WO2012111288A1 - 類似症例検索装置および類似症例検索方法 - Google Patents
類似症例検索装置および類似症例検索方法 Download PDFInfo
- Publication number
- WO2012111288A1 WO2012111288A1 PCT/JP2012/000885 JP2012000885W WO2012111288A1 WO 2012111288 A1 WO2012111288 A1 WO 2012111288A1 JP 2012000885 W JP2012000885 W JP 2012000885W WO 2012111288 A1 WO2012111288 A1 WO 2012111288A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- diagnosis
- similar
- diagnostic
- flow
- item
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H15/00—ICT specially adapted for medical reports, e.g. generation or transmission thereof
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
Definitions
- the present invention relates to a case search apparatus and a case search for searching for case data similar to a target interpretation report from a case database storing case data including medical images and interpretation reports of the medical images in the medical information system field. Regarding the method.
- search reports a large number of document data (hereinafter referred to as “interpretation reports”) describing interpretation / diagnosis results are stored in a database. Therefore, an interpretation report having a character string similar to the character string included in the interpretation report (hereinafter referred to as “search target”) input by the doctor is searched from the database. Then, the searched interpretation report and medical image are output as search results (see Non-Patent Document 1).
- an interpretation report is retrieved using a vector having the number of each keyword included in the interpretation report as an element.
- the similarity between each image interpretation report in the database with respect to the search target is calculated by calculating the distance between the vector to be searched and the vector of the image interpretation report in the database. Then, an interpretation report including an interpretation report having a high degree of similarity is preferentially searched.
- an interpretation report having a high matching rate between a search target and a character string can be searched as an interpretation report similar to the search target.
- doctor case data including an interpretation report with a high match rate between the search target and the character string does not reveal how close it is medically to the currently diagnosed case. For this reason, doctors often cannot determine a disease name. Therefore, a doctor diagnoses a symptom to be searched using a plurality of case data searched using other search keywords. That is, in the conventional method, there is a problem that appropriate case data cannot be retrieved from a plurality of case data.
- the present invention has been made to solve the above-described conventional problems, and an object thereof is to provide a similar case retrieval apparatus that can retrieve appropriate case data from a plurality of case data. To do.
- the similar case retrieval apparatus is based on an interpretation report that is document data in which a diagnosis item to be diagnosed when a medical image is interpreted and a diagnosis result indicating a symptom of the diagnosis item are described.
- a keyword extraction unit that extracts a keyword
- a diagnosis tree storage unit that stores a diagnosis tree in which a plurality of diagnosis flows including a diagnosis item until a disease name is determined and a symptom of the diagnosis item are represented in a tree structure
- a diagnostic tree analysis unit that extracts a target diagnostic flow that is a diagnostic flow corresponding to the interpretation report based on the keyword, and a difficulty level of a diagnostic item that is a degree of difficulty in determining a symptom of the diagnostic item, Or a plurality of diagnosis flows included in the diagnosis tree stored in the diagnosis tree storage unit based on the difficulty level of the disease name, which is the degree of difficulty in determining the disease name Search for case data corresponding to the similar diagnosis flow from among a plurality of case data stored in a case data storage
- case data it is possible to search for case data using the similar diagnosis flow extracted based on the difficulty level. Therefore, since case data can be searched effectively by paying attention to a diagnostic item or disease name having a high degree of difficulty, it is possible to search for appropriate case data.
- the difficulty level of the disease name may be calculated such that the difficulty level increases as the number of diagnosis flows including the same disease name increases in the diagnosis tree.
- the difficulty level of the diagnostic item may be calculated such that the difficulty level increases as the number of diagnostic item branches in the diagnostic tree increases.
- the difficulty level of the disease name may be calculated such that the difficulty level increases as the number of disease names branched from the symptoms of the diagnosis item in the diagnosis tree increases.
- the difficulty level of the disease name may be calculated such that the difficulty level is higher as the value that is predetermined for each disease name and the value indicating the ease of mistake of the disease name is larger.
- the similar diagnosis flow extraction unit extracts the similar diagnosis flows for each diagnosis tree. May be.
- This configuration makes it possible to extract a similar diagnosis flow even when a plurality of target diagnosis flows are extracted.
- the present invention can be realized not only as such a similar case search apparatus, but also as a similar case search method in which the operation of a characteristic component included in such a similar case search apparatus is a step. Can do. Moreover, this invention can also be implement
- a non-temporary recording medium such as a CD-ROM (Compact Disc Only Memory) or a transmission medium such as the Internet.
- case data can be searched effectively by paying attention to a diagnosis item or disease name having a high degree of difficulty, so that appropriate case data can be searched from a plurality of case data.
- FIG. 12 is a block diagram illustrating a detailed functional configuration of the similar case search unit.
- FIG. 13 is a diagram illustrating a usage state of the similar case search system according to the embodiment of the present invention.
- FIG. 14 is a flowchart showing a flow of processing performed by the similar case retrieval system in the embodiment of the present invention.
- FIG. 15 is a flowchart showing the flow of processing performed by the keyword extraction unit.
- FIG. 16 is a flowchart illustrating a flow of processing performed by the difficulty level evaluation unit.
- FIG. 17 is a flowchart showing a flow of processing performed by the similar diagnosis flow extraction unit.
- FIG. 18 is a flowchart showing the flow of processing performed by the similar diagnosis flow extraction unit.
- FIG. 19 is a flowchart showing a flow of processing performed by the similar diagnosis flow extraction unit.
- FIG. 20 is a flowchart showing a flow of processing performed by the similar diagnosis flow extraction unit.
- FIG. 21 is a flowchart showing the flow of processing performed by the similar diagnosis flow extraction unit.
- FIG. 22 is a diagram for explaining a method for changing a plurality of diagnosis trees.
- FIG. 23 is a flowchart showing a flow of processing performed by the similar case search unit.
- FIG. 24 is a flowchart illustrating a flow of processing performed by the search case control unit.
- FIG. 25 is a diagram showing a diagnosis tree used in a similar case search experiment.
- FIG. 26 is a diagram illustrating a result of a similar case search experiment.
- FIG. 27 is a diagram illustrating a result of a similar case search experiment.
- FIG. 28 is a diagram illustrating a result of a similar case search experiment.
- FIG. 29 is a diagram illustrating a result of
- FIG. 1 is a flowchart showing a medical image interpretation procedure by a doctor.
- FIG. 2 is a diagram illustrating an example of an interpretation report and a medical image.
- Diagnosis item refers to the type and site of a lesion that is a diagnosis target when a medical image is interpreted.
- Symptom refers to the state of a lesion or site.
- diagnosis result refers to a symptom of the diagnosis item obtained by a doctor diagnosing the diagnosis item.
- the “interpretation report” is document data in which diagnosis items and diagnosis results are described. For example, when the interpretation report includes a description “the tumor boundary is clear and smooth”, the diagnosis item is “tumor boundary” and the diagnosis result is “clear and smooth”.
- Disease name refers to the name of the disease that is finally determined by the doctor based on the diagnosis result of each diagnosis item.
- Cosmetic data refers to data including a medical image and an interpretation report.
- the doctor interprets the medical image and diagnoses the diagnostic item (S01).
- diagnosis item symptom cannot be determined (No in S02)
- the doctor is similar to the diagnosis item and diagnosis result of the medical image currently being interpreted,
- the diagnosis result is searched (S03), and the diagnosis item is diagnosed using the search result (S01).
- the doctor inputs the diagnostic item and the diagnostic result in the finding column of the interpretation report as shown in FIG.
- step S05 if the diagnosis of the diagnostic item necessary for determining the disease name has not been completed (No in S05), the process returns to step S01 again to diagnose the next diagnostic item.
- Case data (hereinafter referred to as “similar case”) is searched (S07), and it is determined whether or not a disease name can be determined using the search result.
- Physician interprets medical images according to the above procedure. At this time, if he / she is not sure about the diagnosis of the symptom of the diagnosis item or the determination of the disease name, the doctor searches for similar cases. Then, the doctor can obtain a clue for diagnosing the diagnosis item or determining the disease name by comparing the search result with the medical image currently being interpreted.
- Physician diagnoses diagnosis items in order and determines the disease name. At this time, all diagnosis items may be diagnosed, and there may be no disease name determined even though there is no diagnosis item to be diagnosed. That is, there are cases where there are many disease name options even after the diagnosis of the diagnosis item is completed. In such a case, it is difficult for a doctor to determine a disease name. Therefore, it is considered highly likely that a similar case search is required.
- Physician diagnoses various types of diagnostic items (shape, pattern, etc.) before determining the disease name. As the number of types of diagnosis items increases, it is more difficult to determine a disease name, and it is likely that a similar case search is required.
- the above five items all relate to the difficulty of diagnosing symptoms or determining the name of the disease, such as how much the doctor is at a loss when diagnosing the diagnosis item or determining the name of the disease. Therefore, if it is possible to present past case data that is likely to be lost by using how much the doctor is lost for each diagnosis item and disease name, it is considered useful for the doctor.
- a similar case of an interpretation report (symptom A) to be searched a plurality of symptoms (for example, symptoms A, B, C, D) that can be determined for a diagnosis item as well as presenting the symptom A , E, F, G), it is useful for doctors if it is possible to search for case data that includes a symptom (for example, only symptom A, D, G).
- the lost cases (symptoms D and G) are different from the character string to be searched (symptom A), and therefore cannot be searched by the similar case search method according to the matching degree of the character strings in the interpretation report in the prior art.
- doctors determine the symptoms of each diagnostic item based on medical knowledge.
- the doctor determines the final disease name by repeating the determination of the symptom of the diagnosis item a plurality of times.
- a disease name, a diagnostic item to be diagnosed before the disease name is determined, and a symptom of the diagnostic item are referred to as a diagnostic flow.
- the doctor first diagnoses a predetermined diagnostic item. For example, when performing a breast ultrasound diagnosis, a doctor first determines a symptom at a boundary portion (the boundary is clear / smooth / clear rough / unclear). Next, the doctor diagnoses the next diagnosis item according to the diagnosis result of the boundary portion. For example, if the boundary is clear and smooth, the doctor next performs shape diagnosis. If the boundary portion is clearly rough or unclear, the doctor next diagnoses the break of the boundary line. That is, the doctor diagnoses the diagnostic item while changing the diagnostic item to be diagnosed next according to the diagnosis result of the diagnostic item, and finally determines the disease name.
- a predetermined diagnostic item For example, when performing a breast ultrasound diagnosis, a doctor first determines a symptom at a boundary portion (the boundary is clear / smooth / clear rough / unclear). Next, the doctor diagnoses the next diagnosis item according to the diagnosis result of the boundary portion. For example, if the boundary is clear and smooth, the doctor next performs shape diagnosis. If the boundary portion is clearly rough or unclear, the
- FIG. 3 shows an example of a diagnosis tree for the shape of breast ultrasound diagnosis.
- Evaluation index for (1) In order to evaluate the similarity of symptoms that can be determined as symptoms of a diagnostic item, look at how many of the same disease names exist ahead of the symptom that determined the diagnostic item. That's fine. For example, when the diagnosis item a is determined as shown in FIG. 4 and there are many same disease names in the pattern A and the pattern B, it is considered that the symptom is not an element for distinguishing the disease names. Therefore, there is no characteristic that divides the target disease name into different symptoms, that is, since the symptoms to be judged are similar, it is considered difficult to make a judgment between the symptoms. Therefore, the difficulty level is evaluated by counting the number of disease names under a certain diagnosis item from the diagnosis tree.
- Evaluation index for (2) In order to evaluate that there are many types of symptoms from the diagnostic tree, the number of types of symptoms in the diagnostic tree (the number of branches for the diagnostic item) may be used. For example, the degree of difficulty is evaluated from the number of symptoms (4: irregular shape, polygon shape, leaf shape, circle / ellipse) for the diagnostic item of shape shown in FIG. As shown in FIG. 5, as the number of types increases, the degree of difficulty is evaluated as a diagnostic item.
- Evaluation index for (4) A case where a disease name is not determined even if all diagnosis items are diagnosed will be described. Assume that a diagnosis item is diagnosed as indicated by an arrow when there is a diagnosis tree as shown in FIG. In such a case, it is necessary to determine one of the disease names 1, 2, and 3 even after all the diagnosis items have been diagnosed. At this time, the difficulty level may be evaluated from the number of disease names in the diagnosis tree. For example, in FIG. 7, the number of disease names of disease names 1, 2, and 3 (three). In FIG. 3, the difficulty level is evaluated from the number of disease names (2) for a diagnostic tree whose boundary is clear and smooth and whose shape is irregular.
- the ease of mistake of the disease name is evaluated.
- the ease of diagnosing a disease name is calculated by using as an index how much the interpretation result has been changed in a later diagnosis with respect to the interpretation report result.
- a doctor's diagnosis is roughly divided into two stages: interpretation by an interpreting doctor and diagnosis by a clinician.
- the probability that the disease name at the time of interpretation is different from the disease name finally determined by the doctor may be easily mistaken.
- there are a primary interpretation and a secondary interpretation in the interpretation process by an interpreting doctor there are a primary interpretation and a secondary interpretation in the interpretation process by an interpreting doctor. The probability that the results of the primary interpretation and the secondary interpretation are different may be easily mistaken for a disease name.
- these are collectively described as disease data.
- Evaluation index for (5) In order to evaluate that a disease name is not determined for one type of diagnostic item, the number of items may be extracted from the interpretation report and the number of diagnostic trees used may be calculated. A case where there is a diagnostic tree related to shape and a diagnostic tree related to color as shown in FIG. 8 will be described. At this time, it is assumed that a diagnosis is made as indicated by arrows and a disease name 1 is reached. In this case, the interpretation report includes a character string (for example, “item a”) regarding each diagnosis tree. Therefore, by analyzing the character string of the interpretation report and applying it to each diagnostic tree, it is possible to know how many diagnostic trees are used. Further, as the number of diagnosis trees increases, a more complex diagnosis is performed, and it is evaluated that the difficulty level is high.
- a character string for example, “item a”
- the diagnostic difficulty of the above items can be estimated from the diagnostic tree for the above evaluation index.
- the calculation of each difficulty level will be described in detail when the embodiment is described.
- the inventors of the present application have come up with the idea that the target similar case search can be achieved by estimating the difficulty level from the interpretation report and the diagnostic tree including the diagnostic flow.
- FIG. 9 is a block diagram showing a configuration of similar case search system 100 in the present embodiment.
- the similar case search system 100 includes a similar case search device 1, an input unit 2, a similar case display unit 6, and an interpretation support database 10.
- the input unit 2 is a device (for example, a keyboard or a mouse) for a doctor to input an interpretation report.
- the input unit 2 sends text data and selection results input by the doctor to the keyword extraction unit 3.
- the similar case display unit 6 receives the case data from the similar case search unit 5 and presents it to the doctor.
- the similar case display unit 6 is, for example, a display for a PC (Personal Computer), a TV (Television), a medical interpretation monitor, or the like.
- the similar case search device 1 is a device for searching case data similar to the interpretation report input by the input unit 2 from the case data storage unit 9.
- the similar case search device 1 includes a keyword extraction unit 3, a difficulty level evaluation unit 4, and a similar case search unit 5.
- the keyword extraction unit 3 receives the text data of the interpretation report from the input unit 2, extracts keywords related to medical terms, diagnosis items, and diagnosis results from the text data, and sends them to the difficulty level evaluation unit 4. That is, the keyword extraction unit 3 receives an interpretation report from the input unit 2 and extracts a keyword from the received interpretation report.
- FIG. 10 is a block diagram showing a detailed functional configuration of the keyword extraction unit 3. As shown in FIG. 10, the keyword extraction unit 3 includes a character string analysis unit 31 and a character string comparison unit 32.
- the character string analysis unit 31 performs character string analysis on the interpretation report received from the input unit 2, and classifies nouns, particles, and the like. Thereafter, the character string analysis unit 31 sends the analyzed character string to the character string comparison unit 32.
- the character string comparison unit 32 receives the character strings analyzed by the character string analysis unit 31, compares the character strings with the character strings stored in the keyword dictionary storage unit 7, and if they match, the keyword (Hereinafter also referred to as “character string”) is sent to the difficulty level evaluation unit 4.
- the difficulty level evaluation unit 4 receives the character string extracted from the keyword extraction unit 3 and reads the diagnosis tree from the diagnosis tree storage unit 8. The difficulty level evaluation unit 4 calculates the difficulty level of the diagnosis item from these character strings and the diagnosis tree, and sends it to the similar case search unit 5.
- FIG. 11 is a block diagram showing a detailed functional configuration of the difficulty level evaluation unit 4. As shown in FIG. 11, the difficulty level evaluation unit 4 includes a diagnosis tree analysis unit 41 and a similar diagnosis flow extraction unit 42.
- the diagnosis tree analysis unit 41 refers to the diagnosis tree storage unit 8 and determines the target diagnosis flow, which is a diagnosis flow corresponding to the interpretation report input to the input unit 2, based on the keywords extracted by the keyword extraction unit 3. Extract.
- the diagnosis tree analysis unit 41 compares the character string received from the keyword extraction unit 3 with each of a plurality of diagnosis trees stored in the diagnosis tree storage unit 8, and selects the diagnosis tree including the keyword. To do. Next, the diagnosis tree analysis unit 41 analyzes, from the selected diagnosis tree and character string, which diagnosis flow the current interpretation report corresponds to in the aforementioned diagnosis tree. Then, the diagnosis tree analysis unit 41 sends the diagnosis tree and in-tree position information (target diagnosis flow) corresponding to the current interpretation report to the similar diagnosis flow extraction unit 42.
- the similar diagnosis flow extraction unit 42 is similar to the target diagnosis flow among a plurality of diagnosis flows included in the diagnosis tree stored in the diagnosis tree storage unit 8 based on the difficulty level of the diagnosis item or the difficulty level of the disease name. Extract a similar diagnosis flow.
- the similar diagnosis flow extraction unit 42 includes, for example, a plurality of diagnosis flows such that a diagnosis flow including a diagnosis item or a disease name having a high difficulty included in the target diagnosis flow is extracted as a similar diagnosis flow. Similar diagnosis flow is extracted from.
- the similar diagnosis flow extraction unit 42 receives the diagnosis tree and the position information in the tree, and acquires the difficulty level of each diagnosis item included in the target diagnosis flow corresponding to the current interpretation report. Next, the similar diagnosis flow extraction unit 42 extracts a diagnosis flow including a diagnosis item with a high degree of difficulty as a similar diagnosis flow, and sends it to the similar case search unit 5.
- the similar case search unit 5 receives the similar diagnosis flow from the difficulty level evaluation unit 4, calculates the similarity of the case data in the case data storage unit 9, and sends case data having a high similarity to the similar case display unit 6. That is, the similar case search unit 5 searches for case data corresponding to the similar diagnosis flow from among a plurality of case data stored in the case data storage unit 9. That is, the similar case search unit 5 searches for case data diagnosed according to the similar diagnosis flow. Further, the similar case search unit 5 also searches case data diagnosed according to the target diagnosis flow.
- FIG. 12 is a block diagram showing a detailed functional configuration of the similar case search unit 5.
- the similar case search unit 5 includes a search case control unit 51 and a similarity evaluation unit 52.
- the search case control unit 51 receives a similar diagnosis flow from the difficulty level evaluation unit 4. If there is no input from the input unit 2, the search case control unit 51 sends the similar diagnosis flow to the similarity evaluation unit 52 as it is. On the other hand, when there is an input from the input unit 2, the search case control unit 51 weights the similar diagnosis flow and sends the similar diagnosis flow to the similarity evaluation unit 52.
- the similarity evaluation unit 52 receives the similar diagnosis flow, and calculates the similarity between the diagnosis flow of the case data stored in the case data storage unit 9 and the similar diagnosis flow. Next, case data having a high degree of similarity is sent to the similar case display unit 6.
- the interpretation support database 10 includes a keyword dictionary storage unit 7, a diagnosis tree storage unit 8, and a case data storage unit 9.
- Keyword dictionary storage unit 7 stores medical terms used for keyword extraction.
- the diagnosis tree storage unit 8 stores information related to a diagnosis flow such as diagnosis items and types of symptoms thereof. That is, the diagnostic tree storage unit 8 stores a diagnostic tree in which a plurality of diagnostic flows including a diagnostic item until a disease name is determined and symptoms of the diagnostic item are represented in a tree structure.
- the case data storage unit 9 stores past interpretation reports and medical images used for diagnosis. That is, the case data storage unit 9 stores case data that is a set of an interpretation report and a medical image.
- FIG. 13 is a diagram showing a usage pattern of the similar case search system 100 in the present embodiment.
- the similar case search system 100 includes a similar case search device 1, an input unit 2, a similar case display unit 6, and an interpretation support database 10.
- the similar case search system 100 realizes a difficulty evaluation based on the interpretation report input by the input unit 2, searches the interpretation support database 10 for case data having a high degree of difficulty in diagnosis with respect to the input interpretation report, The result is presented on the similar case display unit 6.
- FIG. 14 is a flowchart showing a flow of processing performed in the similar case search system 100.
- FIG. 14 corresponds to steps S03 and S07 in FIG.
- step S ⁇ b> 10 the input unit 2 receives an input of an interpretation report from a doctor and sends it to the similar case search device 1.
- a specific explanation will be given, taking as an example a case where a doctor inputs “interpretation is clear and smooth, shape is irregular. Suspected papillary duct cancer” in an interpretation report as shown in FIG. To do.
- step S ⁇ b> 10 the doctor's input “the boundary portion is clear and smooth and the shape is irregular. Suspected papillary duct cancer” is sent to the similar case retrieval apparatus 1.
- step S11 Keyword extraction
- the keyword extraction unit 3 receives the text data of the interpretation report, extracts a character string related to diagnosis from the text data, and transmits it to the difficulty level evaluation unit 4. Details of step S11 will be described below with reference to FIG.
- step S30 the keyword extraction unit 3 reads the interpretation report received from the input unit 2.
- step S31 the character string analysis unit 31 performs character string analysis on the interpretation report received from the input unit 2, and classifies nouns, particles, and the like. Thereafter, the character string analysis unit 31 sends the analyzed character string to the character string comparison unit 32.
- step S32 the character string comparison unit 32 receives the analyzed character string from the character string analysis unit 31 and compares it with the character string stored in the keyword dictionary storage unit 7.
- step S ⁇ b> 33 the character string comparison unit 32 sets only the character string that matches the character string received from the character string analysis unit 31 among the character strings stored in the keyword dictionary storage unit 7 to the difficulty level evaluation unit 4. Send it.
- step S10 in the case where the input “The boundary portion is clear and smooth and the shape is irregular. Suspected papillary ductal carcinoma” is received, in step S11, the character string analysis unit 31 “ The boundary is clear and smooth, and the shape is irregular. "
- the character string comparison unit 32 reads the character string such as “border part / has / clear smooth //, / shape / has / irregularity / .papillary duct cancer / of / suspect /.”. Disassemble.
- step S 12 difficulty assessment
- the difficulty level evaluation unit 4 receives the character string extracted from the keyword extraction unit 3 and receives a diagnosis tree from the diagnosis tree storage unit 8.
- the difficulty level evaluation unit 4 calculates the difficulty level of the diagnosis item from these character strings and the diagnosis tree, and transmits it to the similar case search unit 5. Details of step S12 will be described below with reference to FIG.
- step S40 the difficulty level evaluation unit 4 reads the character string received from the keyword extraction unit 3.
- step S41 the diagnosis tree analysis unit 41 compares the character string received from the keyword extraction unit 3 with the diagnosis tree stored in the diagnosis tree storage unit 8, and selects a diagnosis tree. That is, the diagnosis tree analysis unit 41 refers to the diagnosis tree storage unit 8 and specifies a diagnosis tree including diagnosis items, symptoms, and disease names that match the keyword extracted by the keyword extraction unit 3.
- step S42 the diagnostic tree analysis unit 41 analyzes which diagnostic flow corresponds to the current diagnostic report among a plurality of diagnostic flows included in the diagnostic tree from the received character string and the selected diagnostic tree. Then, the diagnosis tree analysis unit 41 sends the diagnosis tree and position information in the tree in the current interpretation report to the similar diagnosis flow extraction unit 42. That is, the diagnosis tree analysis unit 41 sends the diagnosis flow corresponding to the interpretation report to the similar diagnosis flow extraction unit 42 as the target diagnosis flow.
- step S43 the similar diagnosis flow extraction unit 42 receives the diagnosis tree and the position information in the tree, and calculates the degree of difficulty for the current interpretation report for each diagnosis item. For example, consider the case where “boundary part”, “clear smooth”, “shape”, “irregular”, and “papillary duct cancer” are extracted as keywords in step S11. At this time, the diagnosis tree analysis unit 41 compares the character string of the diagnosis tree in FIG. 3 with “border”, “clear smooth”, “shape”, “irregular”, and “papillary duct cancer”. The target diagnosis flow corresponding to the currently input interpretation report is extracted. Hereinafter, the evaluation of the difficulty level will be described using the target diagnosis flow extracted in this way.
- step S43 the similar diagnosis flow extraction unit 42 calculates the difficulty level used when extracting the similar diagnosis flow.
- the difficulty level may be determined from the number of matching disease names (number of identical disease names) with respect to the diagnosis item. Even if all diagnosis items are diagnosed, diagnosis is difficult if the disease name is not determined. Therefore, the similar diagnosis flow extraction unit 42 evaluates the degree of difficulty based on how much the disease names under the branch of the diagnosis item match. Specifically, the similar diagnosis flow extraction unit 42 evaluates the difficulty level of the disease name so that the difficulty level is high when the number of the same disease names is large, and the difficulty level of the disease name is low so that the difficulty level is low when the number is the same. Assess degree. That is, the similar diagnosis flow extraction unit 42 calculates the difficulty level of the disease name so that the difficulty level increases as the number of diagnosis flows including the same disease name increases. Details of step S43 will be described below with reference to FIG.
- step S70 the similar diagnosis flow extraction unit 42 reads the diagnosis tree.
- step S71 the similar diagnosis flow extraction unit 42 calculates a search target flow.
- step S72 the similar diagnosis flow extraction unit 42 stores a disease name of a search target diagnosis flow (hereinafter also referred to as a search target flow).
- step S73 the similar diagnosis flow extraction unit 42 counts the number of item branches of the read diagnosis tree.
- step S74 the similar diagnosis flow extraction unit 42 reads an item branch.
- step S75 the similar diagnosis flow extraction unit 42 counts the same number of disease names as the disease names stored in step S72 existing under the branch.
- step S76 the similar diagnosis flow extraction unit 42 calculates the difficulty level of the diagnosis item from the same number of disease names.
- Step S77 is a branching step for confirming whether or not the number of disease names has been counted for all items. If all the items have been counted, the process proceeds to the end. If not, the process proceeds to step S74. Continue to count the same disease name.
- the degree of difficulty is evaluated for “irregular”, the disease name “solid ductal cancer” that also exists under “irregular” is included under “polygon”, so these symptoms are mistaken. Since it is easy, it is evaluated that the degree of difficulty is high. On the other hand, since the “branch shape” is not easily mistaken because there is no matching disease name, it is evaluated that the difficulty level is low.
- the difficulty level may be determined from the number of types of symptoms (number of item branches) for the diagnostic item. Diagnosis is difficult when there are many types of symptoms for the diagnosis item. Therefore, the degree of difficulty is evaluated based on the number of symptoms. At this time, it is evaluated that the difficulty level is high when there are many types of symptoms, and the difficulty level is low when there are few types of symptoms. That is, the difficulty level of a diagnostic item is calculated so that the difficulty level increases as the number of symptom branches of the diagnostic item in the diagnostic tree increases. In other words, the difficulty level of the diagnostic item is calculated so that the difficulty level increases as the number of symptoms that can be determined as symptoms of the diagnostic item increases. Details of step S43 will be described below with reference to FIG.
- step S80 the similar diagnosis flow extraction unit 42 reads the diagnosis tree.
- step S81 the similar diagnosis flow extraction unit 42 calculates a search target flow.
- step S82 the similar diagnosis flow extraction unit 42 counts the number of item branches in the diagnosis tree.
- step S83 the similar diagnosis flow extraction unit 42 counts the number of item branches.
- step S84 the similar diagnosis flow extraction unit 42 calculates the difficulty level of the diagnosis item from the number of item branches.
- Step S85 is a branch for confirming whether or not the number of diagnostic items has been counted for all item branches. If all items have been counted, the process proceeds to the end. If not, the process proceeds to step S83. Continue counting forward item branches.
- the difficulty level is calculated from the number of item branches in Fig. 3.
- each symptom type has the same diagnosis difficulty level, and the difficulty level is determined only from the number.
- the diagnosis item there are four types of symptoms under the diagnosis item “shape”: “irregular”, “polygon”, “leaf shape”, and “circle / ellipse”.
- any of “Irregular”, “Polygon”, “Foliage”, and “Circle / Ellipse” It is evaluated that the degree of difficulty is higher.
- the difficulty level of the disease name may be calculated in the same manner as described above. That is, the difficulty level of the disease name may be calculated such that the difficulty level increases as the number of disease names that branch from the symptoms of the diagnosis item in the diagnosis tree increases. Specifically, for example, in FIG. 3, it may be calculated such that the more difficult the disease name is, the more disease names are listed in the lowermost block.
- the difficulty level may be determined based on the ease with which a disease name for a diagnostic item is mistaken. It is assumed that the diagnosis was difficult when there were many wrong names in the diagnosis items. Therefore, the difficulty level is evaluated based on the ease of mistake of the disease name. Specifically, the similar diagnosis flow extraction unit 42 calculates the difficulty level to be higher as the value that is predetermined for each disease name and that indicates the ease of mistake of the disease name is larger.
- step S43 Details of step S43 are shown in FIG.
- step S90 the similar diagnosis flow extraction unit 42 reads the diagnosis tree.
- step S91 the similar diagnosis flow extraction unit 42 calculates a search target flow.
- step S92 the similar diagnosis flow extraction unit 42 extracts a disease name to be searched.
- step S93 the similar diagnosis flow extraction unit 42 reads the disease data for the disease name extracted in step S93 and the disease name in the diagnosis tree.
- step S94 the similar diagnosis flow extraction unit 42 counts the total number of diagnosis flows.
- step S95 the similar diagnosis flow extraction unit 42 reads the diagnosis flow.
- step S96 the similar diagnosis flow extraction unit 42 evaluates the difficulty level of the disease name in each diagnosis flow based on the disease data.
- Step S97 is a branching step for confirming whether the difficulty level has been evaluated for all the diagnostic flows. If the evaluation of all the flows has been completed, the process proceeds to the end. If not, the process proceeds to step S95. Continue to evaluate the difficulty of progress.
- the ease of diagnosing the disease name may be calculated from the accuracy rate of the pathological examination for the diagnosis result.
- the pathological examination is an examination for collecting mass tissue.
- a pathological examination is performed to collect a mass tissue at the stage of treatment, and as a result of analyzing the tissue, a definitive diagnosis is made.
- the coincidence rate between the pathological examination result and the interpretation result is defined as a correct answer rate.
- the difficulty level may be calculated by combining the plurality of methods described above.
- the difficulty level is calculated based on the ease of mistake of the disease name in FIG.
- the symptom type “polygon” there are three types of disease names “solid ductal carcinoma”, “mucinous carcinoma”, and “medullary carcinoma”.
- step S44 the similar diagnosis flow extraction unit 42 uses the plurality of diagnosis trees extracted by the diagnosis tree analysis unit 41, evaluates the combination of the diagnosis trees, deletes an impossible similar diagnosis flow, and enables an effective diagnosis flow. Is extracted and sent to the similar case search unit 5.
- Extraction of effective diagnosis flows from a plurality of diagnosis trees is performed using as an index whether or not a combination of disease names exists in the plurality of diagnosis trees. Details of step S44 will be described below with reference to FIG.
- step S100 the similar diagnosis flow extraction unit 42 reads a plurality of diagnosis trees.
- step S101 the similar diagnosis flow extraction unit 42 extracts a flow having the same disease name as the search target flow from a plurality of diagnosis trees.
- step S102 the similar diagnosis flow extraction unit 42 reads flows of similar difficulty levels from a plurality of diagnosis trees for flows having the same disease name.
- step S103 the similar diagnosis flow extraction unit 42 counts the number of types of disease names in the flows extracted in steps S101 and S102.
- step S104 the similar diagnosis flow extraction unit 42 counts the number of disease names included in the flow for each disease name and each diagnosis tree.
- step S105 the similar diagnosis flow extraction unit 42 adopts the disease names counted in all diagnosis trees as an effective diagnosis flow.
- Step S106 is a branching step for confirming whether all disease names have been evaluated. If evaluation has been completed for all disease names, the process proceeds to the end. If not, the process proceeds to step S103 for evaluation. to continue.
- step S44 the diagnostic tree to be used may be changed from the diagnostic items described in the interpretation report.
- the diagnostic tree is changed based on the effective diagnostic flow extraction method described above will be described.
- the detailed flow for changing the diagnosis tree in step S44 will be described below with reference to FIG.
- step S110 the similar diagnosis flow extraction unit 42 reads the diagnosis item and the disease name described in the interpretation report.
- step S111 the similar diagnosis flow extraction unit 42 selects and reads a diagnosis tree including the diagnosis item extracted in step S110.
- step S112 the similar diagnosis flow extraction unit 42 extracts a diagnosis flow having the same disease name as the search target flow from a plurality of diagnosis trees.
- step S113 the similar diagnosis flow extraction unit 42 reads similar diagnosis flows from a plurality of diagnosis trees for flows having the same disease name.
- step S114 the similar diagnosis flow extraction unit 42 counts the number of types of disease names in the flows extracted in steps S112 and S113.
- step S115 the similar diagnosis flow extraction unit 42 counts the number of disease names included in the flow for each disease name and each diagnosis tree.
- step S116 the similar diagnosis flow extraction unit 42 adopts the disease names counted in all diagnosis trees as an effective diagnosis flow.
- Step S117 is a branching step for confirming whether all disease names have been evaluated. If evaluation has been completed for all disease names, the process proceeds to the end. If not, the process proceeds to step S114 for evaluation. to continue. For example, in the case where there are a plurality of diagnosis trees as shown in FIG. 22, the doctor now makes a diagnosis regarding the shape, and the search target flow is “item a-pattern A-item b-pattern C-disease name 2”. . It is also assumed that diagnosis of disease names 1, 2, and 3 is difficult in the shape diagnosis tree, and diagnosis of disease names 2 and 3 is difficult in the shape diagnosis tree.
- step S111 the currently diagnosed diagnosis flow is read, and in step S112, the diagnosis flow (in this case, from the color diagnosis tree) whose disease name matches the currently diagnosed flow is referred to as the diagnosis flow “item c-pattern E”.
- the diagnosis flow “item c-pattern E”.
- step S112 the diagnosis flow (in this case, from the color diagnosis tree) whose disease name matches the currently diagnosed flow is referred to as the diagnosis flow “item c-pattern E”.
- step S112 the diagnosis flow (in this case, from the color diagnosis tree) whose disease name matches the currently diagnosed flow is referred to as the diagnosis flow “item c-pattern E”.
- -"Item d-Pattern G-Disease name 2" is read.
- step S113 “item a—pattern A—item b—pattern C—disease name 1”, “item a—pattern A—item b—pattern C—disease name 3”, “item c” are similar
- Diagnosis flow 1 “Item a—Pattern A—Item b—Pattern C—Disease name 1” (Diagnostic tree: shape)
- Diagnosis flow 2 “Item a—Pattern A—Item b—Pattern C—Disease name 2” (Diagnostic tree: shape)
- Diagnosis flow 3 “Item a—Pattern A—Item b—Pattern C—Disease name 3” (Diagnostic tree: shape)
- Diagnosis flow 4 “item c-pattern E-item d-pattern G-disease name 2” (diagnosis tree: color)
- Diagnosis flow 5 “item c-pattern E-item d-pattern G-disease name 3” (diagnosis tree: color)
- the similar diagnosis flow extraction unit 42 extracts a similar diagnosis flow for each diagnosis tree.
- steps S114 and S115 the number of disease names is counted with respect to the shape tree and the color tree.
- the diagnosis flow currently being searched is not similar to the diagnosis flow that is comprehensively diagnosed using another diagnosis flow, and is therefore excluded from the display target. Can be changed.
- display targets can be narrowed down by evaluation combining a plurality of trees.
- step S ⁇ b> 45 the similar diagnosis flow extraction unit 42 sends a diagnosis flow similar in degree of difficulty to the search target to the similar case search unit 5 as a similar diagnosis flow based on the difficulty level. Specifically, the similar diagnosis flow extraction unit 42 extracts a similar diagnosis flow so that a diagnosis flow including a diagnosis item or disease name having a higher difficulty level is extracted as a similar diagnosis flow.
- step S13 Similar case search
- the similar case search unit 5 receives the similar diagnosis flow from the difficulty level evaluation unit 4, calculates the similarity of each case data stored in the case data storage unit 9, and stores case data having a high similarity. Priority is sent to the similar case display unit 6. Details of step S13 will be described below with reference to FIG.
- step S50 the similar case search unit 5 reads the similar diagnosis flow received from the difficulty level evaluation unit 4.
- step S51 the search case control unit 51 receives a search item from the input unit 2, and transmits a similarity diagnosis flow weighted to the search item to the similarity evaluation unit 52. Details of step S51 are shown in FIG.
- step S60 the search case control unit 51 receives a similar diagnosis flow.
- step S61 the search case control unit 51 reads the user input from the input unit 2.
- Step S62 is a branching step for determining whether or not there is a user input in the input unit 2. If there is an input, the process proceeds to step S63, and if there is no input, the process proceeds to step S52.
- step S63 the search case control unit 51 searches the presence / absence and position of the character string input by the user in the similar diagnosis flow.
- step S64 the search case control unit 51 weights the items in the diagnosis flow input by the user. By applying this weighting, it is possible to preferentially search items that the user wants to search. Details of the calculation method for weighting will be described in the examples described later.
- step S52 the similarity evaluation unit 52 receives the user search item and the similarity diagnosis flow, and evaluates the similarity with the interpretation report stored in the case data storage unit 9.
- step S53 the similarity evaluation unit 52 reads from the case data storage unit 9 an interpretation report and a medical image that are included in the case data having a higher similarity evaluation result in step S52.
- step S54 the similarity evaluation unit 52 sends the interpretation report and the medical image included in the case data read from the case data storage unit 9 to the similar case display unit 6.
- step S14 the similar case display unit 6 receives the case data from the similar case search unit 5 and presents it to the doctor.
- the similar case display unit 6 is, for example, a display, a TV, a monitor, or the like.
- case data can be searched using the similar diagnosis flow extracted based on the degree of difficulty. Therefore, since case data can be searched effectively by paying attention to a diagnostic item or disease name having a high degree of difficulty, it is possible to search for appropriate case data.
- FIG. 25 shows the diagnostic tree used in this example. Symptoms are described as a pattern in a square frame, and diagnosis items are described as items. A diagnosis tree having the same disease name in a plurality of diagnosis flows (eg, disease name N in patterns D and E) was used. In this experiment, it is assumed that the interpretation report includes “item a is pattern B. item c is pattern G. Suspected disease name S” and the following keywords are extracted.
- x indicates the comparison result of the character strings, and takes 1 when they match and 0 when they do not match.
- F k (w, x) is a function for determining whether to perform weighting. 0 is taken when both x and w are 0, and 1 is taken otherwise.
- diagnosis item 1 is a hierarchy including diagnosis item a
- symptom 1 is a symptom for item a
- patterns A, B, and C correspond to this.
- Diagnostic item 2 is a diagnostic item under the symptom and refers to items b, c, and d.
- Symptom 2 indicates symptoms under items b, c, and d.
- diagnosis flow is expressed as “diagnosis item 1—symptom 1—diagnosis item 2—symptom 2—disease name”.
- the diagnostic flow of the interpretation report to be searched is “aBCcGS”.
- FIG. 27 shows experimental results of difficulty level evaluation based on the number of matching disease names.
- the degree of difficulty is evaluated using the number of matching disease names as an index
- the number of disease names that match the disease name “disease name S” in the search interpretation report is calculated in each diagnosis flow, and thus the similarity of the diagnosis flow including the disease name S increases.
- the similarities of “aBCcHS” and “aBCcIS” increased from 0.8 to 0.9. Therefore, if this technique is used, it is possible to preferentially search for a diagnosis flow that reaches the name of the disease being searched. That is, a diagnosis flow including “disease name S” which is a disease name having a high difficulty level is extracted as a similar diagnosis float.
- FIG. 28 shows experimental results of difficulty level assessment based on ease of mistaking a disease name.
- the degree of difficulty was evaluated using the ease of mistake of the disease name as an index, the similarity of the diagnosis flow including “disease name T” with the same diagnosis item and symptom as the diagnosis flow including the disease name “disease name S” in the search interpretation report increased. Therefore, if this method is used, it is possible to preferentially search for a diagnosis flow including a disease name that is easily mistaken. That is, a diagnosis flow including a disease name that is easily mistaken for a disease name included in the target diagnosis flow is extracted as a similar diagnosis flow.
- FIG. 29 shows experimental results of difficulty level evaluation based on a combination of difficulty level (2) and difficulty level (3).
- aBCcHT has a degree of similarity of 0.6 in the conventional number of matching character strings, but the degree of similarity is 0.9 in the method proposed this time. Therefore, if this technique is used, it is possible to preferentially search for a diagnosis flow including a disease name that is easy to reach the name of the disease being searched and is easy to make a mistake.
- the keyword extraction unit described above extracts items such as “border”, “clear smooth”, “shape”, “polygon”, and “solid duct cancer” from the report description, and within the diagnosis tree In this case, the diagnosis flow “border-clear smooth-shape-polygon-solid duct cancer” that matches this description is selected as a search target.
- the past case 1 is “boundary”, “clear smooth”, “shape”, “ While the four character strings of “square” match, the past case 1 matches three of “boundary”, “clear smooth”, and “shape”, so the past case 1 with many matching character strings Are output as similar cases.
- the input unit 2 may receive input from a doctor using a template for selecting diagnostic items and symptoms prepared in advance.
- the input unit 2 may create the template from the data in the case data storage unit 9.
- the character string analysis unit 31 may use a general keyword extraction method (for example, keyword extraction by morphological analysis or keyword extraction by N-gram).
- the character string comparison unit 32 may compare with a keyword using a synonym dictionary.
- the character string comparison unit 32 performs not only a comparison with the keyword stored in the keyword dictionary storage unit 7 but also a conversion process for unifying the keywords for difficulty evaluation using the synonym dictionary. Also good.
- the diagnostic tree corresponding to the input interpretation report may be determined not only from the diagnostic items but also from the examination site information (chest, abdomen, etc.) of the patient.
- the keyword stored in the keyword dictionary storage unit 7 may be created with reference to an interpretation report stored in the case data storage unit.
- storage part 7 may be produced from disease classification tables, such as ICD10 (international disease classification 10th edition), for example.
- the diagnostic tree storage unit 8 may be created with reference to an interpretation report in the case DB.
- the current search case may be added to the case data storage unit 9 after the diagnosis is completed.
- the similarity diagnosis flow extraction unit 42 calculates the difficulty level, but it is not always necessary to calculate the difficulty level.
- the similar diagnosis flow extraction unit 42 may read the difficulty stored for each diagnosis item or disease name from the diagnosis tree storage unit 8.
- some or all of the components included in the similar case search apparatus 1 in the above embodiment may be configured by one system LSI (Large Scale Integration).
- the system LSI is an ultra-multifunctional LSI manufactured by integrating a plurality of components on a single chip. Specifically, a microprocessor, a ROM (Read Only Memory), a RAM (Random Access Memory), etc. It is a computer system comprised including. A computer program is stored in the ROM. The system LSI achieves its functions by the microprocessor operating according to the computer program.
- system LSI may be called IC, LSI, super LSI, or ultra LSI depending on the degree of integration.
- method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible.
- An FPGA Field Programmable Gate Array
- reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may be used.
- the present invention can be realized not only as a similar case search apparatus including such a characteristic processing unit, but also as a similar case search method using a characteristic processing unit included in the similar case search apparatus as a step. It can also be realized. It can also be realized as a computer program that causes a computer to execute the characteristic steps included in the similar case retrieval method. Needless to say, such a computer program can be distributed via a computer-readable recording medium such as a CD-ROM or a communication network such as the Internet.
- the present invention is not limited to the above embodiment.
- the present invention is useful as a similar case retrieval apparatus for retrieving a plurality of case data (diagnostic images and interpretation reports) stored in a case data storage unit. It is also widely used in other systems that search similar data from databases that store text data in association with images and drawings in fields where the structure of judgments can be made into a tree (mechanism design, trial case search, patent search, etc.). Applicable.
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Entrepreneurship & Innovation (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
Description
ある診断項目の症状として判定可能な症状同士がよく似ていることを評価するためには、その診断項目を判定した症状の先に同じ病名がどれだけ存在するかを見ればよい。例えば、図4のように診断項目aを判定した後に、パターンAとパターンBとに同じ病名が多く存在する場合には、その症状はそれら病名を区別する要素にはなっていないと考えられる。したがって、症状同士には対象とする病名を大きく分けるような特徴がない、つまり、判定する症状同士が似ているため、その症状間の判定を行うのは難しくなると考えられる。そのため、診断ツリーからある診断項目下の病名数を計数することで難易度を評価する。
診断ツリーから症状の種類が多いことを評価するためには、診断ツリー中の症状の種類数(診断項目に対する枝の数)を用いればよい。例えば、図3の形状という診断項目に対して、症状の個数(4個:不整形、多角形、分葉形、円・楕円)から難易度を評価する。図5のようにこの種類数が多いほど難易度が高い診断項目として評価する。
診断ツリーから複数の診断項目を診断しても、病名が決定しないことを評価するためには、診断ツリーに含まれる診断項目数を用いればよい。図6のような診断ツリーに対して、実線矢印と点線矢印のフローを比較した場合には、診断する回数が多い実線の方が診断は難しく、難易度が高いと評価する。
全ての診断項目を診断しても、病名が決定されない場合について説明する。図7のような診断ツリーがある場合に、矢印で示すように診断項目を診断したとする。このような場合、診断項目の診断を全て終えても、病名1、2、3の中からいずれかを決定しなければならない。このときには、診断ツリー中の病名数から難易度を評価すればよい。例えば、図7においては、病名1、2、3の病名数(3個)。図3においては、境界部が明瞭平滑、形状が不整形の診断ツリーに対する病名数(2個)から難易度を評価する。
1つの診断項目の種類では、病名が決定しないことを評価するためには、読影レポートから項目数を抽出し、使用した診断ツリーの個数を計算すればよい。図8のように形状に関する診断ツリーと色に関する診断ツリーがある場合について説明する。このとき、それぞれ矢印のように診断を行い病名1に至ったとする。この場合には、読影レポートにはそれぞれの診断ツリーに関する文字列(例えば「項目a」など)が含まれている。したがって、読影レポートの文字列を解析し、各診断ツリーにあてはめることで診断ツリーをいくつ使用したかが分かる。また、この診断ツリーの個数が多いほど複雑な診断をしていることになり、難易度が高いと評価する。
図9は、本実施の形態における類似症例検索システム100の構成を示すブロック図である。図9に示すように、類似症例検索システム100は、類似症例検索装置1と、入力部2と、類似症例表示部6と、読影支援データベース10とを備える。
入力部2は、医師が読影レポートを入力するための機器(例えば、キーボードあるいはマウスなど)である。入力部2は、医師が入力したテキストデータや選択結果をキーワード抽出部3に送付する。
類似症例表示部6は、類似症例検索部5から症例データを受け取り、医師に提示する。類似症例表示部6は、例えばPC(Personal Computer)用のディスプレイ、TV(Television)、医療用読影モニタなどである。
類似症例検索装置1は、入力部2で入力された読影レポートと類似する症例データを、症例データ記憶部9から検索するための装置である。類似症例検索装置1は、キーワード抽出部3と、難易度評価部4と、類似症例検索部5とを備える。
キーワード抽出部3は、入力部2から読影レポートのテキストデータを受信し、テキストデータの中から医学用語、診断項目、診断結果に関わるキーワードを抽出し、難易度評価部4に送付する。つまり、キーワード抽出部3は、入力部2から読影レポートを受信し、受信した読影レポートからキーワードを抽出する。
難易度評価部4は、キーワード抽出部3から抽出された文字列を受け取り、診断ツリー記憶部8からは診断ツリーを読み出す。難易度評価部4は、これらの文字列と診断ツリーとから診断項目の難易度を計算し、類似症例検索部5に送付する。
類似症例検索部5は、難易度評価部4から類似診断フローを受け取り、症例データ記憶部9の症例データの類似度を計算し、類似度が高い症例データを類似症例表示部6に送付する。つまり、類似症例検索部5は、症例データ記憶部9に記憶されている複数の症例データの中から、類似診断フローに対応する症例データを検索する。つまり、類似症例検索部5は、類似診断フローに従って診断された症例データを検索する。さらに、類似症例検索部5は、対象診断フローに従って診断された症例データも検索する。
読影支援データベース10は、キーワード辞書記憶部7と、診断ツリー記憶部8と、症例データ記憶部9とを備える。
キーワード辞書記憶部7は、キーワード抽出に用いる医学用語を記憶している。
診断ツリー記憶部8は、診断項目とその症状の種類などの診断フローに関する情報を格納している。つまり、診断ツリー記憶部8は、病名が決定されるまでの診断項目と当該診断項目の症状とを含む複数の診断フローがツリー構造で表された診断ツリーを記憶している。
症例データ記憶部9は、過去の読影レポートと診断に用いた医用画像を格納している。つまり、症例データ記憶部9は、読影レポートと医用画像との組である症例データを記憶している。
図13は、本実施の形態における類似症例検索システム100の利用形態を示す図である。
図14は、類似症例検索システム100において行われる処理の流れを示すフローチャートである。図14は、図1のステップS03、S07に対応する。
(S10:読影レポート入力)
ステップS10において、入力部2は、医師からの読影レポートの入力を受け付け、類似症例検索装置1に送付する。ここで、図2に示すような読影レポートに医師が「境界部は明瞭平滑であり、形状は不整形である。乳頭腺管癌の疑い。」と入力した場合を例にとり、具体的に説明する。ステップS10においては、医師の入力である「境界部は明瞭平滑であり、形状は不整形である。乳頭腺管癌の疑い。」が類似症例検索装置1に送付される。
ステップS11において、キーワード抽出部3は、読影レポートのテキストデータを受け取り、そのテキストデータから診断に関する文字列を抜き出し、難易度評価部4に送信する。以下、ステップS11の詳細を、図15を用いて説明する。
ステップS12において、難易度評価部4は、キーワード抽出部3から抽出された文字列を受け取り、診断ツリー記憶部8から診断ツリーを受け取る。難易度評価部4は、これら文字列と診断ツリーとから診断項目の難易度を計算し、類似症例検索部5に送信する。以下、ステップS12の詳細を図16を用いて説明する。
ステップS43において、類似診断フロー抽出部42は、類似診断フローを抽出する際に利用する難易度を計算する。
なお、ステップS43において、難易度は、診断項目に対する病名の一致数(同一病名数)から決定されてもよい。全ての診断項目を診断しても、病名が決定しない場合には診断が難しくなる。そのため、類似診断フロー抽出部42は、診断項目の分岐下の病名がどれだけ一致しているかを基準に難易度を評価する。具体的には、類似診断フロー抽出部42は、同一病名数が多い場合には難易度が高くなるように病名の難易度を評価し、少ない場合には難易度が低くなるように病名の難易度を評価する。つまり、類似診断フロー抽出部42は、同一の病名を含む診断フローが多いほど難易度が高くなるように、病名の難易度を計算する。以下、ステップS43の詳細を図17を用いて説明する。
なお、ステップS43において、難易度は、診断項目に対する症状の種類数(項目分岐数)から決定されてもよい。診断項目に対する症状の種類が多いときには診断が難しくなる。そのため、症状の数を基準に難易度を評価する。このとき、症状の種類が多い場合には難易度が高く、少ない場合には難易度が低いと評価する。つまり、診断項目の難易度は、診断ツリーにおける診断項目の症状の分岐数が多いほど難易度が高くなるように計算される。言い換えると、診断項目の難易度は、診断項目の症状として判定可能な症状の数が多いほど難易度が高くなるように計算される。以下、ステップS43の詳細について図18を用いて説明する。
なお、ステップS43において、難易度は診断項目に対する病名の間違えやすさから決定されてもよい。診断項目において病名の間違いが多いときには診断が難しかったことが想定される。そのため、病名の間違えやすさを基準に難易度を評価する。具体的には、類似診断フロー抽出部42は、病名ごとに予め定められた値であって病名の間違えやすさ示す値が大きいほど難易度が高くなるように計算する。
ステップS44において、類似診断フロー抽出部42は、診断ツリー解析部41が抽出した複数の診断ツリーを使用し、診断ツリーの組合せを評価することで、ありえない類似診断フローを削除し、有効な診断フローを抽出し、類似症例検索部5に送付する。
なお、ステップS44において、読影レポートに記載された診断項目から使用する診断ツリーを変更してもよい。ここでは、先に述べた有効な診断フローの抽出方法を基に診断ツリーを変更する例について説明する。以下、ステップS44において、診断ツリーを変更する詳細フローについて図21を用いて説明する。
診断フロー2:「項目a-パターンA-項目b-パターンC-病名2」(診断ツリー:形状)
診断フロー3:「項目a-パターンA-項目b-パターンC-病名3」(診断ツリー:形状)
診断フロー4:「項目c-パターンE-項目d-パターンG-病名2」(診断ツリー:色)
診断フロー5:「項目c-パターンE-項目d-パターンG-病名3」(診断ツリー:色)
ステップS45において、類似診断フロー抽出部42は、難易度に基づいて、検索対象と難易度が類似した診断フローを類似診断フローとして類似症例検索部5に送付する。具体的には、類似診断フロー抽出部42は、難易度が高い診断項目または病名が含まれる診断フローほど類似診断フローとして抽出されるように、類似診断フローを抽出する。
ステップS13において、類似症例検索部5は、難易度評価部4から類似診断フローを受け取り、症例データ記憶部9に記憶されている各症例データの類似度を計算し、類似度が高い症例データを優先して類似症例表示部6に送付する。以下、ステップS13の詳細を図23を用いて説明する。
ステップS14において、類似症例表示部6は、類似症例検索部5から症例データを受け取り、医師に提示する。類似症例表示部6は、例えばディスプレイ、TV、モニタなどである。
以下では、実験のために簡単な読影レポートを作成し、類似度を評価した結果を示す。本実施例で用いた診断ツリーを図25に示す。症状をパターンとして四角の枠内に記載し、診断項目は項目として記載した。複数の診断フローに同一の病名が存在する(例:パターンD、Eに病名Nが存在する)診断ツリーを用いた。また、今回の実験において、読影レポートには「項目aはパターンB。項目cはパターンG。病名Sの疑い。」が記載されているとし、以下のキーワードが抽出されるとした。
(2)症状:「パターンB」、「パターンG」
(3)病名:「病名S」
本実験では、結果を分かりやすくするために、診断項目下に存在する病名の一致数として病名に一番近い診断結果である図25のパターンD~Lに関して以下の式を適用した。なお、この計算は診断項目を遡ってパターンA~Cに適用してもよい。ここでnxは分岐下にある病名の一致数であり、naは病名の総数を示す。例えば、パターンDとパターンEを比較する場合には、パターンD下には病名M、N、Oがあり、パターンE下には病名P、N、Oがある。したがって、N、Oが一致しているため、nxは2、総数naは6となり、その重みwkは0.33となる。
本実験では、病名Sが一番間違えやすく、次に病名Tを間違えやすいとした。また、その間違えやすさの数値には式(4)を用いた。この数値は、病名ごとにあらかじめ決定してもよいし、データベースから自動的に決定してもよい。
実験結果を図26~図29に示す。図には類似度の高い10件について類似度と診断フローとを掲載した。ここで、診断項目1は診断項目aを含む階層、症状1は項目aに対する症状であり、パターンA、B、Cがこれに該当する。診断項目2はその症状下にある診断項目であり項目b、c、dのことをいう。症状2は項目b、c、d下の症状を示す。以下では診断フローを「診断項目1-症状1-診断項目2-症状2-病名」と表記する。この表記法では検索対象となる読影レポートの診断フローは「a-B-c-G-S」となる。
従来法の実験結果を図26に示す。文字列の一致数が類似度の指標であるため、例えば今回の検索対象読影レポート「a-B-c-G-S」に対して、類似度が一番高いものは文字列が全て一致する「a-B-c-G-S」であり、その類似度は1となる。また、5つの文字列中4つが一致する「a-B-c-G-T」の類似度は0.8となる。このように、従来法では検索した診断フローと同じ症例データを検索できる。
病名一致数による難易度評価の実験結果を図27に示す。病名一致数を指標に難易度を評価すると、検索読影レポートの病名「病名S」と一致する病名数を各診断フローにおいて計算するため、病名Sが含まれる診断フローの類似度が上昇した。例えば、図27において「a-B-c-H-S」、「a-B-c-I-S」の類似度が0.8から0.9に上昇した。そのため、本手法を使用すれば、検索している病名に至る診断フローを優先的に検索できる。つまり、難易度が高い病名である「病名S」を含む診断フローが類似診断フロートとして抽出される。
病名間違えやすさによる難易度評価の実験結果を図28に示す。病名間違えやすさを指標に難易度を評価すると、検索読影レポートの病名「病名S」が含まれる診断フローと同じ診断項目、症状で「病名T」が含まれる診断フローの類似度が上昇した。そのため、本手法を使用すれば、間違えやすい病名を含む診断フローを優先的に検索できる。つまり、対象診断フローに含まれる病名と間違えやすい病名を含む診断フローが類似診断フローとして抽出される。
難易度(2)と難易度(3)の組み合わせによる難易度評価の実験結果を図29に示す。病名一致数と病名間違えやすさを組み合わせた場合には、文字列の一致数が少なくても、類似度が大きく上昇する診断フローが存在した。例えば、「a-B-c-H-T」は従来の文字列の一致数では類似度が0.6であるが、今回提案した手法では類似度は0.9となる。したがって、本手法を使用すれば、検索している病名に至りやすく、間違えやすい病名を含む診断フローを優先的に検索できる。
2 入力部
3 キーワード抽出部
4 難易度評価部
5 類似症例検索部
6 類似症例表示部
7 キーワード辞書記憶部
8 診断ツリー記憶部
9 症例データ記憶部
10 読影支援データベース
31 文字列解析部
32 文字列比較部
41 診断ツリー解析部
42 類似診断フロー抽出部
51 検索症例制御部
52 類似度評価部
100 類似症例検索システム
Claims (9)
- 医用画像を読影する際に診断の対象となる診断項目と、前記診断項目の症状を示す診断結果とが記載された文書データである読影レポートからキーワードを抽出するキーワード抽出部と、
病名が決定されるまでの診断項目と当該診断項目の症状とを含む複数の診断フローがツリー構造で表された診断ツリーを記憶している診断ツリー記憶部を参照して、前記読影レポートに対応する診断フローである対象診断フローを前記キーワードに基づいて抽出する診断ツリー解析部と、
診断項目の症状を判定する難しさの度合いである診断項目の難易度、または病名を決定する難しさの度合いである病名の難易度に基づいて、前記診断ツリー記憶部に記憶されている診断ツリーに含まれる複数の診断フローの中から前記対象診断フローに類似する類似診断フローを抽出する類似診断フロー抽出部と、
症例データ記憶部に記憶されている複数の症例データの中から、前記類似診断フローに対応する症例データを検索する類似症例検索部とを備える
類似症例検索装置。 - 前記病名の難易度は、前記診断ツリーにおいて、同一の病名を含む診断フローの数が多いほど難易度が高くなるように計算される
請求項1に記載の類似症例検索装置。 - 前記診断項目の難易度は、前記診断ツリーにおける診断項目の症状の分岐数が多いほど難易度が高くなるように計算される
請求項1または2に記載の類似症例検索装置。 - 前記病名の難易度は、前記診断ツリーにおける診断項目の症状から分岐する病名の数が多いほど難易度が高くなるように計算される
請求項1~3のいずれか1項に記載の類似症例検索装置。 - 前記病名の難易度は、病名ごとに予め定められた値であって病名の間違えやすさを示す値が大きいほど難易度が高くなるように計算される
請求項1~4のいずれか1項に記載の類似症例検索装置。 - 少なくとも2つの前記対象診断フローが抽出され、かつ、前記2つの対象診断フローが互いに異なる診断ツリーに含まれる場合に、前記類似診断フロー抽出部は、診断ツリーごとに前記類似診断フローを抽出する
請求項1~5のいずれか1項に記載の類似症例検索装置。 - 前記類似診断フロー抽出部は、難易度が高い診断項目または病名を含む診断フローほど類似診断フローとして抽出されるように、前記診断ツリー記憶部に記憶されている診断ツリーに含まれる複数の診断フローの中から類似診断フローを抽出する
請求項1~6のいずれか1項に記載の類似症例検索装置。 - 医用画像を読影する際に診断の対象となる診断項目と、前記診断項目の症状を示す診断結果とが記載された文書データである読影レポートからキーワードを抽出するキーワード抽出ステップと、
病名が決定されるまでの診断項目と当該診断項目の症状とを含む複数の診断フローがツリー構造で表された診断ツリーを記憶している診断ツリー記憶部を参照して、前記読影レポートに対応する診断フローである対象診断フローを前記キーワードに基づいて抽出する診断ツリー解析ステップと、
診断項目の症状を判定する難しさの度合いである診断項目の難易度、または病名を決定する難しさの度合いである病名の難易度に基づいて、前記診断ツリー記憶部に記憶されている診断ツリーに含まれる複数の診断フローの中から前記対象診断フローに類似する類似診断フローを抽出する類似診断フロー抽出ステップと、
症例データ記憶部に記憶されている複数の症例データの中から、前記類似診断フローに対応する症例データを検索する類似症例検索ステップとを含む
類似症例検索方法。 - 請求項8に記載の類似症例検索方法をコンピュータに実行させるためのプログラム。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012800019350A CN102985924A (zh) | 2011-02-14 | 2012-02-09 | 相似病例检索装置以及相似病例检索方法 |
JP2012529034A JP5139603B2 (ja) | 2011-02-14 | 2012-02-09 | 類似症例検索装置および類似症例検索方法 |
US13/657,240 US9111027B2 (en) | 2011-02-14 | 2012-10-22 | Similar case search apparatus and similar case search method |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2011029169 | 2011-02-14 | ||
JP2011-029169 | 2011-02-14 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US13/657,240 Continuation US9111027B2 (en) | 2011-02-14 | 2012-10-22 | Similar case search apparatus and similar case search method |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012111288A1 true WO2012111288A1 (ja) | 2012-08-23 |
Family
ID=46672235
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2012/000885 WO2012111288A1 (ja) | 2011-02-14 | 2012-02-09 | 類似症例検索装置および類似症例検索方法 |
Country Status (4)
Country | Link |
---|---|
US (1) | US9111027B2 (ja) |
JP (1) | JP5139603B2 (ja) |
CN (1) | CN102985924A (ja) |
WO (1) | WO2012111288A1 (ja) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2015191561A (ja) * | 2014-03-28 | 2015-11-02 | 株式会社東芝 | 読影レポート作成装置、読影レポート作成システムおよび読影レポート作成プログラム |
JP2017515574A (ja) * | 2014-05-12 | 2017-06-15 | コーニンクレッカ フィリップス エヌ ヴェKonink | 症例の難しさに基づくコンピュータ支援患者層別化のための方法及びシステム |
CN109243599A (zh) * | 2018-03-16 | 2019-01-18 | 申朴信息技术(上海)股份有限公司 | 一种基于多维度信息检索的疾病对码方法 |
JP2020013535A (ja) * | 2018-07-06 | 2020-01-23 | 株式会社日立システムズ | 情報処理装置、検査評価システムおよび検査評価方法 |
Families Citing this family (42)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2973348A4 (en) | 2013-03-15 | 2017-01-25 | Rush University Medical Center | Geographic utilization of artificial intelligence in real-time for disease identification and alert notification |
CN104036109A (zh) * | 2014-03-14 | 2014-09-10 | 上海大图医疗科技有限公司 | 基于图像的病例检索、勾画及治疗计划系统和方法 |
KR102251245B1 (ko) | 2014-04-30 | 2021-05-12 | 삼성전자주식회사 | 관심 영역별 추가 정보 제공 장치 및 방법 |
EP3262533A1 (en) * | 2015-02-25 | 2018-01-03 | Koninklijke Philips N.V. | Method and system for context-sensitive assessment of clinical findings |
US10348810B1 (en) | 2015-04-06 | 2019-07-09 | EMC IP Holding Company LLC | Scalable distributed computations utilizing multiple distinct clouds |
US10366111B1 (en) | 2015-04-06 | 2019-07-30 | EMC IP Holding Company LLC | Scalable distributed computations utilizing multiple distinct computational frameworks |
US10706970B1 (en) | 2015-04-06 | 2020-07-07 | EMC IP Holding Company LLC | Distributed data analytics |
US10277668B1 (en) | 2015-04-06 | 2019-04-30 | EMC IP Holding Company LLC | Beacon-based distributed data processing platform |
US10541936B1 (en) | 2015-04-06 | 2020-01-21 | EMC IP Holding Company LLC | Method and system for distributed analysis |
US10791063B1 (en) | 2015-04-06 | 2020-09-29 | EMC IP Holding Company LLC | Scalable edge computing using devices with limited resources |
US10404787B1 (en) | 2015-04-06 | 2019-09-03 | EMC IP Holding Company LLC | Scalable distributed data streaming computations across multiple data processing clusters |
US10425350B1 (en) | 2015-04-06 | 2019-09-24 | EMC IP Holding Company LLC | Distributed catalog service for data processing platform |
US10331380B1 (en) | 2015-04-06 | 2019-06-25 | EMC IP Holding Company LLC | Scalable distributed in-memory computation utilizing batch mode extensions |
US10528875B1 (en) | 2015-04-06 | 2020-01-07 | EMC IP Holding Company LLC | Methods and apparatus implementing data model for disease monitoring, characterization and investigation |
US10812341B1 (en) | 2015-04-06 | 2020-10-20 | EMC IP Holding Company LLC | Scalable recursive computation across distributed data processing nodes |
US10496926B2 (en) | 2015-04-06 | 2019-12-03 | EMC IP Holding Company LLC | Analytics platform for scalable distributed computations |
US10515097B2 (en) | 2015-04-06 | 2019-12-24 | EMC IP Holding Company LLC | Analytics platform for scalable distributed computations |
US10541938B1 (en) | 2015-04-06 | 2020-01-21 | EMC IP Holding Company LLC | Integration of distributed data processing platform with one or more distinct supporting platforms |
US10511659B1 (en) | 2015-04-06 | 2019-12-17 | EMC IP Holding Company LLC | Global benchmarking and statistical analysis at scale |
US10509684B2 (en) | 2015-04-06 | 2019-12-17 | EMC IP Holding Company LLC | Blockchain integration for scalable distributed computations |
US10860622B1 (en) | 2015-04-06 | 2020-12-08 | EMC IP Holding Company LLC | Scalable recursive computation for pattern identification across distributed data processing nodes |
US10776404B2 (en) | 2015-04-06 | 2020-09-15 | EMC IP Holding Company LLC | Scalable distributed computations utilizing multiple distinct computational frameworks |
US10505863B1 (en) | 2015-04-06 | 2019-12-10 | EMC IP Holding Company LLC | Multi-framework distributed computation |
CN106169165B (zh) * | 2015-05-20 | 2020-06-16 | 詹宇 | 面向诊疗数据的症状层次关联及预测方法 |
CN104866979B (zh) * | 2015-06-08 | 2018-10-23 | 苏芮 | 一种突发急性传染病的中医病例数据处理方法和系统 |
JP6697743B2 (ja) * | 2015-09-29 | 2020-05-27 | パナソニックIpマネジメント株式会社 | 情報端末の制御方法及びプログラム |
CN106560827B (zh) * | 2015-09-30 | 2021-11-26 | 松下知识产权经营株式会社 | 控制方法 |
CN105224807B (zh) * | 2015-10-14 | 2018-09-14 | 东软集团股份有限公司 | 病例审核规则提取方法及装置,病例审核方法及系统 |
US10656861B1 (en) | 2015-12-29 | 2020-05-19 | EMC IP Holding Company LLC | Scalable distributed in-memory computation |
CN107958007B (zh) * | 2016-10-18 | 2022-03-29 | 浙江格林蓝德信息技术有限公司 | 病例信息检索方法及装置 |
US10374968B1 (en) | 2016-12-30 | 2019-08-06 | EMC IP Holding Company LLC | Data-driven automation mechanism for analytics workload distribution |
JP6862286B2 (ja) * | 2017-05-31 | 2021-04-21 | キヤノン株式会社 | 情報処理装置、情報処理方法、情報処理システム及びプログラム |
CN109002442B (zh) * | 2017-06-06 | 2023-04-25 | 株式会社日立制作所 | 一种基于医生相关属性检索诊断病例的装置及方法 |
CN109002449B (zh) * | 2017-06-06 | 2023-04-07 | 株式会社日立制作所 | 一种基于疾病相关属性检索诊断病例的装置及方法 |
CN107658004A (zh) * | 2017-09-14 | 2018-02-02 | 武汉互创科技有限公司 | 一种在移动终端上查看云端医学影像信息的方法及系统 |
CN110021424B (zh) * | 2017-10-13 | 2021-11-16 | 广州金域医学检验集团股份有限公司 | 病例提示反射方法及装置 |
US11139080B2 (en) | 2017-12-20 | 2021-10-05 | OrthoScience, Inc. | System for decision management |
CN109346171A (zh) * | 2018-10-31 | 2019-02-15 | 北京惠每云科技有限公司 | 一种辅助诊断方法、装置和计算机设备 |
CN109793491B (zh) * | 2018-12-29 | 2021-11-23 | 维沃移动通信有限公司 | 一种用于色盲检测的终端设备 |
CN109887604B (zh) * | 2019-02-26 | 2023-01-17 | 上海中医药大学 | 一种中医病名相似度的量化判定系统 |
CN113094497B (zh) * | 2021-06-07 | 2021-09-14 | 华中科技大学 | 一种电子健康记录推荐方法及共享边缘计算平台 |
CN113793663A (zh) * | 2021-08-17 | 2021-12-14 | 陈浩宇 | 一种医疗数据处理方法和系统 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004157623A (ja) * | 2002-11-05 | 2004-06-03 | Hitachi Ltd | 検索システム及び検索方法 |
JP2009129108A (ja) * | 2007-11-21 | 2009-06-11 | Toshiba Corp | レポート検索装置、及びレポート検索方法 |
JP2010079568A (ja) * | 2008-09-25 | 2010-04-08 | Canon Inc | データ検索装置およびその制御方法、データ検索システム |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009078085A (ja) | 2007-09-27 | 2009-04-16 | Fujifilm Corp | 医用画像処理システム、医用画像処理方法、及びプログラム |
JP5383431B2 (ja) | 2009-10-27 | 2014-01-08 | キヤノン株式会社 | 情報処理装置、情報処理方法及びプログラム |
-
2012
- 2012-02-09 WO PCT/JP2012/000885 patent/WO2012111288A1/ja active Application Filing
- 2012-02-09 CN CN2012800019350A patent/CN102985924A/zh active Pending
- 2012-02-09 JP JP2012529034A patent/JP5139603B2/ja active Active
- 2012-10-22 US US13/657,240 patent/US9111027B2/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004157623A (ja) * | 2002-11-05 | 2004-06-03 | Hitachi Ltd | 検索システム及び検索方法 |
JP2009129108A (ja) * | 2007-11-21 | 2009-06-11 | Toshiba Corp | レポート検索装置、及びレポート検索方法 |
JP2010079568A (ja) * | 2008-09-25 | 2010-04-08 | Canon Inc | データ検索装置およびその制御方法、データ検索システム |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2015191561A (ja) * | 2014-03-28 | 2015-11-02 | 株式会社東芝 | 読影レポート作成装置、読影レポート作成システムおよび読影レポート作成プログラム |
JP2017515574A (ja) * | 2014-05-12 | 2017-06-15 | コーニンクレッカ フィリップス エヌ ヴェKonink | 症例の難しさに基づくコンピュータ支援患者層別化のための方法及びシステム |
CN109243599A (zh) * | 2018-03-16 | 2019-01-18 | 申朴信息技术(上海)股份有限公司 | 一种基于多维度信息检索的疾病对码方法 |
JP2020013535A (ja) * | 2018-07-06 | 2020-01-23 | 株式会社日立システムズ | 情報処理装置、検査評価システムおよび検査評価方法 |
JP7229761B2 (ja) | 2018-07-06 | 2023-02-28 | 株式会社日立システムズ | 情報処理装置、検査評価システムおよび検査評価方法 |
JP7493638B2 (ja) | 2018-07-06 | 2024-05-31 | 株式会社日立システムズ | 情報処理装置、検査評価システムおよび検査評価方法 |
Also Published As
Publication number | Publication date |
---|---|
US20130044925A1 (en) | 2013-02-21 |
CN102985924A (zh) | 2013-03-20 |
JP5139603B2 (ja) | 2013-02-06 |
US9111027B2 (en) | 2015-08-18 |
JPWO2012111288A1 (ja) | 2014-07-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5139603B2 (ja) | 類似症例検索装置および類似症例検索方法 | |
JP5054252B1 (ja) | 類似症例検索装置、類似症例検索方法、類似症例検索装置の作動方法およびプログラム | |
Si et al. | Fully end-to-end deep-learning-based diagnosis of pancreatic tumors | |
US8238663B2 (en) | Similar image search apparatus and computer readable medium | |
JP5852970B2 (ja) | 症例検索装置および症例検索方法 | |
Azar et al. | Decision tree classifiers for automated medical diagnosis | |
JP4979842B1 (ja) | 類似症例検索装置および類似症例検索方法 | |
JP5475923B2 (ja) | 類似症例検索装置および類似症例検索方法 | |
Nandi et al. | Classification of breast masses in mammograms using genetic programming and feature selection | |
WO2013001584A1 (ja) | 類似症例検索装置および類似症例検索方法 | |
JP5462414B2 (ja) | 類似症例検索装置および関連度データベース作成装置並びに類似症例検索方法および関連度データベース作成方法 | |
US20220301670A1 (en) | Automated information extraction and enrichment in pathology report using natural language processing | |
JP2011118543A (ja) | 症例画像検索装置、方法およびプログラム | |
JP2013525898A (ja) | 放射線ディスクリプタを用いた報告ビューア | |
JP2014029644A (ja) | 類似症例検索装置および類似症例検索方法 | |
Niazi et al. | Visually meaningful histopathological features for automatic grading of prostate cancer | |
US11062448B2 (en) | Machine learning data generation support apparatus, operation method of machine learning data generation support apparatus, and machine learning data generation support program | |
TW200817956A (en) | Clinician-driven example-based computer-aided diagnosis | |
US20180053297A1 (en) | Methods and Apparatuses for Detection of Abnormalities in Low-Contrast Images | |
JP2010172559A (ja) | 医療診断支援システム及び医療診断支援装置 | |
JP2012179336A (ja) | 病理画像診断支援装置 | |
Roge et al. | Evaluating the sensitivity of deep learning to inter-reader variations in lesion delineations on bi-parametric MRI in identifying clinically significant prostate cancer | |
CN111265234A (zh) | 一种肺部纵膈淋巴结的性质判定方法及系统 | |
Sajiv et al. | Machine Learning based Analysis of Histopathological Images of Breast Cancer Classification using Decision Tree Classifier | |
Manickavasagam et al. | Deep Belief Network for Lung Nodule Segmentation and Cancer Detection. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201280001935.0 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2012529034 Country of ref document: JP |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 12747873 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 12747873 Country of ref document: EP Kind code of ref document: A1 |