Analysis device, analysis method, and program

Abstract

This invention provides an analysis device, analysis method, and program that can present the user with the next steps based on the analysis and diagnosis results of medical information. [Solution] The system comprises: an analysis unit that acquires "first medical information" obtained by computer processing of medical information; a generation unit that generates structured data by structuring unstructured data acquired from information created by the user based on the medical information; an acquisition unit that acquires second medical information from the structured data; a comparison processing unit that compares the first medical information acquired by the analysis unit and the second medical information acquired by the acquisition unit; a judgment unit that determines whether additional action is necessary to make a definitive diagnosis based on the comparison results made by the comparison processing unit; and an output unit that outputs information prompting confirmation of the comparison results based on the judgment results of the judgment unit.

Description

Technical Field

[0001] The present invention relates to an analysis device, an analysis method, and a program.

Background Art

[0002] <00……全文翻译较长,仅展示了部分内容。完整内容如下:

Technical Field

[0001] The present invention relates to an analysis device, an analysis method, and a program.

Background Art

[0002] In recent years, with the development of AI (Artificial Intelligence) technology, attempts have been made to introduce AI-based analysis in the medical field to support the analysis and diagnosis of medical information such as image diagnosis, which was conventionally performed by doctors, using AI. For example, Patent Document 1 discloses a diagnostic support device that detects the difference between the first medical information based on the creation information of a user and the second medical information obtained by computer processing, and displays the difference on a display unit in a display form corresponding to the combination of the lesion names included in the first medical information and the lesion names included in the second medical information.

[0003] In the clinical field of medicine, it is required to perform examinations and diagnoses appropriately and quickly, and to streamline and optimize the diagnosis to reduce the burden on doctors. The introduction of AI analysis is expected to contribute to such streamlining and optimization of diagnosis.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, the technology disclosed in Patent Document 1 only shows the difference between the first medical information and the second medical information. In other words, it does not adequately consider what actions should be taken afterward depending on whether or not there is a difference, or what the next steps should be. For this reason, there is a problem that even if AI analysis is introduced, the efficiency and optimization of diagnosis will not necessarily be achieved.

[0006] For example, after a primary radiologist has interpreted a medical image, no decision is made as to whether or not to pass that image on to the next stage, the interpretation process by a secondary radiologist. Therefore, a secondary interpretation process is required for all image interpretations, and the burden on the secondary radiologist is not reduced.

[0007] The present invention has been made in view of the problems of the prior art described above, and aims to provide an analysis device, analysis method, and program that can present the next step to the user according to the analysis results and diagnostic results of medical information. [Means for solving the problem]

[0008] To solve the above problems, the invention described in claim 1 is an analysis device comprising an analysis unit that acquires first medical information obtained by computer processing of medical information, A generation unit that generates structured data by structuring unstructured data obtained from information created by the user based on the aforementioned medical information, An acquisition unit that acquires second medical information from the aforementioned structured data, A comparison processing unit that compares the first medical information obtained by the analysis unit with the second medical information obtained by the acquisition unit, Based on the comparison results obtained by the comparison processing unit, a determination unit determines whether or not additional action is necessary to perform a definitive diagnosis. An output unit outputs information prompting confirmation of the comparison result based on the judgment result of the judgment unit, It is characterized by having the following features.

[0009] Furthermore, the invention described in claim 4 is an analysis method in which each step is performed by a computer, An analysis process to obtain first medical information obtained by computer processing of medical information, A generation process that generates structured data by structuring unstructured data obtained from information created by the user based on the aforementioned medical information, The acquisition process involves obtaining second medical information from the aforementioned structured data, A comparison processing step that compares the first medical information obtained in the analysis step with the second medical information obtained in the acquisition step, Based on the comparison results obtained in the comparison processing step, a determination step is made to determine whether or not additional action is necessary to perform a definitive diagnosis. An output step outputs information prompting confirmation of the comparison result based on the judgment result in the judgment step, It is characterized by including.

[0010] Furthermore, the invention described in claim 5 is a program. On the computer, An analysis function that obtains first medical information obtained through computer processing of medical information, A generation function that generates structured data by structuring unstructured data obtained from information created by the user based on the aforementioned medical information, The acquisition function obtains second medical information from the aforementioned structured data, A comparison processing function that compares the first medical information obtained by the analysis function with the second medical information obtained by the acquisition function, Based on the comparison results obtained by the comparison processing function described above, a judgment function is provided to determine whether or not additional action is necessary to make a definitive diagnosis. An output function that outputs information prompting confirmation of the comparison results based on the judgment result of the aforementioned judgment function, It is characterized by achieving this. [Effects of the Invention]

[0011] According to the present invention, the next process can be presented to the user according to the analysis result and diagnosis result of medical information.

Brief Description of Drawings

[0012] [Figure 1] It is an overall configuration diagram of the medical image system in this embodiment. [Figure 2] It is a main part block diagram showing the functional configuration of one embodiment of the analysis device according to the present invention. [Figure 3] It is a table showing an example when natural language, which is unstructured data, is structured. [Figure 4] It is a flowchart showing the analysis process in the first QA pattern. [Figure 5] It is an explanatory diagram schematically showing the flow of the analysis process shown in FIG. 4. [Figure 6] It is a flowchart showing the comparison process. [Figure 7] It is a flowchart showing the analysis process in the second QA pattern.

Modes for Carrying Out the Invention

[0013] Hereinafter, an embodiment of an analysis device, an analysis method, and a program according to the present invention will be described. However, the scope of the invention is not limited to the illustrated examples.

[0014] 〔Configuration of Medical Image System〕 The analysis device in this embodiment performs analysis of medical images, which are medical information, for example, within a medical image system. FIG. 1 shows the system configuration of the medical image system 100.

[0015] As shown in Figure 1, the medical imaging system 100 includes modality 1, console 2, analysis device 3, image interpretation terminal 4, image server 5, etc., which are connected via a communication network N such as a LAN (Local Area Network), WAN (Wide Area Network), or the Internet. Each device constituting the medical imaging system 100 conforms to the HL7 (Health Level Seven) or DICOM (Digital Image and Communications in Medicine) standards, and communication between each device is conducted in accordance with HL7 or DICOM. The number of modality 1, console 2, image interpretation terminal 4, etc., is not particularly limited.

[0016] Modality 1 is an image generation device such as an X-ray imaging device (DR, CR), an ultrasound diagnostic device (US), a CT scanner, or an MRI scanner. Based on examination order information transmitted from a RIS (Radiology Information System) or the like (not shown), it captures images of the patient's examination target area and generates medical images as medical information. The medical images generated in Modality 1 have supplementary information (patient information, examination information, image ID, etc.) written to the image file header in accordance with the DICOM standard. The medical images with this supplementary information are then transmitted to the analysis device 3 or the image interpretation terminal 4 via the console 2 or the like.

[0017] Console 2 is an imaging control device that controls imaging in Modality 1. Console 2 outputs imaging conditions and image reading conditions to Modality 1 and acquires image data of medical images captured in Modality 1. Console 2 is composed of a control unit, display unit, operation unit, communication unit, storage unit, etc. (not shown), and each unit is connected by a bus.

[0018] The analysis device 3 is a device that performs various analyses on medical images, which are medical information. The analysis device 3 is configured as a PC, a mobile terminal, or a dedicated device. In this embodiment, the analysis device 3 includes, for example, a medical image management device such as a PACS (Picture Archiving and Communication System).

[0019] Figure 2 is a block diagram showing the functional configuration of the analysis device 3. As shown in Figure 2, the analysis device 3 is configured to include a control unit 31, a storage unit 32, a data acquisition unit 33, a data output unit 34, an operation unit 35, a display unit 36, etc., and each unit is connected by a bus 37.

[0020] The data acquisition unit 33 is an acquisition unit that acquires various types of data from external devices (for example, the console 2 or the image interpretation terminal 4 described later). The data acquisition unit 33 is configured, for example, as a network interface, and is configured to receive data from an external device connected via a communication network N by wire or wireless connection. In this embodiment, the data acquisition unit 33 is configured as a network interface, but it can also be configured as a port into which a USB memory stick or SD card can be inserted. In this embodiment, the data acquisition unit 33 acquires image data of medical images from, for example, the console 2. The data acquisition unit 33 also acquires information from the image interpretation terminal 4, such as diagnostic results related to medical images created by a user (e.g., a doctor) based on the medical image (information on the detection of lesions that can be read from the medical image), and image interpretation reports, which are the interpretation results by a radiologist (e.g., a radiologist who performs primary and secondary interpretations). Furthermore, if supplementary information is attached to the medical image, such as when a region of interest (ROI) is set by a user such as a radiologist, the data acquisition unit 33 also acquires this supplementary information.

[0021] The data output unit 34 is an output unit that outputs information processed by the analysis device 3. The destination to which the data output unit 34 outputs various types of information is not particularly limited. For example, it may be the display unit 36 ​​of the analysis device 3, or it may be the image interpretation terminal 4 or image server 5 described later, or various external display devices not shown. The data output unit 34 can be, for example, a network interface for communicating with the image interpretation terminal 4 or the image server 5, a connector for connecting to external devices (e.g., display devices not shown, printers, etc.), or ports for various media such as USB memory.

[0022] The data output unit 34 outputs information about the next process (also called "next process information") when the next process is set based on the comparison result of the control unit 31, which functions as a comparison processing unit. The process information (next process information) output from the data output unit 34 includes, for example, information on whether or not to perform additional inspections, and information on whether or not to use the "first medical information" or "second medical information" as "definitive diagnosis" information.

[0023] For example, if the comparison result from the control unit 31, which functions as a comparison processing unit, does not match between the "first medical information" and the "second medical information," the data output unit 34 may output information requesting a diagnosis from a second user as process information (next process information). Alternatively, if the comparison result from the control unit 31, which functions as a comparison processing unit, matches between the "first medical information" and the "second medical information," the data output unit 34 may output information not requesting a diagnosis from the second user as process information (next process information). Here, the second user is, for example, the secondary radiologist (or final radiologist) who performs the secondary interpretation when the "second medical information" was created by the primary radiologist performing the primary interpretation. For example, the system could output information requesting the user (e.g., the primary radiologist) to re-examine the results of comparing the "first medical information" and the "second medical information." This would allow the user to understand if the AI ​​analysis differs from their own judgment. Furthermore, the response when the "first medical information" and the "second medical information" do not match is not limited to requesting a diagnosis from a secondary radiologist (or final radiologist). For example, the system could request the user who created the "second medical information" (e.g., the primary radiologist) to reconfirm the diagnosis.

[0024] The operation unit 35 consists of a keyboard equipped with various keys, a pointing device such as a mouse, or a touch panel attached to the display unit 36. The operation unit 35 is capable of user input, and specifically outputs operation signals to the control unit 31 through key operations on the keyboard, mouse operations, or touch operations on the touch panel.

[0025] The display unit 36 ​​is configured with a monitor such as an LCD (Liquid Crystal Display) and displays various screens according to the instructions of the display signals input from the control unit 31. Note that the monitor is not limited to one, and may be provided with multiple monitors. As will be described later, the display unit 36 ​​displays statistical information and the like output from the control unit 31 (control unit 31 as a comparison processing unit) as appropriate.

[0026] The control unit 31 is composed of a CPU (Central Processing Unit), RAM (Random Access Memory), etc., and comprehensively controls the operation of each part of the analysis device 3. Specifically, the CPU reads various processing programs stored in the program storage unit 321 of the memory unit 32, loads them into RAM, and executes various processes according to the program. In this embodiment, the control unit 31 works in cooperation with the program to realize various functions as follows.

[0027] The control unit 31 functions as an analysis unit that acquires "first medical information" through computer processing of medical information. Specifically, the medical images acquired by the data acquisition unit 33 are subjected to lesion detection and analysis processing, and the detection and analysis results of one or more types of lesions are output as "first medical information." Here, the computer processing includes, for example, AI analysis using AI (Artificial Intelligence) which performs image diagnosis and image analysis, including lesion detection by CAD (Computer-Aided Diagnosis).

[0028] The control unit 31 also functions as a learning unit that learns the correspondence between medical information (medical images in this embodiment) and medical information (such as the name of a lesion), and the control unit 31 as an analysis unit obtains "first medical information" by computer processing the medical information (medical images) based on the correspondence between medical information (medical images) and medical information learned by the learning unit. In other words, a machine learning model created by training with a large amount of training data (for example, pairs of medical images showing lesions and correct labels (such as the lesion area in the medical image and the diagnostic name of the lesion (type of lesion))) is used to detect and analyze lesions from input medical images. The "first medical information" obtained in this way includes information such as the name and location of the lesion, and is attached to the image data of the medical image as supplementary information.

[0029] In this embodiment, the control unit 31 also functions as an acquisition unit (comparison target acquisition unit) that acquires a comparison target to be compared with the "first medical information" obtained by AI analysis. The control unit 31, acting as a comparison target acquisition unit, acquires "second medical information" as a comparison target to be compared with "first medical information," based on information created by the user based on medical information at the image interpretation terminal 4, etc.

[0030] Furthermore, the information created by users (e.g., radiologists) is generally unstructured data. Here, unstructured data refers to, for example, medical information itself (e.g., medical images, electrocardiogram waveform data, etc., and especially medical images in this embodiment), or information created by the user based on medical information (e.g., natural language interpretation reports created regarding medical images), etc. In contrast, the "first medical information," which is the result of AI analysis, is structured data. Therefore, in order to compare the two, the "second medical information" to be used as a comparison target must also be structured, and it must be structured data consisting of strings and other elements that can be compared with the "first medical information."

[0031] In this embodiment, structuring includes, for example, analyzing text, images, audio, video, etc., and tagging metadata. In this embodiment, structuring includes, for example, dividing unstructured data (e.g., natural language from a radiology report) obtained from user-created information (e.g., radiology report information) based on medical information (e.g., medical images) into words (morphemes) and assigning meaning to the words. Assigning meaning to words includes, for example, classifying sentences into subject, predicate, object, and complement (SVOC). Assigning meaning to words also includes classifying words according to their attributes (location, findings, disease name, affirmative / negative, affirmative / negative judgment, important findings, numerical values). In this embodiment, structured data refers to data that has undergone the above-described structure.

[0032] In this embodiment, the control unit 31 functions as a generation unit that structures unstructured data obtained from information created by the user based on medical information (e.g., medical images) and generates structured data. Specifically, for example, the memory unit 32 is equipped with a structured dictionary 323 in which words are pre-classified into predetermined attributes. The control unit 31, acting as a generation unit, classifies the words constituting the unstructured data according to their attributes and assigns meaning to them. The structured dictionary 323 also includes those obtained through machine learning.

[0033] Figure 3 shows an example of applying the structured dictionary 323 to divide and structure a radiology report (natural text, which is unstructured data) into words. As shown in Figure 3, for example, if the original text (natural language) of the radiology report is "There is a slightly lobulated mass of about 4.5 × 4 × 4.5 cm on the right side of the anterior mediastinum," we first break down the sentence into words to describe where, what size, and what kind of mass it is. This reveals that the location is "anterior mediastinum" and "on the right side," the size (numerical value) is "about 4.5 × 4 × 4.5 cm," the finding is "lobulated," and the disease name is "mass." The affirmation / negation is "affirmation" based on the phrase "there is," and it corresponds to an important finding (indicated by "〇" in Figure 3). By dividing each word into semantically assigned words in this way, the unstructured data of natural language becomes structured data. The control unit 31 also functions as an acquisition unit (comparison target acquisition unit) that acquires "second medical information" that can be compared with "first medical information" from the structured data thus structured.

[0034] Furthermore, the control unit 31 functions as a comparison processing unit that compares the "first medical information" acquired by the analysis unit with the structured "second medical information" acquired by the acquisition unit. In other words, it compares the two and outputs the results. Specifically, the goal is to clarify the agreements and disagreements (differences) between "the first medical information" and "the second medical information." Since both "Medical Information 1" and "Medical Information 2" are structured data, they can be cross-referenced and compared.

[0035] The method by which the control unit 31, acting as a comparison processing unit, compares the "first medical information" and the "second medical information" is not particularly limited. The control unit 31 may also function as a classification unit that, as a comparison processing unit, classifies the "second medical information" into information that is easily compared with the "first medical information" (referred to as "third medical information"), as a prerequisite for comparing the "first medical information" and the "second medical information". Here, "second medical information" includes one of the following: image finding name, disease name, or anatomical location name. The expressions used by users (radiologists, etc.) to represent image finding names, disease names, anatomical location names, etc., vary from person to person and are diverse. Therefore, in order to match "second medical information" with "first medical information," it is effective to classify it as "third medical information" and consolidate it into a single expression as much as possible.

[0036] For example, if the user (radiologist, etc.) uses expressions such as "nodular shadow," "tumor," and "faint circular shadow," the control unit 31, acting as a classification unit, will consolidate all of these into the expression "nodular shadow." Then, if the term "nodular shadow" appears in the "first medical information" which is the result of the AI ​​analysis, even if the original expression used by the user (radiologist, etc.) was "tumor," and the words themselves do not match, the control unit 31, acting as a comparison processing unit, will determine that it matches the AI ​​analysis result by grouping the word "tumor" as meaning "nodular shadow."

[0037] Furthermore, the control unit 31, which functions as a classification unit, is not limited to classifying "second medical information" into "third medical information." For example, the control unit 31, acting as a classification unit, may classify the expressions constituting the "second medical information" into the "first medical information," which is the result of AI analysis, when consolidating them into a single entity.

[0038] For example, if a user (such as a radiologist) uses expressions like "upper lung field," "upper lung field," or "upper lung," and the AI ​​analysis result, "First Medical Information," expresses all of these as "upper lung field," then the control unit 31, acting as a classification unit, will consolidate all of these into the expression "upper lung field," which is used in the AI ​​analysis result. This allows the "second medical information" and the "first medical information" to be compared with perfect terminology matching, improving the accuracy of the comparison.

[0039] Furthermore, the method by which the control unit 31, as the classification unit, classifies "second medical information" into "third medical information" or "second medical information" into "first medical information" is not particularly limited. For example, the memory unit 32 is equipped with a correspondence registration unit 324 that pre-registers correspondences between "second medical information" and "third medical information," and between "second medical information" and "first medical information," and the control unit 31, which acts as a classification unit, performs classification by referring to this correspondence registration unit 324.

[0040] Furthermore, the control unit 31, which functions as a classification unit, may also function as a learning unit that learns the correspondence between "second medical information" and "first medical information," and may establish the correspondence between the two through learning. When the two are matched through learning, the similarity between the "second medical information" and the "first medical information" is calculated. In this case, the control unit 31 also functions as a similarity calculation unit that calculates the similarity between the two.

[0041] Similarity is calculated by performing text mining (similarity calculation) on the data of "Second Medical Information" and "First Medical Information," for example, if "Second Medical Information" is structured text data. For calculating similarity, methods such as creating a vector space model that represents the frequency of each word's appearance in a document as vectors can be used. This method allows for the calculation of similarity between two sets of data by comparing their vectors. When calculating similarity in this way, it is preferable that the control unit 31, which acts as the similarity calculation unit, has previously learned the similarity between the "second medical information" and the "first medical information" through machine learning or the like. The method for calculating similarity is not particularly limited, and various methods can be used.

[0042] Furthermore, when the control unit 31 analyzes medical information (e.g., medical images) as an analysis unit, if the terminology for lesion information and other similar terms is broadly defined to allow for flexible matching, then even if there are slight discrepancies in expression between the "first medical information" and the "second medical information," the system can include the definition of the terms constituting the "first medical information" and determine whether they match or not. In this case, there is no need to classify the "second medical information" into the "first medical information" or the "third medical information," and the control unit 31 can perform comparison processing between the two even if it does not function as a classification unit.

[0043] Furthermore, if the control unit 31, acting as the classification unit, can classify "second medical information" by referring to information pre-registered in the correspondence registration unit 324, it is preferable to classify based on the information registered in the correspondence registration unit 324, and if the information is not registered in the correspondence registration unit 324, to classify the "second medical information" through learning as the learning unit. This allows the user's intentions to be prioritized if the user has registered correspondences in advance.

[0044] Furthermore, if the structured "second medical information" acquired by the control unit 31 as an acquisition unit is unknown lesion information, it will not correspond to anything previously registered in the correspondence registration unit 324, nor will it be possible to calculate the similarity to the "first medical information" which is the result of the AI ​​analysis. In this case, the control unit 31 functions as a lesion information classification unit, classifying the "second medical information" into "first medical information," and presenting the classification result to the user and requesting their approval.

[0045] Specifically, for example, by analyzing words that constitute the "second medical information," the system extracts the words that are deemed to be the closest match from among the words used in the AI ​​analysis results registered in the correspondence registration unit 324. The classification results of the lesion information classification unit are then output to the display unit 36, etc., and presented to the user as a classification result presentation unit, requesting approval of the classification results to determine whether or not it is acceptable to associate the unknown lesion information constituting the "second medical information" with the relevant words. As a result, once the classification result is approved by the user, the control unit 31, acting as a learning unit, learns the unknown lesion information as predetermined lesion information according to the classification result, and registers the learning result in the correspondence registration unit 324. This allows for correct classification the next time the same lesion information is read, by referring to the correspondence registration unit 324.

[0046] The storage unit 32 is composed of an HDD (Hard Disk Drive) and semiconductor memory, and stores programs for performing various processes, including the analysis of medical information such as medical images described later, as well as parameters and files necessary for the execution of said programs. The memory unit 32 of this embodiment includes, for example, a program memory unit 321 for storing various programs, as well as a learning data memory unit 322, a structured dictionary 323, an association registration unit 324, and the like.

[0047] As mentioned above, the learning data storage unit 322 stores, for example, a large amount of learning data and machine learning models created by training with this data using deep learning or the like. Furthermore, as mentioned above, structured dictionary 323 is dictionary data used to structure unstructured data such as radiology reports created by users (radiologists) and generate structured data, which is "secondary medical information." The correspondence registration unit 324 is a correspondence table or the like in which the correspondence relationships between each word when structured data are matched are pre-registered.

[0048] The image interpretation terminal 4 is a computer device that includes, for example, a control unit, an operation unit, a display unit, a storage unit, a communication unit, etc., and reads medical images, which are medical information, from an image server 5 or the like and displays them for interpretation. Users (primary radiologists and secondary radiologists) interpret medical images on the image interpretation terminal 4 and create interpretation reports, etc., which are the radiologists' diagnostic results regarding the medical images. Furthermore, users (primary radiologists, secondary radiologists) may add information such as regions of interest (ROIs) to the medical images on the image interpretation terminal 4. Information indicating a region of interest (ROI) is, for example, a mark or frame placed on a part that is judged to be a lesion, and is set on the image by the user (primary radiologist, secondary radiologist) touching the display screen or specifying it using an operating unit such as a pointing device. Information indicating a region of interest (ROI) is composed of coordinate information indicating the location and region, and is itself structured data generated based on medical information. In this case, the generation unit that generates the structured data is the control unit of the image interpretation terminal 4 that sets information indicating a region of interest (ROI) (coordinate information, etc.) in response to the operation of the user (primary radiologist, secondary radiologist).

[0049] Image server 5 is, for example, a server that constitutes a PACS (Picture Archiving and Communication Systems), and stores in a database the medical images output from modality 1, along with patient information (patient ID, patient name, date of birth, age, gender, height, weight, etc.), examination information (examination ID, examination date and time, modality type, examination site, requesting department, examination purpose, etc.), image ID of the medical image, AI analysis results output from the control unit 31 of analysis device 3 (i.e., "first medical information"), and "second medical information" which is the diagnosis result of the radiologist, such as the reading report created by the user (radiologist) on the reading terminal 4, and the matching results (comparison results) output from the control unit 31 (comparison processing unit) of analysis device 3. Furthermore, the image server 5 reads medical images and various accompanying information from the database as requested by the image interpretation terminal 4 and displays them on the image interpretation terminal 4.

[0050] [Regarding the analysis method in this embodiment] In this embodiment, the analysis method includes: an analysis step of acquiring "first medical information" obtained by computer processing of medical information such as medical images; a generation step of structuring unstructured data acquired from information created by a user based on the medical information to generate structured data; an acquisition step of acquiring "second medical information" from the generated structured data; a comparison processing step of comparing the "first medical information" acquired in the analysis step and the "second medical information" acquired in the acquisition step; and an output step of outputting the next step information based on the comparison results compared in the comparison processing step.

[0051] According to the analysis method in this embodiment, by considering both the AI ​​analysis results and the radiologist's diagnosis results, quality assurance (QA) can be performed regarding the diagnostic accuracy of medical information (medical images). In terms of QA patterns, the analysis method flow differs depending on how reliable the AI ​​analysis results are. The first QA pattern is a flow adopted when the AI ​​analysis results are not very reliable, or when the AI ​​analysis cannot be fully trusted.

[0052] Figure 4 is a flowchart showing the analysis process in the first QA pattern, and Figure 5 is an explanatory diagram schematically showing the flow of the process. As shown in Figures 4 and 5, in the first QA pattern, the target medical information (medical image) is first analyzed by AI (AI application) in the control unit 31 of the analysis device 3, and the "first medical information," which is the result of the analysis (diagnosis), is obtained (Step S1; analysis process).

[0053] Furthermore, information such as image interpretation reports created by the user (a radiologist performing the initial interpretation) when they interpret the medical information (in this case, medical images) is acquired by the data acquisition unit 33 (step S2). Then, the unstructured data acquired from this information is structured by the control unit 31, which acts as the generation unit, and structured data is generated (step S3; generation process). From this structured data, the control unit 31, which acts as the acquisition unit, acquires "second medical information" (step S4; acquisition process). Then, the control unit 31, which acts as a comparison processing unit, performs a comparison (matching) of the "first medical information" and the "second medical information" (step S5; comparison processing step).

[0054] Figure 6 is a flowchart showing the comparison process. As shown in Figure 6, in this comparison processing step, the control unit 31 first determines whether or not the lesion information, such as findings and disease names, which constitute the "second medical information," is pre-registered in the correspondence registration unit 324 (step S11). If lesion information constituting the "second medical information" is pre-registered in the correspondence registration unit 324 (Step S11; YES), the control unit 31 compares the "first medical information" and the "second medical information" based on the registered information and determines whether they match or do not match (are different) (Step S12). If they match (Step S12; YES), the control unit 31 determines that the "first medical information" and the "second medical information" match (Step S13), and if they do not match (Step S12; NO), the control unit 31 determines that the "first medical information" and the "second medical information" do not match (Step S14).

[0055] On the other hand, if the lesion information constituting the "second medical information" is not pre-registered in the correspondence registration unit 324 (step S11; NO), the control unit 31 calculates the similarity between the "first medical information" and the "second medical information" using machine learning (step S15). Then, it determines whether the calculated similarity is above a predetermined threshold (step S16), and if it is above the predetermined threshold (step S16; YES), it determines that the "first medical information" and the "second medical information" are a match (step S17). If it is not above the predetermined threshold (step S16; NO), it determines that the "first medical information" and the "second medical information" are not a match (step S18).

[0056] Furthermore, if the lesion information, such as findings or disease names, that constitute the "second medical information" is unknown, the control unit 31 classifies it into one of the words that appear in the AI's analysis results and presents the classification result to the user (e.g., a radiologist) for approval. If approved, it is registered as new correspondence information in the correspondence registration unit 324. Once a new correspondence is registered in the correspondence registration unit 324, the correspondence registered in the correspondence registration unit 324 will be referenced the next time the same lesion information is entered. On the other hand, if it is not approved, it is reclassified into a different word and the process is repeated until it is approved by the user.

[0057] In the comparison processing step, once it is determined whether the "first medical information" and the "second medical information" match or not, the process returns to Figures 4 and 5. Based on the comparison results obtained in the comparison processing step, the control unit 31 determines what should be done in the next step (step S6), and outputs the determination result as next step information from the data output unit 34, etc. (step S7; output step). For example, if "the first medical information" and "the second medical information" match, the matching information will be considered "the definitive diagnostic information."

[0058] Furthermore, if there is a discrepancy between the two, the "next process information" will output that the case should be passed on to a second user (i.e., a secondary or final radiologist different from the primary radiologist). In this case, medical image data, "first medical information," "second medical information," and matching results (for example, what content matched and what was judged to be inconsistent) are sent to the image interpretation terminal 4 operated by the secondary radiologist. The secondary radiologist then refers to this information to perform a secondary interpretation and creates a diagnostic result such as an image interpretation report as "(second medical information in the secondary interpretation)," which becomes the "definitive diagnostic information." Furthermore, "next-step information" is not limited to this; if further examinations are deemed necessary (for example, if the "second medical information" based on the primary radiologist's diagnosis includes a statement indicating the need for further examination), instructions for additional examinations will be output as "next-step information."

[0059] As shown in Figures 4 and 5, the process involves comparing the "second medical information" based on the primary radiologist's diagnosis with the "first medical information" which is the AI's analysis result (diagnosis result), and then informing the user of the next step information corresponding to the result (comparison result). This means that, for example, if the "first medical information" and the "second medical information" match, the case does not require a second interpretation, reducing the burden on the secondary radiologist. As a result, secondary radiologists can focus on cases that require a second interpretation, enabling them to perform interpretations efficiently and appropriately. Furthermore, by structuring unstructured data, such as image interpretation reports created by users, and then comparing it with the "primary medical information" which is the result of AI analysis, appropriate comparison processing can be performed.

[0060] When "definitive diagnostic information" is issued for a medical image as medical information, the medical image is associated with the "definitive diagnostic information" determined by the analysis device 3 and stored in the image server 5. Furthermore, there are no particular limitations on the storage locations for medical images and related information (such as "primary medical information," "secondary medical information," etc.), "definitive diagnostic information," etc. For example, if the analysis device 3, the image interpretation terminal 4, and the image server 5 constitute a PACS, this information may be stored and managed collectively on an information management server on the PACS. Furthermore, it is not necessary to save all of the above information; for example, only medical images and "definitive diagnostic information" may be saved together.

[0061] In this context, "definitive diagnostic information" refers to a diagnosis determined and confirmed by a physician based on medical images and the findings and analysis results obtained therefrom. The final diagnosis for a patient may be made by a clinician, for example, by comprehensively considering not only the judgment of the physician who prepared this "definitive diagnostic information," but also test data and examination data obtained from various tests and consultations.

[0062] Furthermore, if the reliability of the AI ​​analysis results is exceptionally high, and the AI ​​analysis can be fully trusted, the second QA pattern flow will be adopted. Figure 7 is a flowchart showing the analysis process in the second QA pattern. As shown in Figure 7, in the second QA pattern, the target medical information (medical image) is first analyzed by AI in the control unit 31 of the analysis device 3, and the "first medical information," which is the result of the analysis (diagnosis), is obtained (step S21). In this case, obtaining the "second medical information (from the primary radiologist's interpretation)," which is the diagnosis of the primary radiologist, is not mandatory.

[0063] The control unit 31 then determines whether the "first medical information" indicates that the medical image is normal (step S22). In other words, it determines whether the "first medical information" includes any abnormal findings regarding the medical image. If the "first medical information" determines that the medical image is normal (Step S22; YES), then the "first medical information" which is the result of the AI ​​analysis is designated as the "definitive diagnostic information" (Step S23).

[0064] In this case, the AI ​​can reduce the workload of radiologists by eliminating the need for them to interpret images that it deems normal, thereby easing their burden and enabling more efficient image diagnosis. Even if the "first medical information" determines that the medical image is normal, this does not have to be immediately considered "definitive diagnostic information." Instead, a "normal label" indicating that the medical image is normal and does not contain any abnormal findings may be attached to the "first medical information," which is the result of the AI ​​analysis. The medical image and the "first medical information" can then be sent to a secondary radiologist (or final radiologist), and the secondary radiologist's diagnosis may be considered "definitive diagnostic information." In this case, while still seeking a diagnosis from a secondary radiologist (the final radiologist), assigning a "normal" label can reduce the burden on the radiologist.

[0065] Furthermore, the application of the second QA pattern is not limited to cases where the reliability of the AI ​​analysis results is comprehensive; it may also apply to cases where the scope of that reliability is limited. In other words, for example, if a single image requires interpretation from different perspectives, the diagnosis of a specific lesion can be left to AI analysis, and the "first medical information" resulting from that analysis can be trusted. For other lesions, the "first medical information" from the AI ​​analysis and the "second medical information" based on the radiologist's diagnosis can be used in combination. Specifically, when X-ray images from health checkups are expanded to capture images of the chest and abdomen (or upper body in a longer format), the "first medical information" derived from the AI ​​analysis is trusted as the "definitive diagnostic information" for chest diagnoses. However, for abdominal diagnoses, it may be appropriate to use a combination of the "first medical information" derived from the AI ​​analysis and the "second medical information" based on the radiologist's diagnosis. In particular, "medical information" is not limited to still images; it can also include moving images, in which case the diagnostic elements increase even more than with still images. For this reason, even with a limited application that differs depending on the medical department or the subject of diagnosis, such as relying on the "first medical information," which is the result of AI analysis, as a general rule in respiratory system diagnoses, and using a combination of the "first medical information," which is the result of AI analysis, and the "second medical information," which is based on the diagnosis of a radiologist, in circulatory (blood flow) system diagnoses, it is possible to expect effects such as increased diagnostic efficiency and reduced burden on radiologists.

[0066] On the other hand, if the "first medical information" does not indicate that the medical image is normal (including abnormal findings, step S22; NO), the medical image and the "first medical information" are forwarded to a secondary radiologist (or final radiologist) for secondary interpretation (or final interpretation) (step S24). The secondary radiologist's diagnosis is then considered the "definitive diagnostic information" (step S25). In this case as well, the control unit 31, acting as a comparison processing unit, may compare the "first medical information," which is the analysis result of the AI, with the "definitive diagnostic information," which is the diagnosis result of the secondary radiologist, to determine whether they match or not. When the matching result is output, the matching result is also stored in the image server 5, etc., along with the medical image and the "definitive diagnostic information." By storing the matching result linked to the medical image and the "definitive diagnostic information," it becomes possible to refer to it later when a clinician makes a final diagnosis.

[0067] 〔effect〕 As described above, the analysis device 3 according to this embodiment includes a control unit 31 as an analysis unit that acquires "first medical information" obtained by computer processing (i.e., AI analysis) of medical information, a control unit 31 as a generation unit that structures unstructured data acquired from information created by the user based on medical information to generate structured data, a control unit 31 as an acquisition unit that acquires "second medical information" from the structured data, a control unit 31 as a comparison processing unit that compares the "first medical information" and the "second medical information" acquired by the control unit 31, and an output unit (data output unit 34) that outputs the next process information based on the comparison result. By comparing the radiologist's judgment with the AI's analysis results and indicating the next steps to the user based on the comparison, efficient and effective diagnosis can be achieved. In particular, in image diagnosis, the need for a second interpretation is determined by comparing the results with the AI's analysis. This reduces the burden on the secondary radiologist and enables more efficient image diagnosis.

[0068] If the process information output from the output unit (data output unit 34) is information on whether or not to perform additional inspections, the necessary inspections can be presented to the user based on the results of matching the "first medical information" and the "second medical information". Therefore, users can properly understand the necessary processes.

[0069] Furthermore, the process information output from the output unit (data output unit 34) can present the user with a decision on whether to use the "first medical information," which is the result of AI analysis, or the "second medical information," which is based on the radiologist's interpretation, as a definitive diagnosis, or to refer it to a secondary radiologist. Therefore, users can recognize when "primary medical information" or similar information should be used as the definitive diagnosis, thus avoiding the cumbersome process of referring all image diagnoses to secondary radiologists.

[0070] Furthermore, if the comparison results obtained by the control unit 31, which acts as a comparison processing unit, do not match between the "first medical information" and the "second medical information," the control unit outputs information requesting a diagnosis from a second user (e.g., a secondary radiologist) as process information. Therefore, users can recognize when an image needs to be referred to a secondary radiologist, avoiding the cumbersome process of having all image diagnoses referred to secondary radiologists.

[0071] Furthermore, the control unit 31 may function as a classification unit that classifies the "second medical information" into "third medical information," and compare it with the "first medical information" and the "third medical information." The content and expression used in radiology reports and other documents vary widely depending on the user (radiologist), and may not always correspond to the results of AI analysis. In such cases, it is possible to standardize the input variations from user to user, enabling accurate matching (comparison) with the AI ​​analysis results.

[0072] Furthermore, if the system includes a registration unit (correspondence registration unit 324) for pre-registering the correspondence between "second medical information" and "third medical information," and the control unit 31, acting as a classification unit, classifies "second medical information" into "third medical information" based on the correspondence between "second medical information" and "third medical information" registered in the correspondence registration unit 324, then the classification can reflect the intentions of the user who made the pre-registration.

[0073] Furthermore, if the system learns the correspondence between "second medical information" and "third medical information," and the control unit 31, acting as a classification unit, classifies "second medical information" into "third medical information" based on the learned correspondence between "second medical information" and "third medical information," then appropriate classification can be performed even if there is no prior registration.

[0074] Furthermore, the control unit 31 may function as a classification unit that classifies the "second medical information" into "first medical information," and the "first medical information" may be compared with the "second medical information." The content and expression used in radiology reports and other documents vary widely depending on the user (radiologist), and may not always correspond to the results of AI analysis. In such cases, it is possible to standardize the input variations among users and achieve a more accurate match (comparison) with the AI ​​analysis results.

[0075] Furthermore, if the system includes a registration unit (correspondence registration unit 324) for pre-registering the correspondence between "second medical information" and "first medical information," and the control unit 31, acting as a classification unit, classifies "second medical information" into "first medical information" based on the correspondence between "second medical information" and "first medical information" registered in the correspondence registration unit 324, then the classification can reflect the intentions of the user who made the pre-registration.

[0076] Furthermore, if the system learns the correspondence between "second medical information" and "first medical information," and the control unit 31, acting as a classification unit, classifies "second medical information" into "first medical information" based on the learned correspondence between "second medical information" and "first medical information," then appropriate classification can be performed even if there is no prior registration.

[0077] Furthermore, if, when there is no prior registration in the correspondence registration unit 324, the "second medical information" is classified into "first medical information" based on the learned correspondence between "second medical information" and "first medical information," then the classification can be performed appropriately even when there is no prior registration, while reflecting the user's intent.

[0078] Furthermore, the system may include a similarity calculation unit that calculates the similarity between "second medical information" and "first medical information." In this case, the system learns the correspondence between "first medical information" and "second medical information" based on the similarity. This allows for the proper learning of the correspondence between "second medical information" and "first medical information."

[0079] Furthermore, if the "second medical information" is unknown lesion information, the control unit 31 may classify the "second medical information" as "first medical information" as a lesion information classification unit, and present the classification result to the user for approval. In this case, if the classification result is approved, the unknown lesion information is learned as predetermined lesion information according to the classification result and registered in the registration unit (correspondence registration unit 324). As a result, from the next time onward, the information registered in the correspondence registration unit 324 can be used for classification, enabling efficient and appropriate classification.

[0080] Furthermore, if location information or other data is automatically added as supplementary information by the image interpretation terminal 4, such as when a user (radiologist, etc.) assigns a region of interest (ROI) to a medical image, the medical image is acquired by the analysis device 3 with structured data attached. According to this embodiment, the data attached in this structured form is also included in the "second medical information" that is compared with the "first medical information," and can be appropriately included in the comparison.

[0081] [Variation] Although embodiments of the present invention have been described above, it goes without saying that the present invention is not limited to these embodiments, and various modifications are possible without departing from the spirit of the invention.

[0082] For example, in the above embodiment, the medical information to be analyzed by the analysis device 3 is described as a medical image, but the medical information is not limited to medical "images". Information obtained through various examinations of patients may be broadly included in medical information. For example, electrocardiogram waveform data, heart sound data, blood flow data, and results obtained from various examinations may also be included in medical information.

[0083] Furthermore, in this embodiment, although the analysis device 3, image interpretation terminal 4, and image server 5 are shown as separate and independent devices in Figure 1, the analysis device 3 and image server 5, or the analysis device 3, image interpretation terminal 4, and image server 5, may be configured as a single device or a single system.

[0084] It goes without saying that the present invention is not limited to the above embodiments or modifications, and can be modified as appropriate without departing from the spirit of the present invention. [Explanation of Symbols]

[0085] 1 Modality 2 Console 3 Analysis device 4. Terminal for image interpretation 5 Image Server 31 Control Unit 32 Storage section 33 Data Acquisition Unit 36 Display section 100 Medical Imaging Systems

Claims

1. An analysis unit that acquires first medical information obtained by computer processing of medical information, A generation unit that generates structured data by structuring unstructured data obtained from information created by the user based on the aforementioned medical information, An acquisition unit that acquires second medical information from the aforementioned structured data, A comparison processing unit that compares the first medical information obtained by the analysis unit with the second medical information obtained by the acquisition unit, Based on the comparison results obtained by the comparison processing unit, a determination unit determines whether or not additional action is necessary to perform a definitive diagnosis. An output unit outputs information prompting confirmation of the comparison result based on the judgment result of the judgment unit, An analytical device equipped with the following features.

2. The aforementioned additional measures include additional tests, additional diagnoses, and additional confirmations. The analysis apparatus according to claim 1.

3. The structured data is obtained by dividing the unstructured data into words and classifying the words according to their attributes. The aforementioned attributes include findings of the lesion, size of the lesion, The analysis apparatus according to claim 1.

4. An analysis method in which each process is executed by a computer, An analysis process for obtaining first medical information obtained by computer processing of medical information, A generation process that generates structured data by structuring unstructured data obtained from information created by the user based on the aforementioned medical information, The acquisition process involves obtaining second medical information from the aforementioned structured data, A comparison processing step that compares the first medical information obtained in the analysis step with the second medical information obtained in the acquisition step, Based on the comparison results obtained in the comparison processing step, a determination step is made to determine whether or not additional action is necessary to perform a definitive diagnosis. An output step outputs information prompting confirmation of the comparison result based on the judgment result in the judgment step, An analysis method that includes this.

5. On the computer, An analysis function that obtains first medical information obtained through computer processing of medical information, A generation function that generates structured data by structuring unstructured data obtained from information created by the user based on the aforementioned medical information, The acquisition function obtains second medical information from the aforementioned structured data, A comparison processing function that compares the first medical information obtained by the analysis function with the second medical information obtained by the acquisition function, Based on the comparison results obtained by the comparison processing function described above, a judgment function is provided to determine whether or not additional action is necessary to make a definitive diagnosis. An output function that outputs information prompting confirmation of the comparison results based on the judgment result of the aforementioned judgment function, A program that makes this possible.

Images