Pathological diagnosis assistance device, pathological diagnosis assistance method, and recording medium

Abstract

In this pathological diagnosis assistance device, an endoscopic examination information acquisition means acquires endoscopic examination information including information on a lesion. A pathological slide acquisition means acquires a pathological slide of the lesion. A malignancy degree determination means determines the degree of malignancy of the lesion on the basis of the endoscopic examination information and the pathological slide. The pathological diagnosis assistance device can assist a user's decision making.

Description

Pathological diagnosis support device, pathological diagnosis support method, and recording medium

[0001] The present disclosure relates to a method for assisting pathological diagnosis.

[0002] In pathological diagnosis, a technique is known in which a pathological specimen slide is converted into a digital image and diagnostic support is provided using an image analysis technique. For example, Patent Document 1 describes an image diagnostic system and an image diagnostic method for diagnosing pathological image data.

[0003] Japanese Patent Application Laid-Open No. 2022-146822

[0004] If an area is found to be suspected to be abnormal during an endoscopic examination, tissue or polyps from that area are sampled for a pathological diagnosis, but the information from the endoscopic examination is not always used effectively when making a pathological diagnosis.

[0005] One object of the present disclosure is to provide a pathological diagnosis support device that supports pathological diagnosis using information from endoscopic examinations and pathological slides.

[0006] In one aspect of the present disclosure, a pathology diagnosis support device includes: an endoscopic examination information acquisition means for acquiring endoscopic examination information including information about a lesion; a pathology slide acquisition means for acquiring a pathology slide of the lesion; and a malignancy determination means for determining the malignancy of the lesion based on the endoscopic examination information and the pathology slide.

[0007] In another aspect of the present disclosure, a pathology diagnosis support method includes acquiring endoscopic examination information including information about a lesion, acquiring a pathology slide of the lesion, and performing a malignancy assessment to determine the malignancy of the lesion based on the endoscopic examination information and the pathology slide.

[0008] In yet another aspect of the present disclosure, a recording medium records a program that causes a computer to execute a process of acquiring endoscopic examination information including information about a lesion, acquiring a pathology slide of the lesion, and performing a malignancy assessment to determine the malignancy of the lesion based on the endoscopic examination information and the pathology slide.

[0009] According to the present disclosure, it is possible to provide a pathology diagnosis support device that supports pathology diagnosis using information from endoscopic examinations and pathology slides.

[0010] Fig. 1 is a diagram conceptually illustrating a pathological diagnosis support device according to the present disclosure; Fig. 2 is a block diagram illustrating a hardware configuration of a pathological diagnosis support device according to the present disclosure; Fig. 3 is a block diagram illustrating a functional configuration of a pathological diagnosis support device according to the present disclosure; Fig. 4 is a flowchart of pathological diagnosis processing; Fig. 5 is a block diagram illustrating a functional configuration of another pathological diagnosis support device according to the present disclosure; Fig. 6 is a flowchart of processing by another pathological diagnosis support device according to the present disclosure.

[0011] Preferred embodiments of the present disclosure will be described below with reference to the drawings. First Embodiment Overall Configuration Fig. 1 is a diagram conceptually illustrating a pathology diagnosis support device. The pathology diagnosis support device 10 generates a draft of a diagnostic report based on input endoscopic examination information and pathology slides. The endoscopic examination information is information related to the endoscopic examination. The pathology slides are created by collecting tissue from the lesion during the endoscopic examination.

[0012] 2 is a block diagram showing the hardware configuration of the pathological diagnosis support device 10. As shown in the figure, the pathological diagnosis support device 10 includes a processor 11, an interface (IF) 12, a ROM (Read Only Memory) 13, a RAM (Random Access Memory) 14, a recording medium 15, an input unit 16, and a display unit 17. Each component is connected to the other via, for example, a bus 18.

[0013] The processor 11 is a computer such as a CPU (Central Processing Unit), and executes a pre-prepared program to control the entire pathological diagnosis support device 10. Specifically, the processor 11 may be a CPU, a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an MPU (Micro Processing Unit), an FPU (Floating Point number Processing Unit), a PPU (Physics Processing Unit), a TPU (Tensor Processing Unit), a quantum processor, a microcontroller, or a combination thereof.

[0014] The processor 11 also loads programs stored in the ROM 13, the recording medium 15, or the like, and executes each process coded in the program. The processor 11 also functions as part or all of the pathological diagnosis support device 10. The processor 11 then executes the pathological diagnosis support process described below.

[0015] The IF 12 inputs and outputs data to and from external devices. Specifically, endoscopic examination information and pathology slides are input to the pathology diagnosis support device 10 via the IF 12.

[0016] The ROM 13 stores various programs executed by the processor 11. The RAM 14 stores endoscopic examination information and pathology slides input from external devices. The RAM 14 may also store machine learning models such as a determination model, a large-scale language model, and an image generation model, which will be described later. The RAM 14 is also used as a working memory while the processor 11 is executing various processes.

[0017] The recording medium 15 is a non-volatile, non-temporary storage device such as a disk-shaped recording medium or a semiconductor memory. The recording medium 15 may be configured to be detachable from the pathological diagnosis support device 10. The recording medium 15 records various programs executed by the processor 11.

[0018] The input unit 16 is, for example, a mouse, a keyboard, etc., and is used by the user to input data. The display unit 17 is, for example, a liquid crystal display device, etc., and displays a draft of the diagnostic report.

[0019] 3 is a block diagram showing the functional configuration of the pathology diagnosis support device 10 according to the first embodiment. The pathology diagnosis support device 10 functionally includes an endoscopic examination information input unit 111, a pathology slide input unit 112, a treatment plan determination unit 113, a post-follow-up prediction unit 114, and a treatment plan creation unit 115.

[0020] Endoscopic examination information and pathological slides are input to the pathological diagnosis support device 10 via the IF 12 .

[0021] The endoscopic examination information includes information about the lesion (hereinafter also referred to as "lesion information") and endoscopic operation information. The lesion information includes an endoscopic image containing the lesion (hereinafter also referred to as "lesion image"), the lesion location, the lesion size, and the imaging position of the endoscopic image. The "lesion location" refers to the location of the lesion in the lesion image, and the "lesion size" refers to the size of the lesion found. The lesion location and size can be obtained by a doctor or by AI (artificial intelligence) diagnosis. The "imaging position of the endoscopic image" refers to information indicating the location of the endoscopic image in the digestive tract at which the endoscopic image was captured. The imaging position of the endoscopic image is estimated, for example, based on the elapsed time since the endoscope was inserted. The endoscopic operation information refers to information indicating the endoscopic operation performed by the doctor and the time of that operation. Examples of endoscopic operations include switching the light wavelength from white light to highlighted light and switching the observation mode, such as magnified observation. The endoscopic examination information may also include comments from the doctor regarding the endoscopic examination, such as audio data during the endoscopic examination. The endoscopic examination information may include data in multiple formats, such as image data, video data, audio data, and text data.

[0022] A pathology slide is digital data of a pathology slide specimen created from tissue of a lesion. The pathology diagnosis support device 10 can use, as the pathology slide specimen, not only a hematoxylin-eosin (HE)-stained specimen but also specimens stained by a staining method for assessing the malignancy of a lesion, such as an immunostained specimen. A pathology slide is created by scanning the pathology slide specimen at a predetermined magnification, such as 200x or 400x, using a slide scanner. The pathology slide may also include text data containing comments from the creator of the pathology slide specimen.

[0023] Endoscopic examination information is input to an endoscopy information input unit 111, and pathology slides are input to a pathology slide input unit 112. The endoscopy information input unit 111 outputs the endoscopy information to a policy determination unit 113. The pathology slide input unit 112 outputs the pathology slides to the policy determination unit 113.

[0024] The policy decision unit 113 determines whether to perform follow-up observation or treatment for the lesion. Specifically, the policy decision unit 113 determines the malignancy of the lesion from the endoscopic examination information and the pathology slide using a previously prepared judgment model. The malignancy is, for example, a numerical value between 0 and 1, and the closer the malignancy is to 1, the higher the malignancy. The policy decision unit 113 then determines to treat the lesion if the malignancy is equal to or greater than a predetermined threshold TH1, and determines to perform follow-up observation for the lesion if the malignancy is less than a predetermined threshold TH2 and equal to or greater than a predetermined threshold TH3. Furthermore, the policy decision unit 113 determines that the lesion is benign if the malignancy is less than a predetermined threshold TH4. The predetermined thresholds TH1 to TH4 can be set arbitrarily. In this case, the thresholds have the relationship TH1>TH2>TH3>TH4.

[0025] The judgment model is, for example, a machine learning model equipped with an attention mechanism such as a Transformer, and is assumed to have been trained in advance using a dataset that combines endoscopic examination information (lesion images and lesion size), pathology slides, and correct labels of malignancy.

[0026] If the determination result is "follow-up observation," the policy determination unit 113 outputs the endoscopic examination information, the pathology slide, and the malignancy level to the post-follow-up observation prediction unit 114. On the other hand, if the determination result is "treatment," the policy determination unit 113 outputs the endoscopic examination information, the pathology slide, and the malignancy level to the treatment policy creation unit 115. Furthermore, if the determination result is "benign," the policy determination unit 113 outputs the result indicating benignity to the display unit 17 and ends the processing.

[0027] The post-follow-up prediction unit 114 generates a draft of a follow-up report. Specifically, the post-follow-up prediction unit 114 first generates feature quantities of a pathology slide from the pathology slide using a feature extraction model or the like. The feature extraction model is, for example, a neural network such as a convolutional neural network (CNN), which takes an image as input and generates feature quantities of the image. The post-follow-up prediction unit 114 then inputs the endoscopic examination information and the feature quantities of the pathology slide into a large language model (LLM) to generate a draft of the report.

[0028] The LLM is a model that learns the relationships between words in a sentence and generates related strings from a target string. By using a language model trained on sentences and paragraphs from various contexts, it is possible to generate related strings with appropriate content related to the target string. Note that the data input to the LLM is not limited to strings; for example, image data may be input and strings based on the content of the image data may be output. Examples of LLMs include ChatGPT by OpenAI. When endoscopic examination information and features of a pathology slide are input, the LLM outputs strings such as "There is a 5 mm adenomatous polyp in the XX location. Histologically, it is ..."

[0029] The post-follow-up prediction unit 114 may also generate a draft report using an LLM specialized for the medical field. The LLM specialized for the medical field may be generated by, for example, performing transfer learning on ChatGPT using medical data. The medical data may be, for example, medical records collected from medical institutions, including patient examination records and doctor findings.

[0030] Furthermore, the post-follow-up prediction unit 114 predicts the state of the lesion after a predetermined period of time has elapsed, and includes the prediction result in a draft report. Specifically, the post-follow-up prediction unit 114 uses a pre-prepared image generation model to generate a predicted image that predicts the lesion image or pathology slide after a predetermined period of time has elapsed (e.g., six months or one year). The predicted image is generated to include the lesion and its surroundings. The image generation model used by the post-follow-up prediction unit 114 is generated by performing transfer learning using image data showing the progression of the lesion on an image generation model that has been pre-trained using an open dataset.

[0031] An example of a pre-trained image generation model is Stable Diffusion using Stability AI. Furthermore, image data showing the progression of a lesion may be, for example, a dataset that combines a lesion image and malignancy level at a certain point in time with a lesion image after a predetermined period of time has elapsed, or a dataset that combines a pathology slide and malignancy level at a certain point in time with a pathology slide after a predetermined period of time has elapsed. Transfer learning may be performed using a doctor's knowledge and empirical rules in addition to image data showing the progression of a lesion.

[0032] The post-follow-up prediction unit 114 outputs a draft of a report on follow-up to the display unit 17. The report generated by the post-follow-up prediction unit 114 includes a predicted image of the lesion, allowing the user to estimate the risk from the predicted image when deciding whether or not to perform follow-up.

[0033] The treatment policy creation unit 115 generates a draft report regarding the treatment policy. Specifically, the treatment policy creation unit 115 first generates pathology slide feature values ​​from the pathology slide using a feature extraction model or the like. Then, the treatment policy creation unit 115 inputs the endoscopic examination information and the pathology slide feature values ​​into an LLM to generate a draft report including the treatment policy. The LLM of the treatment policy creation unit 115 is generated, for example, by performing transfer learning on ChatGPT using a dataset that combines the endoscopic examination information, the pathology slide feature values, and the treatment policy. When the endoscopic examination information and the pathology slide feature values ​​are input, the LLM of the treatment policy creation unit 115 outputs a string such as, "There is a 20 mm adenomatous polyp in the XX location. Histologically, it is .... Therefore, we believe that treatment such as ... is appropriate."

[0034] The treatment policy creation unit 115 may predict the state of the lesion after treatment and include the prediction result in the draft report. Specifically, the treatment policy creation unit 115 generates predicted images that predict the lesion image and pathology slide after treatment using a pre-prepared image generation model. The image generation model used by the treatment policy creation unit 115 is generated by performing transfer learning on a diffusion model that has been pre-trained using an open dataset using image data showing the lesion site before and after treatment. The image data showing the lesion site before and after treatment may be, for example, a dataset combining a pre-treatment lesion image and a post-treatment lesion image, or a dataset combining a pre-treatment pathology slide and a post-treatment pathology slide.

[0035] The treatment policy creation unit 115 outputs a draft of a report on the treatment policy to the display unit 17. The user can decide on a treatment policy by referring to the draft of the report.

[0036] In this way, the pathological diagnosis support device 10 performs a comprehensive pathological diagnosis using the endoscopic examination information and the pathological slides. This allows the pathological diagnosis support device 10 to perform a pathological diagnosis including information that is difficult for the person performing the endoscopic examination to convey to the person performing the pathological diagnosis, such as the presence of mucus, thereby making it possible to obtain more accurate diagnostic results.

[0037] In the above configuration, the endoscopic examination information input unit 111 is an example of an endoscopic examination information acquisition means, the pathology slide input unit 112 is an example of a pathology slide acquisition means, the policy determination unit 113 is an example of a malignancy assessment means and a policy determination means, and the post-observation prediction unit 114 and the treatment policy creation unit 115 are examples of a draft creation means.

[0038] [Pathological diagnosis support processing] Next, the pathological diagnosis support processing will be described. Fig. 4 is a flowchart of the pathological diagnosis support processing by the pathological diagnosis support device 10. This processing is realized by the processor 11 shown in Fig. 2 executing a program prepared in advance and operating as each element shown in Fig. 3.

[0039] First, endoscopic examination information and pathology slides are input to the pathology diagnosis support device 10 via the IF 12 (step S111). The endoscopic examination information is input to the endoscopic examination information input unit 111, and the pathology slides are input to the pathology slide input unit 112. The endoscopic examination information input unit 111 outputs the endoscopic examination information to the policy decision unit 113. The pathology slide input unit 112 outputs the pathology slides to the policy decision unit 113.

[0040] The policy decision unit 113 determines the malignancy of the lesion from the endoscopic examination information and the pathology slide using a previously prepared determination model (step S112). Then, based on the malignancy, the policy decision unit 113 determines whether to perform follow-up observation or treat the lesion (step S113). Specifically, if the malignancy is equal to or greater than a predetermined threshold TH1, the policy decision unit 113 determines to treat the lesion. If the malignancy is less than a predetermined threshold TH2 and equal to or greater than a predetermined threshold TH3, the policy decision unit 113 determines to perform follow-up observation. If the malignancy is less than a predetermined threshold TH4, the policy decision unit 113 determines that the lesion is benign and terminates processing.

[0041] If the determination result is "follow-up observation" (step S114: Yes), the policy decision unit 113 outputs the endoscopic examination information, the pathology slide, and the malignancy level to the post-follow-up observation prediction unit 114. On the other hand, if the determination result is "treatment" (step S114: No), the policy decision unit 113 outputs the endoscopic examination information, the pathology slide, and the malignancy level to the treatment policy creation unit 115.

[0042] The post-follow-up prediction unit 114 generates a draft of a follow-up report (step S115). The post-follow-up prediction unit 114 generates a draft of a follow-up report using the LLM. The post-follow-up prediction unit 114 also uses an image generation model to predict the state of the lesion after a predetermined period of time has passed, and includes the prediction result in the draft of the report. The post-follow-up prediction unit 114 outputs the generated draft of the follow-up report to the display unit 17 (step S117). Then, the processing ends.

[0043] The treatment policy creation unit 115 generates a draft of a report on a treatment policy (step S116). The treatment policy creation unit 115 generates a draft of the report on a treatment policy using the LLM. The treatment policy creation unit 115 may also use an image generation model to predict the state of the lesion after treatment and include the prediction result in the draft of the report. The treatment policy creation unit 115 outputs the generated draft of the report on a treatment policy to the display unit 17 (step S117). Then, the processing ends.

[0044] 5 is a block diagram showing the functional configuration of a pathology diagnosis support device according to Embodiment 2. The pathology diagnosis support device 200 includes an endoscopic examination information acquisition unit 201, a pathology slide acquisition unit 202, and a malignancy determination unit 203.

[0045] 6 is a flowchart of processing by the pathology diagnosis support device of the second embodiment. The endoscopic examination information acquisition means 201 acquires endoscopic examination information including information about a lesion (step S201). The pathology slide acquisition means 202 acquires a pathology slide of the lesion (step S202). The malignancy determination means 203 determines the malignancy of the lesion based on the endoscopic examination information and the pathology slide (step S203).

[0046] According to the pathology diagnosis support device 200 of the second embodiment, it is possible to provide a pathology diagnosis support device that supports pathology diagnosis using endoscopic examination information and pathology slides, thereby enabling the pathology diagnosis support device 200 to support user decision-making.

[0047] A part or all of the above-described embodiments can be described as, but not limited to, the following supplementary notes.

[0048] (Supplementary Note 1) A pathology diagnosis support device comprising: an endoscopic examination information acquisition means for acquiring endoscopic examination information including information related to a lesion; a pathology slide acquisition means for acquiring pathology slides of the lesion; and a malignancy determination means for determining the malignancy of the lesion based on the endoscopic examination information and the pathology slides.

[0049] (Supplementary Note 2) The pathology diagnosis support device according to Supplementary Note 1, further comprising a policy decision unit that decides a course of action for the lesion based on the malignancy level.

[0050] (Supplementary Note 3) The pathology diagnosis support device according to Supplementary Note 2, further comprising a draft creation means for creating a draft of a diagnostic report in accordance with the response policy.

[0051] (Appendix 4) The pathology diagnosis support device according to Appendix 3, wherein the policy decision means decides that the response policy is treatment if the malignancy level is equal to or higher than a first threshold, and decides that the response policy is observation if the malignancy level is lower than a second threshold and equal to or higher than a third threshold.

[0052] (Supplementary Note 5) The pathology diagnosis support device according to Supplementary Note 4, wherein the draft creation means generates a draft of the text of the diagnostic report using a machine learning model that receives the endoscopic examination information, the pathology slide, and the malignancy grade as input and outputs a draft of the text of the diagnostic report.

[0053] (Appendix 6) The pathology diagnosis support device according to Appendix 5, wherein the draft creation means, when the response policy is observation, uses a machine learning model that inputs the endoscopic examination information, the pathology slide, and the malignancy grade and outputs a predicted image representing the state of the lesion after a predetermined period of time has passed, to generate a predicted image, and the draft of the diagnostic report includes a draft of the text of the diagnostic report and the predicted image.

[0054] (Appendix 7) The pathology diagnosis support device according to Appendix 5, wherein the draft creation means, when the response policy is treatment, generates a post-treatment image using a machine learning model that inputs the endoscopic examination information, the pathology slide, and the malignancy grade and outputs a post-treatment image representing the state of the lesion after treatment, and the draft of the diagnostic report includes a draft of the text of the diagnostic report and the post-treatment image.

[0055] (Supplementary Note 8) The pathology diagnosis support device according to Supplementary Note 1, wherein the malignancy determination means determines the malignancy using a machine learning model that receives the endoscopic examination information and the pathology slide as input and outputs the malignancy.

[0056] (Appendix 9) A pathology diagnosis support method that acquires endoscopic examination information including information about a lesion, acquires a pathology slide of the lesion, and performs a malignancy determination that determines the malignancy of the lesion based on the endoscopic examination information and the pathology slide.

[0057] (Supplementary Note 10) The pathology diagnosis support method according to Supplementary Note 9, further comprising: determining a course of action for the lesion based on the malignancy level.

[0058] (Supplementary Note 11) The pathology diagnosis support method according to Supplementary Note 10, further comprising: creating a draft of a diagnostic report in accordance with the response policy.

[0059] (Supplementary Note 12) The pathology diagnosis support method according to Supplementary Note 11, wherein the policy decision is to decide that the response policy is treatment if the malignancy level is equal to or higher than a first threshold, and to decide that the response policy is observation if the malignancy level is lower than a second threshold and equal to or higher than a third threshold.

[0060] (Appendix 13) The pathology diagnosis support method according to Appendix 12, wherein the draft creation generates a draft of the text of the diagnostic report using a machine learning model that inputs the endoscopic examination information, the pathology slide, and the malignancy grade and outputs a draft of the text of the diagnostic report.

[0061] (Appendix 14) The pathology diagnosis support method according to Appendix 13, wherein, when the response policy is observation, the draft creation involves generating a predicted image using a machine learning model that inputs the endoscopic examination information, the pathology slide, and the malignancy grade and outputs a predicted image representing the state of the lesion after a predetermined period of time has passed, and the draft of the diagnostic report includes a draft of the text of the diagnostic report and the predicted image.

[0062] (Appendix 15) The pathology diagnosis support method according to Appendix 13, wherein, when the response policy is treatment, the draft creation involves generating a post-treatment image using a machine learning model that inputs the endoscopic examination information, the pathology slide, and the malignancy grade and outputs a post-treatment image representing the state of the lesion after treatment, and the draft of the diagnostic report includes a draft of the text of the diagnostic report and the post-treatment image.

[0063] (Supplementary Note 16) The pathology diagnosis support method according to Supplementary Note 9, wherein the malignancy determination is performed using a machine learning model that receives the endoscopic examination information and the pathology slide as input and outputs the malignancy.

[0064] (Appendix 17) A recording medium having recorded thereon a program that causes a computer to execute a process of acquiring endoscopic examination information including information about a lesion, acquiring a pathology slide of the lesion, and performing a malignancy assessment that assesses the malignancy of the lesion based on the endoscopic examination information and the pathology slide.

[0065] (Supplementary Note 18) The recording medium according to Supplementary Note 17, further comprising a policy decision for deciding a course of action for the lesion based on the malignancy level.

[0066] (Supplementary Note 19) The recording medium according to Supplementary Note 18, further comprising: a step of creating a draft of a diagnostic report in accordance with the response policy.

[0067] (Appendix 20) The recording medium described in Appendix 19, wherein the policy decision is to decide that the response policy is treatment if the malignancy level is equal to or higher than a first threshold, and to decide that the response policy is observation if the malignancy level is less than a second threshold and equal to or higher than a third threshold.

[0068] (Appendix 21) The recording medium according to Appendix 20, wherein the draft creation is performed using a machine learning model that inputs the endoscopic examination information, the pathology slides, and the malignancy grade, and outputs a draft of the diagnostic report text.

[0069] (Appendix 22) When the response policy is observation, the draft is created by using a machine learning model that inputs the endoscopic examination information, the pathology slide, and the malignancy grade and outputs a predicted image representing the state of the lesion after a predetermined period of time has passed to generate a predicted image, and the draft of the diagnostic report is a recording medium described in Appendix 21 that includes a draft of the text of the diagnostic report and the predicted image.

[0070] (Appendix 23) When the response policy is treatment, the draft creation generates a post-treatment image using a machine learning model that inputs the endoscopic examination information, the pathology slide, and the malignancy grade and outputs a post-treatment image representing the state of the lesion after treatment, and the draft of the diagnostic report is the recording medium described in Appendix 21 that includes a draft of the text of the diagnostic report and the post-treatment image.

[0071] (Supplementary Note 24) The recording medium according to Supplementary Note 17, wherein the malignancy determination is performed using a machine learning model that receives the endoscopic examination information and the pathology slide as input and outputs the malignancy.

[0072] Although the present disclosure has been described above with reference to the embodiments and examples, the present disclosure is not limited to the above-described embodiments and examples. Various modifications that can be understood by a person skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure.

[0073] 10 Pathological diagnosis support device 111 Endoscopy examination information input unit 112 Pathological slide input unit 113 Policy decision unit 114 Post-follow-up prediction unit 115 Treatment policy creation unit

Claims

1. An endoscopic examination information acquisition means for acquiring endoscopic examination information including information on a lesion, a pathological slide acquisition means for acquiring a pathological slide of the lesion, and a malignancy determination means for determining the malignancy of the lesion based on the endoscopic examination information and the pathological slide. A pathological diagnosis support apparatus comprising the same.

2. The pathological diagnosis support apparatus according to claim 1, further comprising a policy determination means for determining a response policy for the lesion based on the malignancy.

3. The pathological diagnosis support apparatus according to claim 2, further comprising a draft creation means for creating a draft of a diagnosis report according to the response policy.

4. The policy determination means determines the response policy as treatment when the malignancy is equal to or higher than a first threshold value, and determines the response policy as follow-up observation when the malignancy is less than a second threshold value and equal to or higher than a third threshold value. The pathological diagnosis support apparatus according to claim 3.

5. The draft creation means generates a draft of the text of the diagnosis report using a machine learning model that takes the endoscopic examination information, the pathological slide, and the malignancy as inputs and outputs a draft of the text of the diagnosis report. The pathological diagnosis support apparatus according to claim 4.

6. When the response policy is follow-up observation, the draft creation means generates a predicted image representing the state of the lesion after a predetermined period has elapsed using a machine learning model that takes the endoscopic examination information, the pathological slide, and the malignancy as inputs and outputs the predicted image. The draft of the diagnosis report includes the draft of the text of the diagnosis report and the predicted image. The pathological diagnosis support apparatus according to claim 5.

7. When the response policy is treatment, the draft creation means generates a post-treatment image representing the state of the lesion after treatment using a machine learning model that takes the endoscopic examination information, the pathological slide, and the malignancy as inputs and outputs the post-treatment image. The draft of the diagnosis report includes the draft of the text of the diagnosis report and the post-treatment image. The pathological diagnosis support apparatus according to claim 5.

8. The malignancy determination means determines the malignancy using a machine learning model that takes the endoscopic examination information and the pathological slide as inputs and outputs the malignancy. The pathological diagnosis support apparatus according to claim 1.

9. A pathological diagnosis support method for acquiring endoscopic examination information including information on a lesion, acquiring a pathological slide of the lesion, and performing a malignancy determination for determining the malignancy of the lesion based on the endoscopic examination information and the pathological slide.

10. The pathological diagnosis support method according to claim 9, further comprising making a policy decision to determine a response policy for the lesion based on the malignancy degree.

11. The pathological diagnosis support method according to claim 10, further comprising making a draft creation to create a draft of a diagnosis report according to the response policy.

12. In the policy decision, when the malignancy degree is equal to or higher than a first threshold value, the response policy is determined as treatment; when the malignancy degree is less than a second threshold value and equal to or higher than a third threshold value, the response policy is determined as follow-up observation. The pathological diagnosis support method according to claim 11.

13. In the draft creation, using a machine learning model that takes the endoscopic examination information, the pathological slide, and the malignancy degree as inputs and outputs a draft of the text of the diagnosis report, a draft of the text of the diagnosis report is generated. The pathological diagnosis support method according to claim 12.

14. In the draft creation, when the response policy is follow-up observation, using a machine learning model that takes the endoscopic examination information, the pathological slide, and the malignancy degree as inputs and outputs a predicted image representing the state of the lesion after a predetermined period, a predicted image is generated. The draft of the diagnosis report includes the draft of the text of the diagnosis report and the predicted image. The pathological diagnosis support method according to claim 13.

15. In the draft creation, when the response policy is treatment, using a machine learning model that takes the endoscopic examination information, the pathological slide, and the malignancy degree as inputs and outputs a post-treatment image representing the state of the lesion after treatment, a post-treatment image is generated. The draft of the diagnosis report includes the draft of the text of the diagnosis report and the post-treatment image. The pathological diagnosis support method according to claim 13.

16. In the malignancy degree determination, using a machine learning model that takes the endoscopic examination information and the pathological slide as inputs and outputs the malignancy degree, the malignancy degree is determined. The pathological diagnosis support method according to claim 9.

17. A recording medium recording a program for causing a computer to execute a process of obtaining endoscopic examination information including information about a lesion, obtaining a pathological slide of the lesion, and performing a malignancy degree determination to determine the malignancy degree of the lesion based on the endoscopic examination information and the pathological slide.

18. The recording medium according to claim 17, further comprising making a policy decision to determine a response policy for the lesion based on the malignancy degree.

19. The recording medium according to claim 18, further comprising making a draft creation to create a draft of a diagnosis report according to the response policy.

20. The recording medium according to claim 19, wherein when the malignancy is greater than or equal to a first threshold, the policy decision determines the countermeasure policy as treatment, and when the malignancy is less than a second threshold and greater than or equal to a third threshold, the countermeasure policy is determined as observation of the course of the disease.

Images