JP2010211452A - Case image retrieval device, method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To retrieve a proper case image according to the experience or speciality of a user. <P>SOLUTION: A similar image retrieval part 3 specifies a retrieval range corresponding to the set of classification conditions and a group included in user attribute information; and retrieves only a case image applied with the image attribute information belonging to the specified retrieval range. A similar image retrieval part 3 displays the specified case image (similar case image) at a display part 5. Various case information stored in an image information DB 4 corresponding to the similar case image can be also displayed at the display part 5. When the similar image retrieval part 3 or the display part 5 receives the user attribute information of "outside of facility" from a user authentication part 7, the display part 5 anonymitizes and displays personal information (information for specifying an individual such as a patient name and birth date) included in diagnostic information by replacing it with a meaningless character string or does not display personal information at all. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a technique for retrieving and presenting a case image similar to a diagnosis target image based on a feature amount extracted from the diagnosis target image.

  In Patent Document 1, an ROI (Region of Interest) is provided for image data, and information such as image data having similar features on the image and diagnosis results of similar image data is retrieved from a database and displayed.

  Patent Documents 2 and 3 are examples of techniques for automatically extracting a lesion area.

  Patent Document 4 is an example of a technique for extracting a lesion area based on position information of an attention area.

  Patent document 5 is an example of a similar image search system.

  Non-Patent Document 1 is an example of a technique for calculating an image feature amount in a region of interest.

JP 2004-5364 A JP 2005-198887 A Japanese Patent Laid-Open No. 2005-246032 JP 2008-245719 A JP 2008-257292 A

Couts, Edwards, Taylor, “Dynamic Appearance Model”, 5th Computer Vision European Conference Report, Germany, Springer, 1998, Vol. 2, 484-498 (TF Coootes, GJ Edwards, CJ Taylor “Active Appearance Models” , In Proc. 5th European Conference on Computer Vision, Springer)

  In the prior art, an ROI is provided for image data, and information such as image data having similar features on the image and diagnosis results of the similar image data is retrieved from a database and displayed. Such techniques can also be used by inexperienced users, such as residents and non-specialists, to learn diagnosis. However, when a typical case and a rare case are mixed and displayed for an inexperienced user, the case may be misunderstood, for example, the findings of the rare case may be characteristic of the case. . In addition, when staff outside the hospital or students use this technology, hospital patient information may be leaked.

  The present invention searches for an appropriate case image according to the user's experience and specialization. Further, preferably, the present invention presents an appropriate case image and its accompanying information according to the user's affiliation.

  The case image search device according to the present invention includes an acquisition unit that acquires attribute information indicating an attribute relating to a user's experience or expertise, and a determination that determines a search range of case images in a database corresponding to the attribute information acquired by the acquisition unit An input unit that inputs a desired diagnosis target image from the imaging device, a calculation unit that calculates a first feature amount that is a feature amount of the diagnosis target image input by the input unit, a first feature amount, A search unit that searches a similar case image similar to the diagnosis target image from the database by comparing the second feature amount that is a feature amount of the case image in the database belonging to the search range determined by the determination unit; Prepare.

  Preferably, the retrieval unit retrieves a similar case image from the database by comparing the feature amount related to at least one of the pixel value and shape of the diagnosis target image with the feature amount related to at least one of the pixel value and shape of the case image. .

  Preferably, the search unit calculates a similarity based on the first feature value and the second feature value, and searches the database for similar case images similar to the diagnosis target image according to the calculated similarity.

  Preferably, a selection unit that selects a lesion area from the diagnosis target image is provided, and the search unit searches the database for a similar case image using the feature quantity related to the lesion area selected by the selection unit as a first feature quantity.

  Preferably, the selection unit selects a lesion area based on a feature amount related to at least one of a pixel value and a shape of the diagnosis target image.

  Preferably, a display unit that displays similar case images searched by the search unit is provided.

  Preferably, the database stores incidental information corresponding to the case image, and the display unit displays incidental information of the database corresponding to the similar case image.

  Preferably, the display unit displays the similar case image in a format corresponding to the attribute information.

  Preferably, the attribute information includes identification information of a terminal used by the user.

  Preferably, an authentication unit that authenticates the user is provided, and the acquisition unit acquires attribute information indicating an attribute of the user authenticated by the authentication unit.

  Preferably, the case information corresponding to the target image, which is a desired case image to be determined as to whether registration in the database is necessary, and the feature amount of the target image, and a case image different from the target image registered in the database The case information corresponding to the registered image and the feature amount of the registered image are compared, and it is determined whether or not the target image is registered in the database based on the comparison result. The target image determined to be registered in the database is stored in the database. A registration unit for registration is provided.

  In the case image search method according to the present invention, a computer acquires attribute information indicating attributes relating to a user's experience or expertise, and a step of determining a search range of case images in a database corresponding to the acquired attribute information; A step of inputting a desired diagnosis target image from the imaging device, a step of calculating a first feature amount that is a feature amount of the input diagnosis target image, a first feature amount, and a database belonging to the determined search range A step of searching the database for a similar case image similar to the diagnosis target image is performed by comparing the second feature amount, which is a feature amount of the case image.

  A program for causing a computer to execute this case image search method is also included in the present invention.

  In the present invention, since the search range of case images in the database is determined according to the user's experience or specialty, appropriate similar case images can be output for each user.

Schematic configuration diagram of similar case search system Block diagram showing details of the registration unit Flow chart of registration process Diagram showing an example of judgment conditions The figure which shows an example of a search range determination table The figure which shows an example of a display format table

  FIG. 1 is a schematic configuration diagram of a similar case search system 100 according to a preferred embodiment of the present invention. This system includes an image data input unit 1, an image analysis unit 2, a similar image search unit 3, an image information DB 4, a display unit 5, a registration unit 6, and a user authentication unit 7. The similar case retrieval system 100 may be configured by a computer (a circuit including a CPU, a RAM, a ROM, and other circuits necessary for arithmetic processing, a data storage medium, a data input / output circuit, a display circuit, an operation device, a communication circuit, etc.). it can.

  Each block of the similar case search system 100 may be integrally configured in one computer. However, the similar case search system 100 is configured by configuring each block with a single computer and connecting them with a network. It may be configured. The image analysis unit 2, the similar image search unit 3, and the registration unit 6 can be replaced with a program module (stored in a computer-readable recording medium such as a ROM) executed by a computer, and a specific hardware configuration is not necessarily required. Is not required. The display unit 5 can be composed of a liquid crystal display, and the image information DB 4 can be composed of a hard disk or the like.

  The image data input unit 1 is a diagnostic image (query image) such as a CT image, an MRI image, a PET image, an X-ray image (including a CR image), a US (ultrasound) image, an endoscopic image, a mammography, or a pathological image. Enter. The image data input unit 1 can also input these query images and incidental information such as the size of the query image and the imaging date and time and case information. The communication protocol between the image data input unit 1 and the medical diagnostic imaging apparatus, and the format of the query image and supplementary information comply with a predetermined standard such as DICOM (Digital Imaging and Communication in Medicine).

  The image analysis unit 2 analyzes the query image input from the image data input unit 1 and creates a feature amount (first feature amount) necessary for searching for similar images stored in the image information DB 4. A method of creating the feature amount is performed in the same manner as the image analysis unit 21 described later.

  The similar image search unit 3 compares the first feature amount created by the image analysis unit 2 with the feature amount (second feature amount) of the case image stored in the image information DB 4 and calculates the degree of similarity between the two. To do. The second feature amount is created and stored in advance from the case image according to the same method as the feature amount calculation method taken by the image analysis unit 2. The similar image search unit 3 has the highest similarity with the feature amount information of the query image among all the case images stored in the image information DB 4 or the case images corresponding to the parts designated in advance by the operation device. A case image corresponding to the feature amount is identified. Then, the similar image search unit 3 displays the specified case image and various case information stored in the image information DB 4 corresponding to the case image on the display unit 5.

  The display unit 5 can be replaced with an output unit other than the video output unit. For example, instead of the display unit 5 or together with the display unit 5, the specified case image and various types of information stored in the image information DB 4 corresponding to the case image may be output by the printing unit. If it is not image information, a search result can be output by outputting a synthesized voice from the voice output means. Alternatively, the output means may be a network I / F, and the search result may be output to a desired terminal 10 (such as a personal computer, a mobile phone, or a PDA). Although the desired terminal 10 may include the display unit 5, when displaying the searched case image on the display unit 5 of the terminal 10, the user authentication unit 7 authenticates the user of the terminal 10 and is similar Case image search needs to be permitted.

  The image information DB 4 stores case information including case images that are search target images. Here, case information may be classified and stored for each disease, and statistical information, disease information, and disease information and medical information for diseases that are easily mistaken for specific diseases are linked and registered. May be.

  Further, the case information includes text-based diagnostic information such as a diagnostic report created by an interpretation doctor, in addition to a case image of each disease that has been confirmed and diagnosed.

The statistical information for each disease includes, for example, the following information.
・ Similarity in representative cases (highest similarity in disease)
・ Average similarity ・ Total number of cases registered in the case DB ・ Number of cases (case pattern) registered near the representative case ・ Characteristics in the case pattern… Characteristics of patients in the main symptoms and case patterns… Average age, medical history, Smoking history, etc., national / regional morbidity, etc. The image information DB 4 stores feature quantities (second feature quantities) extracted from lesion sites of registered case images. However, if the case image itself is stored in the image information DB 4, the second feature amount can be obtained by analyzing the case image later by the image analysis unit 2 or the like. It is not always necessary to register the feature quantity 2 in the implementation of the present invention.

  The display unit 5 may display not only image data with high similarity but also statistical information such as a similar image search result for each disease of similar images in the database as in Patent Document 3.

  Case images registered in the image information DB 4 by the registration unit 6 are various medical imaging apparatuses (CT apparatus, MRI apparatus, PET apparatus, X-ray apparatus, US apparatus, endoscope, mammography apparatus, microscope, and the like as the query image. For example). Further, the second feature amount corresponding to the case image registered in the image information DB 4 by the registration unit 6 is obtained by the same image analysis as that of the image analysis unit 2. The case information corresponding to the case image registered in the image information DB 4 by the registration unit 6 is obtained by an input operation of the operation device by the user. In the embodiment of the present application, the registration unit 6 performs registration so that image data included in the image information DB 4 has no feature amount bias.

  The user authentication unit 7 authenticates whether or not the user of the terminal 10 that is to perform the similar image search is a regular user. For example, the user authentication unit 7 receives input of authentication information (a set of ID and password, biometric information, etc.) from the operation device, and the received authentication information and authentication information stored in advance in the image information DB 4 or other recording medium. If the two match, the user is allowed to start the similar image search process by the similar image search unit 3, and if they do not match, the start of the process is rejected. However, as will be described later, even if both match, the search range of case images by the similar image search unit 3 is limited according to the contents of the authentication information. The method of inputting authentication information is arbitrary. For example, the user himself / herself inputs a key for the ID and password, or from a magnetic stripe, IC chip, or QR code embedded in the hospital staff card or employee card carried by the user. Reading or reading biometric information such as a user's own retina, vein, and fingerprint with a known biometric sensor device.

  The user authentication unit 7 can authenticate the user for each desired terminal 10 that has accessed the similar case search system 100. The user authentication unit 7 identifies the terminal 10 to which the authentication information has been input using unique information such as a MAC address or an IP address. Therefore, even if there are a plurality of terminals 10 that have accessed the user authentication unit 7, each of them can be identified and a user for each terminal 10 can be individually authenticated. A user who is permitted to start a similar image search process by the user authentication unit 7 is called an authenticated user.

  FIG. 2 shows details of the registration unit 6. The registration unit 6 includes an image analysis unit 21 and a DB analysis unit 22. The image analysis unit 21 includes an attention area extraction unit 23 and a feature amount conversion unit 24. Each block of the registration unit 6 may be integrally configured in one computer, but each block may be configured as a single computer and the registration unit 6 may be configured by connecting them through a network. Good. In addition, the DB analysis unit 22, the attention area extraction unit 23, and the feature amount conversion unit 24 can be replaced with a program module (stored in a computer-readable recording medium such as a ROM) executed by a computer. Does not require hardware configuration.

  The image analysis unit 21 uses, as registration candidate images, desired case images such as CT images, MRI images, PET images, X-ray images (including CR images), US (ultrasound) images, endoscopic images, and pathological images. input. Also, case information corresponding to the registration candidate image is input.

  The attention area extraction unit 23 extracts a part to be searched from the input registration candidate image, that is, an attention area such as a lesion or an organ included in the case image. The method of extracting the attention area is arbitrary. For example, a method of automatic detection such as Patent Documents 1 and 2, a method of extracting a region based on position information of a region of interest as in Non-Patent Document 1, and the like, known image contours / pixel values, position information, etc. An image processing technique using the above feature may be used. The region of interest may not necessarily be fully automatically extracted, but may be extracted semi-automatically by an image processing technique using a feature in the vicinity of the region designated by the authentication user on the operation device.

  Next, the feature amount conversion unit 24 calculates a feature amount for the extraction region of the attention region extraction unit 23. The calculated feature values include feature values related to pixel values such as the average, variance, maximum value, minimum value, and luminance histogram of image data, the position of the extraction region, the circularity and moment of the contour of the extraction region, the radius of the cross section, the volume, In addition to the feature amount relating to the shape such as the area, the shape information and texture information of the lesioned part of the ROI (region of interest) obtained by the method of Patent Document 4 may be used. Alternatively, a value obtained by processing information once registered in the image information DB 4 with a predetermined mathematical formula (for example, principal component analysis, independent component analysis) may be used as the feature amount. In short, the outline / pixel value, position information, etc. of an image obtained by a known method can be used as the feature amount.

  Further, as a preprocessing for calculating the feature amount, a lesion area of the image may be specified, and the feature quantity may be calculated for the lesion area. Specific examples of the lesion area include an automatic detection technique as described in Patent Documents 2 and 3, and a technique for extracting an area based on position information of the attention area as described in Non-Patent Document 1.

  The identification of the lesion area is not necessarily fully automatic, but semi-automatically by an image processing technique using a feature amount (a feature amount related to a pixel value or a feature related to a shape such as a contour) in the vicinity of the region designated by the user on the operation device. It may be extracted. In this way, by using a feature amount specialized for a lesion site, a case image similar to the focused lesion site can be searched.

  Based on the comparison between the feature amount of the registration candidate image calculated by the feature amount conversion unit 24 and the feature amount of the case image stored in the image information DB 4, the DB analysis unit 22 stores the registration candidate image in the image information DB 4. It is determined whether or not to register, and registration candidate images are registered according to the determination result, or discarded without registration.

  Specifically, the DB analysis unit 22 performs a registration process as shown in FIG.

  First, in S1, the DB analysis unit 22 has registered case information (determination conditions) corresponding to any of the case images already registered in the image information DB 4 and case information (registration candidate case information) corresponding to the registration candidate image. To determine whether or not they are the same. FIG. 4 shows an example of determination conditions. The determination conditions include information related to the patient such as the patient name, age, and gender indicated by the report information stored in the header of the image data, the type (CT, MRI, etc.), model number, and imaging of the device that performed the imaging. Examples include information on imaging such as conditions, findings, information on disease names and other definitive diagnoses, imaging purposes, anatomical positions of lesion sites, lesion sites (organs), and the like. The determination (matching) of the identity between the determination condition and the registered candidate case information is performed for each item included in the case information. In addition, whether or not the registrant of diagnostic information is a person who has a certain qualification (certified physician, specialist), the length of report information, the date and time of image capture, the date and time of registration of case information, and the like can also be used as judgment conditions.

  In S2, the DB analysis unit 22 calculates the number of items that store the same information as both the determination condition and the registration candidate case information as a result of the matching in S1. When the number exceeds the first threshold value specified in advance by the operation of the operation device by the user having the administrator authority (administrator), the registration candidate image corresponding to the registration candidate case information is the registration analysis target image. And the process proceeds to S3.

  For example, if the threshold = 3 and the patient ages stored in the case information of both the determination condition and the registration candidate case information, the imaging conditions, and the anatomical position are all the same, the registration candidate case information The corresponding registration candidate image is determined as a registration analysis target image.

  However, if at least the disease name included in the determination condition is the same as the disease name included in the registration candidate case information, the registration candidate image is unconditionally determined to be a registration analysis target image, and the process proceeds to S3. This is to eliminate duplicate registration of images with the same disease name and similar image findings.

  For the same purpose, if at least the disease name included in the determination condition and the disease name included in the registration candidate case information are not the same, the registration candidate image is unconditionally determined to be an image to be registered. In this case, S3 is omitted and the process proceeds to S4.

  In S3, the DB analysis unit 22 calculates the feature amount of the registration analysis target image calculated by the feature amount conversion unit 24 and the feature amount of the case image corresponding to the case information storing the same determination condition as the registration analysis target image. The similarity is calculated. The feature amount conversion unit 24 can omit analysis of an image that is not a registered analysis target image. As a method for calculating the similarity, a known method, for example, a difference between feature value values, a least square method on a feature amount space (a weighted space is also possible), or the like can be employed. For convenience of explanation, the similarity S is defined by the following mathematical formula (the mathematical formula described in paragraph 0048 of Patent Document 5). However, in order to implement the present invention, the calculation standard (norm) of the similarity S is not limited to this definition.

The feature quantity m _i (i = 1, 2,..., N) is the feature quantity of the registered case image, and M _i (i = 1, 2,..., N) is the feature quantity extracted from the registered analysis target image. is there. w _i (i = 1, 2,..., n) is a weighting coefficient for each feature amount defined in advance for each disease. A subscript _i of m _i , M _i , and w _i indicates a coordinate axis on the feature amount space. The closer the two feature quantities are, the smaller the value of the similarity S is, and the farther the feature quantities are, the larger the value of the similarity S is.

Preferably, in S1, if the disease name included in the determination condition and the disease name included in the registration candidate case information are the same and the region and the anatomical position are the same, the registration candidate image is registered unconditionally. If it is determined that the image is an analysis target image, it is possible to prevent comparison of feature quantities of unrelated parts in S3. For example, if the feature quantity m _i (i = 1, 2,..., N) extracted from the registration analysis target image is a value (n-dimensional multivalued data) related to the lung segment S1 (part and anatomical position). For example, the feature amount M _i (i = 1, 2,..., N) of the registered image that is the target of calculating the similarity S is also a value (n-dimensional multi-value data) related to the lung segment S1.

  In S4, if the calculated similarity is below a predetermined second threshold, the DB analysis unit 22 determines that the registration analysis target image is a non-registration image and does not register it in the image information DB4. If the calculated similarity is not less than a predetermined threshold, the registration analysis target image is determined to be a registration target image and registered in the image information DB 4. At this time, the case information corresponding to the registration analysis target image is also registered in the image information DB 4.

  The second threshold value TH is determined by the following formula, for example.

TH = α × N + β
N is the total number of case images registered in the image information DB 4, and α and β are constants. α, β, or the second threshold value itself may be arbitrarily specified by an input operation of the operating device by the administrator. Generally, TH = f (N). f is an arbitrary function that tends to increase as N increases. The function f can be arbitrarily specified by an input operation of the operating device by the user, and the arithmetic device (CPU) may calculate the actual second threshold according to the specified function.

  In other words, when the number of registered case images is small, the threshold value becomes small, and the similarity to the registered case images becomes a relatively small value (high correlation). When the number of registered images is small, the registration analysis target image having a high threshold and a small similarity and a high correlation with the registered case images is not registered. Alternatively, when the registration number N is less than a certain threshold (for example, 100), the registration analysis target image and the case information may be registered unconditionally, and when the number of cases is small, enhancement of the image information DB 4 may be prioritized.

  It is not necessary to register an image that has been determined to be a non-registration image, but to delete a registered image in the image information DB 4 whose similarity with the non-registration image is lower than the second threshold and corresponding case information. However, if an image other than the registration target and case information corresponding thereto are registered instead, it is possible to avoid that images having similar characteristics overlap in the image information DB 4. The user can arbitrarily specify whether to select an image that has been determined to be a non-registration image or replace it with a registered case image, and either one of the images depending on the designation. May be registered in the image information DB 4. Alternatively, if the case information of the registered case image contains high priority information, for example, if the date information is new, the report information is long, the registrant is qualified, or the amount of image data is large, the registered information The case image may be left in the image information DB 4, and if not, the registered case image may be replaced with the registration analysis target image. On the other hand, if the case information of the registered case image does not include high priority information, the registered case image is deleted from the image information DB 4 and a new image not registered is registered in the image information DB 4. Also good.

  In addition, the registration candidate image is not necessarily a newly input image, and may be arbitrarily selected from case images already registered in the image information DB 4. In this case, the registration unit 6 leaves the image determined to be a registration target in the image information DB 4 as it is, and deletes the image determined not to be a registration target from the image information DB 4. Similarly to the above, when the case information includes high priority information, the case image corresponding to the case information may be left and the case image corresponding to the case information that does not include the high priority information may be deleted. In short, if there are two images whose similarity is less than the second threshold, one of them may be deleted.

  In general, when there are two or more images in the image information DB 4, the similarity is calculated from a pair composed of any two different images in the image information DB 4, and the magnitude of the similarity value and the second threshold value are calculated. If one of the images in the pair is deleted according to the relationship and this is exhaustively repeated for all the image pairs, it is possible to eliminate duplication of similar images from the images registered in the image information DB 4. it can. This may be executed by the user specifying a command such as “refresh image information DB 4” from the operation device, or may be executed periodically with a desired scheduling.

  Further, after registering the image that is not the registration target in the image information DB 4 together with the flag indicating the image that is determined to be the image that is not the registration target, the case image to which the flag is given from the search target of the similar image search unit 3 is displayed. It is possible to exclude only the images without the flag and to search for cases. That is, the similar image search unit 3 compares only the second feature amount of the case image to which the flag is not assigned with the first feature amount.

  As described above, by controlling the registration of the case image in the image information DB 4 according to the similarity of the feature amount, a lot of similar case images appear as search results, which hinders diagnosis by the diagnostician. The possibility of showing similar search results is reduced.

  Referring to FIG. 1 again, if the similar search is permitted by the authentication, the acquiring unit 7a identifies the organization to which the authenticated user belongs, the title of the authenticated user corresponding to the input authentication information, and the terminal 10 used by the authenticated user. User attribute information related to the attribute of the authenticated user such as information is acquired from the image information DB 4, and this user attribute information is transmitted to the similar image search unit 3. It is assumed that user attribute information corresponding to authentication information of a user permitted by the administrator to use the system 100 is stored in the image information DB 4 in advance. Further, when the acquisition unit 7a identifies the terminal 10 used by the authenticated user, the acquisition unit 7a transmits the identification information to the similar image search unit 3 and the display unit 5.

  In addition to the image information DB 4, the user attribute information may be stored in a portable recording medium in which information cannot be tampered with, for example, an IC chip of a card of a hospital staff card, which is read by the acquisition unit 7 a. You may send it.

  When the similar image search unit 3 receives the user attribute information received from the acquisition unit 7a and the identification information of the terminal 10, the similar image search unit 3 refers to the search range determination table stored in advance in the image information DB 4 or other recording medium, and receives the received user Sets the search range of images corresponding to attribute information.

  Here, the user attribute information includes at least one of “experience” and “professional” classification conditions, and can also include a “facility” classification condition. Two or all three of these three classification conditions may be included in the user attribute information, and the three classification conditions are not in a logical exclusive relationship incompatible with each other. For example, user attribute information in which “experience” is “certified physician / specialist”, “specialty” is “clinical department”, and “facility” is “intra-facility” is transmitted from the user authentication unit 7 to the similar image search unit 3 or the display unit 5. May be sent to.

  In addition, the user attribute information includes the groups “clinic department”, “radiology department” or “group” corresponding to the classification conditions of the group “resident” or “certified physician” / specialist ”/“ specialty ”corresponding to the classification condition of“ experience ”. It shall include the group “inside facility” or “outside facility” corresponding to “pathology” / “facility”. However, the user attribute information may be anything as long as it is information that categorizes the user, and is not limited to the above example.

  FIG. 5 shows an example of a search range determination table stored in the image information DB 4. The search range determination table includes the classification conditions of “experience”, “specialty”, and “facility”, “residential physician”, “certified physician” / specialist ”,“ clinical department ”,“ radiology department ”,“ pathology department ” It is assumed that information defining the search range of the image information DB 4 corresponding to each combination of “inside facility” and “outside facility” groups is stored.

  In FIG. 5, the “experience” is a “residential doctor” group, “typical case only” search range, the “experience” is a “certified physician / specialist” group, “all diseases” search range, and “specialty” The “Clinical Department” group includes the “Disease for each department”, the “Specialist” group includes the “Radiology Department”, the “Disease obtained by CT and MR images”, and the “Specialist” group includes the “Pathology Department” group. The search range includes “all diseases” for the group “pathological image”, “institution” “in-facility”, and “public cases” for the group “institution” “out of facility”. It is prescribed.

  As for “experience”, inexperienced authenticated users such as “residents” can only search for case images to which the incidental information of “typical cases” has been added, and have abundant experience such as “certified physicians / specialists”. The user sets all diseases including rare cases within the search range. Whether “experience” is “resident” or “certified physician / specialist” may be determined from the years of experience of the authenticated user recorded in a recording medium such as an IC chip.

  In addition, regarding “specialized”, all the images and regions of diseases specialized by the authenticated user are included in the search range, and images and regions of diseases that are not specialized are out of the search range. It is assumed that supplementary information related to specialization is given to each case image. Whether the “specialty” is “clinical department”, “radiology department” or “pathology department” may be determined from the department to which the authenticated user belongs recorded on a recording medium such as an IC chip.

  Also, instead of adding incidental information corresponding to user attribute information to the case image file itself as the case image header or meta information, the image information DB 4 is provided with an attribute item indicating the attribute of the search range for each case image. The attribute item may be input for each case image via the operation device. Alternatively, the keyword “typical” may be extracted from the diagnostic information such as header information and report information of the case image, and the attribute “typical case” corresponding to the extracted keyword may be set in the above item. Alternatively, the number of times that the similar image search unit 3 outputs the case image to the display unit 5 is counted for each case image, and the case images are ranked according to the number of times. Then, “typical” may be set for incidental information and attribute items corresponding to case images whose rank is equal to or higher than a predetermined threshold (for example, the rank is up to the top ten). Alternatively, information such as “shape” and “concentration” is extracted from the diagnosis information such as header information and report information of the case image, and the combination of the extracted information and the diagnosis name is used to determine whether “specialty” is “clinical department” or “ It may be determined whether it is “radiology” or “pathology”, and this determination content may be set in the incidental information or attribute item.

  In addition, for “facility”, when the authentication of the user authentication unit 7 is requested from the “outside facility” terminal 10 such as telemedicine, only the case images to which the diagnosis information “public case” is given are searched. When the authentication of the user authentication unit 7 is requested from “within the facility”, all the case images are within the search range.

  Whether the user attribute information of the authenticated user is “outside facility” or “inside facility” is whether the identification information of the terminal 10 to which the authenticated user has input the user attribute information is “internal network address” (such as a local IP address). It can be determined by whether it is an “external network address” (global IP address). Alternatively, whether the user attribute information of the authenticated user is a “related person” such as a doctor or an engineer employed at the facility, or an “external person” such as another doctor or student. It is also possible to discriminate by the user ID read from the recording medium such as the IC chip, and “inside the facility” if it is determined as “related person”, and “outside of the facility” if it is “outside person”.

  In short, how to identify or give the attribute of the search range for the case image is arbitrary as long as it corresponds to the user attribute information. Information indicating search range attributes for case images is referred to as image attribute information. The image attribute information can be information stored in the attribute item of the image information DB 4 and incidental information of the case image.

  The similar image search unit 3 specifies a search range corresponding to a set of classification conditions and groups included in the user attribute information from the search range determination table, and only case images to which image attribute information belonging to the specified search range is given. Search for. That is, the similar image search unit 3 uses the first feature amount created by the image analysis unit 2 as the image attribute information belonging to the search range specified from the search range determination table among the case images stored in the image information DB 4. Is compared with the feature quantity (second feature quantity) of only the case image to which is assigned, and the similarity between the two is calculated, corresponding to the second feature quantity having the highest similarity with the feature quantity information of the query image The case image to be identified is specified. Then, the similar image search unit 3 displays the identified case image (similar case image) on the display unit 5. Various case information stored in the image information DB 4 corresponding to similar case images can also be displayed on the display unit 5. The same can be done by printing out similar case images.

  When there are two or more classification conditions for the user attribute information, the similar image search unit 3 searches only case images belonging to the overlapping search range corresponding to each of them. For example, in the case of an “authentician” and “outside facility” authenticated user, only the case images to which the “typical case” attribute and the “publicly available” attribute are assigned are searched. Alternatively, in the case of an “authenticated physician / specialist” and “clinical department” authenticated user, a case image linked to the specialized field of the user separately designated from the terminal 10 and a “typical case” of other case images Search only those with attributes. The specialized field may be determined from the department to which the user attribute information belongs. For example, if the department is “pathology”, the “specialty” is “pathology”. Alternatively, when the department belongs to “internal medicine”, “surgery”, “ophthalmology”, “dermatology”, “psychiatry”, etc., “specialty” becomes “clinical department”.

  The method of outputting the case image to the display unit 5 or a printer is arbitrary. For example, the similar case images corresponding to the query image are arranged in the descending order of similarity and displayed on the display unit 5 or printed out. Alternatively, an image is displayed for each case based on diagnostic information such as a report.

  Moreover, when the diagnostic information of image information DB4 contains the personal information of the patient who does not plan browsing of an outsider, it is not preferable to disclose this to outsiders without limitation. Therefore, when the similar image search unit 3 or the display unit 5 receives the user attribute information “outside facility” from the user authentication unit 7, as described above, when it is determined that the authenticated user is an “outsider” Alternatively, when it is determined that the identification information of the terminal 10 is “external network address”, the display unit 5 uses personal information (information that can specify an individual such as a patient name and date of birth) included in the diagnostic information as meaningless. Display anonymized by replacing with a character string or do not display personal information at all. A display format table that defines in what format the diagnostic information is to be displayed for which user attribute information is stored in the image information DB 4 or other storage medium, and the user attribute information of the authenticated user according to the display format table The diagnostic information may be displayed in a format corresponding to (see FIG. 6).

  Alternatively, instead of removing case images that do not fall within the search range corresponding to the user attribute information of the authenticated user defined by the search range determination table, all the case images are taken into the search range, but the display unit 5 Case images that do not fall within the search range may be hidden, or image information that is different from the case images that fall within the search range may be added to the case images that do not fall within the search range so that both can be identified. .

  For example, when “resident” is the user attribute information of the authenticated user, the display unit 5 uses a mark or window color different from the “typical” case image for the case image to which “typical case” is not assigned. Make things. Alternatively, when “pathology” is the user attribute information of the authenticated user, the name of the imaging device (CT, MRI, microscope camera, etc.) that is the input source of the case image is attached to each case image.

  The display format of the similar case image by the display unit 5 may correspond to the user attribute information of the authenticated user. For example, when the “resident” is the user attribute information of the authenticated user, the display unit 5 displays the case image and the diagnosis report in parallel, and when the “certified physician / specialist” is the user attribute information of the authenticated user, Only the case image is displayed on the display unit 5, and the diagnosis report is displayed only when there is an instruction from the user. In this way, it is possible to provide detailed information to inexperienced users such as resident doctors, and omit providing detailed information to experienced specialists. Similar output control is also possible when printing out similar case images and diagnostic information thereof.

  Thus, since the search range of the case image is determined according to the user's attribute, an appropriate similar case image can be output for each user. Also, since the output format is changed according to the user's attributes, personal information can be prevented from being disclosed to outsiders, and necessary information can be appropriately provided.

1: image data input unit, 2: image analysis unit, 3: similar image search unit, 4: image information DB, 5: display unit, 6: registration unit, 7: user authentication unit, 7a: acquisition unit, 10: terminal

Claims (13)

An acquisition unit that acquires attribute information indicating attributes related to the user's experience or expertise;
A determination unit for determining a search range of case images in a database corresponding to the attribute information acquired by the acquisition unit;
An input unit for inputting a desired diagnosis target image from the imaging device;
A calculation unit that calculates a first feature amount that is a feature amount of the diagnosis target image input by the input unit;
Similar cases similar to the diagnosis target image by comparing the first feature amount and a second feature amount that is a feature amount of the case image in the database belonging to the search range determined by the determination unit A search unit for searching for an image from the database;
A case image retrieval apparatus comprising:   The search unit compares the feature amount related to at least one of the pixel value and shape of the diagnosis target image with the feature amount related to at least one of the pixel value and shape of the case image, thereby obtaining the similar case image from the database. The case image search device according to claim 1 for searching.   The search unit calculates a similarity based on the first feature quantity and the second feature quantity, and searches the database for similar case images similar to the diagnosis target image according to the calculated similarity degree The case image search device according to claim 1 or 2. A selection unit that selects a lesion area from the diagnosis target image,
The case image search device according to any one of claims 1 to 3, wherein the search unit searches the database for the similar case image using the feature amount relating to the lesion area selected by the selection unit as the first feature amount.   The case image search device according to claim 4, wherein the selection unit selects the lesion area based on a feature amount related to at least one of a pixel value and a shape of the diagnosis target image.   The case image search device according to claim 1, further comprising a display unit that displays a similar case image searched by the search unit. The database stores incidental information corresponding to the case image,
The case image search device according to claim 6, wherein the display unit displays incidental information of the database corresponding to the similar case image.   The case image search device according to claim 6 or 7, wherein the display unit displays the similar case image in a format according to the attribute information.   The case image search device according to claim 1, wherein the attribute information includes identification information of a terminal used by a user. It has an authentication unit that authenticates users,
The case image search device according to claim 1, wherein the acquisition unit acquires attribute information indicating an attribute of a user authenticated by the authentication unit.   Case information corresponding to a target image that is a desired case image to be determined as to whether or not registration in the database is necessary, a feature amount of the target image, and a case image different from the target image registered in the database The case information corresponding to a registered image and the feature amount of the registered image are compared, it is determined whether to register the target image in the database based on the comparison result, and it is determined to register in the database The case image search device according to any one of claims 1 to 10, further comprising a registration unit that registers the target image in the database. Computer
Obtaining attribute information indicating attributes relating to the user's experience or expertise;
Determining a search range of case images in the database corresponding to the acquired attribute information;
Inputting a desired diagnosis target image from the imaging device;
Calculating a first feature amount that is a feature amount of the input diagnosis target image;
By comparing the first feature quantity with a second feature quantity that is a feature quantity of the case image in the database belonging to the determined search range, a similar case image similar to the diagnosis target image is obtained. Searching from the database;
Case image search method for executing   A program for causing a computer to execute the case image search method according to claim 12.

Images