conditions of the current evaluation test with the test conditions of the past evaluation test, and the weighting factors W1 and W3 adapted to the derived similarity may be automatically set.

[0091] [Third Embodiment] As an example, as shown in Fig. 25 , the third embodiment deals with a slide specimen 120 in which tissue slices of multiple organs are placed on a single glass slide 16. Fig. 25 illustrates an example in which a heart slice HS, a brain slice BS, and a bone marrow slice BMS are placed in addition to a liver slice LVS. In this case, the slice image 15 (first slice image 151 and second slice image 152) shows the heart slice HS, the brain slice BS, and the bone marrow slice BMS in addition to the liver slice LVS.

[0092] The CPU 32 of the drug discovery support device 10 of the third embodiment functions as an identification unit 121 in addition to the processing units 50 to 53 of the first embodiment. The identification unit 121 identifies the tissue slices of each organ from the slice image 15 (first slice image 151 and second slice image 152) using, for example, a template for identifying the tissue slices of each organ or a machine learning model. The identification unit 121 outputs, as an identification result, coordinate information of boxes 122 to 125 surrounding the tissue slices of each organ. Box 122 is a box surrounding a heart slice HS, and box 123 is a box surrounding a liver slice LVS. Furthermore, box 124 is a box surrounding a brain slice BS, and box 125 is a box surrounding a bone marrow slice BMS.

[0093] 26 shows how the first feature extraction unit 511 subdivides the heart slice HS in the frame 122 into multiple first patch images 651, and how the feature extractor 130 for the heart slice HS extracts first feature amounts 601 from the first patch images 651. The first feature extraction unit 511 also extracts first feature amounts 601 for the liver slice LVS in the frame 123, the brain slice BS in the frame 124, and the bone marrow slice BMS in the frame 125. The second feature extraction unit 512 similarly extracts second feature amounts 602. That is, the first feature extraction unit 511 and the second feature extraction unit 512 extract the first feature amount 601 and the second feature amount 602 for each tissue slice of each organ identified by the identification unit 121. The subsequent processing by the judgment unit 52 is the same as the processing shown in the first embodiment or the second_1 embodiment, except that judgment is made for each tissue slice of each organ appearing in the target first slice image 151T, so illustration and explanation will be omitted.

[0094] As described above, in the third embodiment, the first slice image 151 and the second slice image 152 are images obtained by capturing a slide specimen 120 on which tissue slices of multiple organs are placed. The identification unit 121 identifies the tissue slices of each organ from the first slice image 151 and the second slice image 152. The first feature extraction unit 511 and the second feature extraction unit 512 extract a first feature 601 and a second feature 602 for each of the identified tissue slices of each organ. The determination unit 52 determines whether or not an abnormality exists in each of the tissue slices of each organ depicted in the target first slice image 151T based on the first feature 601, the second feature 602, and the third feature information 42. This allows the system to handle slide specimens 120 on which tissue slices of multiple organs are placed. The slide specimen 120, on which tissue slices from multiple organs are mounted, as in the third embodiment, is more common than the slide specimen 18, on which tissue slices from a single organ are mounted, as in the first embodiment. This allows for drug discovery support that is more suited to general operations.

[0095] The operator may manually define frames showing tissue slices of each organ in the first slice image 151 and the second slice image 152 .

[0096] As an example, as shown in FIG. 27 , a drug discovery support screen 135 may be configured such that a check box 136 is provided for each of the multiple first section images 151 displayed, allowing the operator to select a first section image 151 to be used as a target first section image 151T. In addition to the abnormality determination button 102, a select / cancel all button 137 is provided at the bottom of the drug discovery support screen 135. The select / cancel all button 137 is a button for selecting all of the multiple first section images 151 as target first section images 151T and a button for canceling that selection. This allows the operator to select the target first section image 151T for which judgment is desired. FIG. 27 illustrates an example in which only the first section image 151 with the image ID "SP001" is selected. In other words, only one target first section image 151T may be used for judgment.

[0097] Although the encoder unit 71 of the autoencoder 70 has been exemplified as the feature extractor 41, the feature extractor 41 is not limited to this. The encoder unit of a convolutional neural network trained to output a class of input patch images 65, for example, a determination result as to whether or not an abnormality exists in the liver slice LVS shown in the patch images 65, may be used as the feature extractor 41. Furthermore, other machine learning models such as a support vector machine may be used as the feature extractor 41.

[0098] In the tissue section shown in the third learning patch image 653L, regions that are likely to be mistakenly detected as abnormal may be masked before use for learning. Examples of regions that are likely to be mistakenly detected as abnormal include blood vessel regions in the liver section LVS and dust particles that adhered during the preparation of the slide specimen 18 or 120.

[0099] The feature 60 is not limited to that extracted by the feature extractor 41. It may be the average value, maximum value, minimum value, mode, or variance of the pixel values ​​of the patch image 65.

[0100] The organ is not limited to the liver LV, but may be the stomach, lung, small intestine, or large intestine. The subject S is not limited to a rat, but may be a mouse, guinea pig, gizzard shad, hamster, ferret, rabbit, dog, cat, monkey, or the like.

[0101] The technology described in JP-A-2021-508373 may also be applied. If the technology described in JP-A-2021-508373 determines that an abnormality has occurred in the tissue section of subject S in current administration group 25, but the technology disclosed herein results in a determination result 61 of subject S showing no abnormality, it can be determined that the cause of the abnormality is not candidate substance 26 but the attribute and rearing environment.

[0102] The drug discovery support device 10 may be a personal computer installed in a pharmaceutical facility as shown in FIG. 1, or may be a server computer installed in a data center independent of the pharmaceutical facility.

[0103] When the drug discovery support device 10 is configured as a server computer, the first slice image 151 and the second slice image 152 are transmitted from a personal computer installed in each pharmaceutical facility to the server computer via a network such as the Internet. The server computer distributes the drug discovery support screen 100 or 135 to the personal computer in the form of screen data for web distribution created using a markup language such as XML (Extensible Markup Language). The personal computer reproduces the drug discovery support screen 100 or 135 to be displayed on a web browser based on the screen data and displays it on a display. Note that other data description languages ​​such as JSON (Javascript (registered trademark) Object Notation) may be used instead of XML.

[0104] The drug discovery support device 10 according to the technique of the present disclosure can be widely used across all stages of drug development, from the earliest stage of setting drug discovery targets to the final stage of clinical trials.

[0105] The hardware configuration of the computer constituting the drug discovery support apparatus 10 according to the technology of the present disclosure can be modified in various ways. For example, the drug discovery support apparatus 10 can be configured with multiple computers separated as hardware in order to improve processing power and reliability. For example, the functions of the first feature extraction unit 511 and the second feature extraction unit 512 and the function of the determination unit 52 can be distributed and performed by two computers. In this case, the drug discovery support apparatus 10 is configured with two computers.

[0106] In this way, the hardware configuration of the computer of the drug discovery support apparatus 10 can be changed as appropriate depending on the required performance, such as processing power, safety, and reliability. Furthermore, not only the hardware, but also application programs such as the operating program 40 can be duplicated or stored in multiple storage devices in order to ensure safety and reliability.

[0107] In each of the above embodiments, the hardware structure of the processing units that perform various processes, such as the RW control unit 50, the first feature extraction unit 511, the second feature extraction unit 512, the determination unit 52, the display control unit 53, the instruction reception unit 110, and the identification unit 121, can be any of the various processors listed below. As described above, the various processors include the CPU 32, which is a general-purpose processor that executes software (operating program 40) and functions as various processing units, as well as programmable logic devices (PLDs) that are processors whose circuit configuration can be changed after manufacture, such as FPGAs (Field Programmable Gate Arrays), and dedicated electrical circuits that are processors having a circuit configuration designed specifically for executing specific processing, such as ASICs (Application Specific Integrated Circuits).

[0108] A single processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs and / or a combination of a CPU and an FPGA).Furthermore, multiple processing units may be configured with a single processor.

[0109] Examples of configuring multiple processing units with a single processor include, first, a form in which one processor is configured with a combination of one or more CPUs and software, as typified by computers such as client and server, and this processor functions as multiple processing units. Second, a form in which a processor is used to realize the functions of the entire system including multiple processing units with a single IC (Integrated Circuit) chip, as typified by systems on chips (SoCs). In this way, various processing units are configured using one or more of the above-mentioned various processors as a hardware structure.

[0110] Furthermore, more specifically, the hardware structure of these various processors can be an electric circuit (circuitry) that combines circuit elements such as semiconductor elements.

[0111] The technology of the present disclosure can be appropriately combined with the various embodiments and / or various modified examples described above. Furthermore, it is not limited to the above-described embodiments, and various configurations can be adopted without departing from the spirit of the present disclosure. Furthermore, the technology of the present disclosure extends not only to programs but also to storage media that non-temporarily store programs.

[0112] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[0113] In this specification, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed by connecting them with "and / or."

[0114] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

Claims

1. A processor is provided. The processor, acquiring a first slice image of a tissue slice of at least one of a plurality of subjects constituting an administration group to which a drug candidate substance has been administered; Extracting a first feature amount from the first slice image; acquiring a plurality of second slice images of tissue slices of a plurality of subjects constituting a control group, the control group having the same attributes as the subjects constituting the administration group and being kept in the same rearing environment, and to which the drug candidate substance was not administered; extracting second feature amounts from each of the second slice images; acquiring third feature amount information relating to third feature amounts extracted from each of a plurality of third slice images showing tissue slices of a plurality of subjects constituting a control group to which the candidate substance was not administered in a past evaluation test; determining whether or not an abnormality exists in the tissue slice depicted in the first slice image based on the first feature amount, the second feature amount, and the third feature amount information; Drug discovery support device.

2. The processor, The drug discovery support device according to claim 1 , wherein the first feature amount and the second feature amount are extracted by a machine learning model trained using a plurality of the third section images.

3. the first feature amount, the second feature amount, and the third feature amount have the same number of dimensions and can be compared in the same feature amount space; the third feature amount information is information on representative positions of the third feature amounts in the feature amount space, The processor, calculating a first distance between the representative position and a position of the first feature in the feature space; calculating second distances between the representative position and each of the positions of the second features in the feature space; The drug discovery support apparatus according to claim 1 , wherein the determination is made based on the first distance and the second distance.

4. The processor, extracting the first feature amount for each of a plurality of first patch images obtained by dividing the first slice image, and calculating the first distance; extracting the second feature amount for each of a plurality of second patch images obtained by dividing the second slice image, and calculating the second distance; The drug discovery support device of claim 3, wherein the judgment is made based on a first area ratio of the first slice image of the first patch image in which the first distance is outside a set range, and a second area ratio of the second slice image of the second patch image in which the second distance is outside the set range.

5. The processor, The drug discovery support device according to claim 4 , wherein when there is a statistically significant difference between the first area ratio and the second area ratio, it is determined that an abnormality exists in the tissue slice depicted in the first slice image.

6. The processor, The drug discovery support device according to claim 4 , wherein it is determined that an abnormality exists in a tissue slice captured in the first slice image in which the first area ratio exceeds a threshold value set based on the second area ratio.

7. the first feature amount, the second feature amount, and the third feature amount have the same number of dimensions and can be compared in the same feature amount space; the third feature amount information is information on representative positions of the third feature amounts in the feature amount space, The processor, calculating a first distance between the representative position and a position of the first feature in the feature space; calculating a third distance between a representative position of the second feature amounts and a position of the first feature amount in the feature amount space; The drug discovery support apparatus according to claim 1 , wherein the determination is made based on the first distance and the third distance.

8. The processor, The drug discovery support device according to claim 7 , wherein, when a weighted sum of the first distance and the third distance is outside a set range, it is determined that an abnormality exists in the tissue slice captured in the first slice image.

9. The processor, The drug discovery support apparatus according to claim 8 , further comprising an instruction for setting a weighting coefficient by which the first distance and the third distance are multiplied in the weighted sum.

10. the first slice image and the second slice image are images obtained by photographing a slide specimen on which tissue slices of a plurality of types of organs are placed, The processor, Identifying a tissue slice of each organ from the first slice image and the second slice image; extracting the first feature amount and the second feature amount from each of the identified tissue sections of each organ; The drug discovery support device according to claim 1 , wherein whether or not an abnormality exists in each of the tissue slices of each organ depicted in the first slice image is determined based on the first feature amount, the second feature amount, and the third feature amount information.

11. acquiring a first slice image of a tissue slice of at least one subject among a plurality of subjects constituting an administration group to which a drug candidate substance has been administered; extracting a first feature amount from the first slice image; acquiring a plurality of second slice images of tissue slices of a plurality of subjects constituting a control group, the control group having the same attributes as the subjects constituting the administration group and being kept in the same rearing environment, to which the drug candidate substance was not administered; extracting second feature amounts from each of the second slice images; acquiring third feature amount information regarding third feature amounts extracted from each of a plurality of third slice images showing tissue slices of a plurality of subjects constituting a control group to which the candidate substance was not administered in a past evaluation test; and determining whether or not an abnormality exists in the tissue slice depicted in the first slice image based on the first feature amount, the second feature amount, and the third feature amount information; A method for operating a drug discovery support device comprising the steps of:

12. acquiring a first slice image of a tissue slice of at least one subject among a plurality of subjects constituting an administration group to which a drug candidate substance has been administered; extracting a first feature amount from the first slice image; acquiring a plurality of second slice images of tissue slices of a plurality of subjects constituting a control group, the control group having the same attributes as the subjects constituting the administration group and being kept in the same rearing environment, to which the drug candidate substance was not administered; extracting second feature amounts from each of the second slice images; acquiring third feature amount information regarding third feature amounts extracted from each of a plurality of third slice images showing tissue slices of a plurality of subjects constituting a control group to which the candidate substance was not administered in a past evaluation test; and determining whether or not an abnormality exists in the tissue slice depicted in the first slice image based on the first feature amount, the second feature amount, and the third feature amount information; An operating program for a drug discovery support device that causes a computer to execute processing including the steps of: