Control program and case search device

Abstract

To provide a control program and a case retrieval device that can perform retrieval with higher accuracy.SOLUTION: A control program controls a case retrieval device (diagnostic console 3) to execute case retrieval (case retrieval processing). The control program controls a computer (control unit 31) of the case retrieval device to execute retrieval steps (step S36, S37) of performing retrieval using a dynamic image and outputting a similar case image similar to the dynamic image or a case candidate related to the dynamic image.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to a control program and a case search device. [Background technology]

[0002] Conventionally, when searching for similar cases using a search device for searching cases, a doctor first checks patient information such as the patient's chief complaint and additional patient information, as well as objective information based on the captured test images. The doctor then inputs the patient's symptoms and possible disease names as keywords into the search device, which then performs a search based on the input keywords. In such keyword searches, there is a possibility that the cases found will vary depending on the types of keywords the doctor uses and the doctor's skill in interpreting symptoms.

[0003] In this regard, Patent Document 1 describes a similar case search device that performs machine learning on features obtained from case images and diagnosis results of the case images, and searches for case images similar to the diagnostic target image based on the features obtained from the diagnostic target image. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2007-279942 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the medical images that are the case images and diagnostic target images in the invention described in Patent Document 1 are still images. The amount of information contained in a still image is less than the amount of information contained in a dynamic image, which is a moving image. Therefore, a search using a still image in a search device may have lower accuracy than a search using a moving image.

[0006] An object of the present invention is to provide a control program and a case search device that are capable of performing searches with higher accuracy. [Means for solving the problem]

[0007] In order to solve the above problem, the control program of the invention described in claim 1 comprises: A control program for causing a case search device to perform a case search, The control program is configured to: an acquisition step of acquiring a dynamic image; a feature amount calculation step of calculating a feature amount of a first image from the dynamic image acquired by the acquisition step; The aforementioned a search step in which a search is performed using a dynamic image, and a similar case image similar to the dynamic image or a case candidate related to the dynamic image is output; and, Run 、 The search step determines a group to which the dynamic image belongs by a trained model based on the feature amount of the first image calculated in the feature amount calculation step, and outputs the similar case image from the dynamic image belonging to the determined group based on the feature amount of the first image and the feature amount of a second image calculated from a frame image of the dynamic image belonging to the determined group, or outputs the case candidate from a diagnosis result associated with the dynamic image belonging to the determined group; The trained model is a model that has trained group classification with disease information as the correct answer based on second feature amounts calculated in advance from multiple frame images of any dynamic image. .

[0008] The invention described in claim 2 is the control program described in claim 1, The information about the disease is the name of the disease.

[0009] The invention described in claim 3 is the control program according to claim 1 or 2, The dynamic image used for the search is at least one frame image among a plurality of frame images that make up the dynamic image.

[0010] The invention described in claim 4 is the control program described in claim 3, The dynamic image used for the search is an image of a region of interest in at least one frame image among a plurality of frame images that make up the dynamic image.

[0011] The invention described in claim 5 is the control program described in claim 4, The region of interest is a region designated by the user.

[0012] The invention described in claim 6 is the control program according to any one of claims 3 to 5, The dynamic image used for the search is a plurality of frame images that make up the dynamic image.

[0013] The invention described in claim 7 is the control program according to any one of claims 3 to 5, The dynamic image used for the search is one of the frame images that constitute the dynamic image.

[0014] The invention described in claim 8 is the control program according to any one of claims 3 to 5, The dynamic image used for the search is an image of a region of interest within one of the frame images that constitute the dynamic image.

[0015] The invention described in claim 9 is the control program according to any one of claims 3 to 5, The dynamic image used for the search is only a portion of the frame images among the plurality of frame images that make up the dynamic image.

[0016] The invention described in claim 10 is the control program according to any one of claims 3 to 9, The dynamic image used for the search is an image selected by the user from among a plurality of frame images that make up the dynamic image.

[0017] The invention described in claim 11 is the control program according to any one of claims 3 to 5, The dynamic image used for the search is a series of frame images among a plurality of frame images that make up the dynamic image.

[0018] The invention described in claim 12 is 1 12. The control program according to any one of claims 1 to 11, The dynamic images are images obtained by continuously radiographing the dynamic state of a target region along the time axis.

[0019] The invention described in claim 13 is 1 13. The control program according to any one of claims 1 to 12, The dynamic image is an image obtained by continuously radiographing the dynamics of a periodic target region along the time axis.

[0020] The invention described in claim 14 is 1 14. The control program according to any one of claims 1 to 13, The dynamic image used for the search is a dynamic analysis image obtained by analyzing a dynamic image.

[0021] The invention described in claim 15 is the control program described in claim 14, The dynamic analysis image is any one of a blood flow analysis image obtained by dynamic analysis of blood flow function, a ventilation analysis image obtained by dynamic analysis of ventilation function, and an adhesion analysis image obtained by dynamic analysis of adhesion.

[0022] The invention described in claim 16 is a control program according to any one of claims 1 to 15, The similar case images are dynamic images.

[0023] The invention described in claim 17 is the control program described in claim 16, The similar case image is at least one frame image among a plurality of frame images constituting a dynamic image.

[0024] The invention of claim 18 is the control program of claim 16 or 17, The similar case images are images obtained by continuously radiographing the dynamics of the target area along the time axis.

[0025] The invention described in claim 19 is the control program according to any one of claims 16 to 18, The similar case images are images obtained by continuously radiographing the dynamics of a target region having periodicity along the time axis.

[0026] The invention described in claim 20 is the control program according to any one of claims 16 to 19, The similar case image is a dynamic analysis image obtained by analyzing a dynamic image.

[0027] The invention described in claim 21 is the control program described in claim 20, The similar case images The dynamic analysis image is any one of a blood flow analysis image obtained by dynamic analysis of blood flow function, a ventilation analysis image obtained by dynamic analysis of ventilation function, and an adhesion analysis image obtained by dynamic analysis of adhesion.

[0028] The invention described in claim 22 is a control program according to any one of claims 1 to 21, The searching step A plurality of similar case images that are similar to the dynamic image or case candidates that are related to the dynamic image are output.

[0029] The invention of claim 23 is a control program according to any one of claims 1 to 22, The searching step Similar case images similar to the dynamic image or case candidates related to the dynamic image are output in order of similarity.

[0030] The invention described in claim 24 is the control program described in claim 22, The searching step outputs similar case images similar to the dynamic image or case candidates related to the dynamic image for each disease name.

[0031] The invention described in claim 25 is the control program described in claim 24, The searching step Similar case images similar to the dynamic image or case candidates related to the dynamic image are output for each disease name and in order of similarity.

[0032] The case search device of the invention described in claim 26 comprises: Acquire dynamic images, Calculating a feature amount of a first image from the acquired dynamic image; The aforementioned A control unit is provided to perform a search using a dynamic image and output a similar case image similar to the dynamic image or a case candidate related to the dynamic image. 、 the control unit determines a group to which the dynamic image belongs using a trained model based on the feature amount of the first image, and outputs the similar case image from the dynamic image belonging to the determined group based on the feature amount of the first image and the feature amount of a second image calculated from a frame image of the dynamic image belonging to the determined group, or outputs the case candidate from a diagnosis result associated with the dynamic image belonging to the determined group; The trained model is a model that has trained group classification with disease information as a correct answer based on second feature amounts calculated in advance from a plurality of frame images of any dynamic image. do. [Effects of the Invention]

[0033] According to the present invention, a more accurate search can be performed. [Brief explanation of the drawings]

[0034] [Figure 1] 1 is a diagram showing the overall configuration of a case search system according to an embodiment of the present invention. [Figure 2] 2 is a flowchart showing an imaging control process executed by a control unit of the imaging console of FIG. 1. [Figure 3] 10 is a flowchart showing a case study process executed by a control unit of the diagnostic console of FIG. [Figure 4] FIG. 10 is a diagram for explaining a method for dividing a dynamic image into a plurality of frame image groups. [Figure 5] 10 is a flowchart showing a case search process executed by a control unit of the diagnostic console of FIG. [Figure 6] 2 is a diagram showing an example of a search screen displayed on a display unit of the diagnostic console of FIG. 1. FIG. [Figure 7] 10 is a flowchart showing a modified example of a case search process executed by a control unit of the diagnostic console of FIG. 1. DETAILED DESCRIPTION OF THE INVENTION

[0035] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings, but the scope of the invention is not limited to the illustrated examples.

[0036] [Configuration of case search system 100] FIG. 1 shows the overall configuration of a case search system 100 according to this embodiment. 1, the case retrieval system 100 is configured by connecting an imaging device 1 and an imaging console 2 via a communication cable or the like, and by connecting the imaging console 2 and a diagnostic console 3 serving as a case retrieval device via a communication network NT such as a LAN (Local Area Network). Each device constituting the case retrieval system 100 conforms to the DICOM (Digital Image and Communications in Medicine) standard, and communication between the devices is performed in accordance with DICOM.

[0037] [Configuration of the imaging device 1] The imaging device 1 is an imaging means for capturing periodic dynamics of a subject, such as changes in the shape of lung expansion and contraction due to breathing, or heartbeat. Dynamic imaging refers to obtaining multiple images by repeatedly irradiating a subject with pulsed radiation such as X-rays at predetermined time intervals (pulsed irradiation) or by continuously irradiating the subject with a low dose rate without interruption (continuous irradiation). In other words, dynamic imaging refers to continuous radiographic imaging of the dynamics of a periodic target area along a time axis. Note that dynamic imaging may be performed using ultrasound or magnetism in addition to radiation such as X-rays. Dynamic imaging includes video imaging, but does not include capturing still images while displaying the video. A series of images obtained by dynamic imaging is called dynamic images. Dynamic images can also be acquired by imaging using a semiconductor image sensor such as an FPD (Flat Panel Detector). Dynamic images include moving images, but do not include images obtained by capturing still images while displaying moving images. Each of the multiple images constituting a dynamic image is called a frame image. In the following embodiment, a case where dynamic imaging is performed using pulse irradiation will be described as an example. In the following embodiment, a case where the subject M is the chest of a subject will be described as an example, but the present invention is not limited to this.

[0038] The radiation source 11 is disposed at a position facing the radiation detection unit 13 across the subject M, and irradiates the subject M with radiation (X-rays) under the control of the radiation irradiation control device 12. The radiation irradiation control device 12 is connected to the imaging console 2 and controls the radiation source 11 to perform radiation imaging based on radiation irradiation conditions input from the imaging console 2. The radiation irradiation conditions input from the imaging console 2 include, for example, a pulse rate, a pulse width, a pulse interval, the number of imaging frames per imaging, the value of the X-ray tube current, the value of the X-ray tube voltage, and the type of additional filter. The pulse rate is the number of radiation irradiations per second and corresponds to the frame rate described below. The pulse width is the radiation irradiation time per radiation irradiation. The pulse interval is the time from the start of one radiation irradiation to the start of the next radiation irradiation and corresponds to the frame interval described below.

[0039] The radiation detection unit 13 is composed of a semiconductor image sensor such as an FPD. The FPD has, for example, a glass substrate or the like, and a plurality of detection elements (pixels) are arranged in a matrix at predetermined positions on the substrate. The detection elements (pixels) detect radiation emitted from the radiation source 11 and transmitted through at least the subject M according to its intensity, and convert the detected radiation into an electrical signal and store it. Each pixel is configured with a switching unit such as a TFT (Thin Film Transistor). FPDs can be of an indirect conversion type, in which X-rays are converted into an electrical signal by a photoelectric conversion element via a scintillator, or a direct conversion type, in which X-rays are directly converted into an electrical signal, and either type may be used. In this embodiment, the pixel values ​​(signal values) of the image data generated by the radiation detection unit 13 are density values, and the greater the amount of transmitted radiation, the higher the density value. The radiation detection unit 13 is disposed opposite the radiation source 11 with the subject M interposed therebetween.

[0040] The reading control device 14 is connected to the radiography console 2. The reading control device 14 controls the switching units of each pixel of the radiation detection unit 13 based on the image reading conditions input from the radiography console 2, switches the reading of the electrical signals accumulated in each pixel, and acquires image data by reading the electrical signals accumulated in the radiation detection unit 13. This image data is a frame image. The reading control device 14 then outputs the acquired frame image to the radiography console 2. The image reading conditions include, for example, the frame rate, frame interval, pixel size, image size (matrix size), etc. The frame rate is the number of frame images acquired per second and coincides with the pulse rate. The frame interval is the time from the start of acquisition of one frame image to the start of acquisition of the next frame image and coincides with the pulse interval.

[0041] The radiation irradiation control device 12 and the reading control device 14 are connected to each other and exchange synchronization signals with each other to synchronize the radiation irradiation operation and the image reading operation.

[0042] [Configuration of imaging console 2] The imaging console 2 outputs radiation irradiation conditions and image reading conditions to the imaging device 1 to control the radiation imaging and radiation image reading operations by the imaging device 1, and also displays dynamic images acquired by the imaging device 1 so that the imaging technician or other person performing the imaging can check the positioning and whether the images are suitable for diagnosis. As shown in FIG. 1, the radiography console 2 comprises a control unit 21, a storage unit 22, an operation unit 23, a display unit 24, and a communication unit 25, and each unit is connected by a bus .

[0043] The control unit 21 is composed of a CPU (Central Processing Unit), RAM (Random Access Memory), etc. In response to an operation of the operation unit 23, the CPU of the control unit 21 reads out a system program and various processing programs stored in the storage unit 22 and loads them into the RAM, and executes various processes including an imaging control process described below in accordance with the loaded programs, thereby centrally controlling the operation of each unit of the imaging console 2 and the radiation irradiation operation and reading operation of the imaging device 1.

[0044] The storage unit 22 is configured with a non-volatile semiconductor memory, a hard disk, etc. The storage unit 22 stores various programs executed by the control unit 21, parameters required for executing processes by the programs, data such as processing results, etc. For example, the storage unit 22 stores a program for executing the imaging control process shown in FIG. 2. The storage unit 22 also stores radiation irradiation conditions and image reading conditions in association with the imaging region. The various programs are stored in the form of readable program code, and the control unit 21 sequentially executes operations in accordance with the program code.

[0045] The operation unit 23 is configured with a keyboard having cursor keys, numeric input keys, various function keys, etc., and a pointing device such as a mouse, and outputs instruction signals input by operating the keys on the keyboard or the mouse to the control unit 21. The operation unit 23 may also have a touch panel on the display screen of the display unit 24, and in this case, outputs instruction signals input via the touch panel to the control unit 21.

[0046] The display unit 24 is composed of a monitor such as an LCD (Liquid Crystal Display) or CRT (Cathode Ray Tube), and displays input instructions and data from the operation unit 23 according to instructions of a display signal input from the control unit 21.

[0047] The communication unit 25 includes a LAN adapter, a modem, a TA (Terminal Adapter), etc., and controls data transmission and reception between each device connected to the communication network NT.

[0048] [Configuration of Diagnostic Console 3] The diagnostic console 3 acquires dynamic images from the imaging console 2. The diagnostic console 3 displays the acquired dynamic images, analyzes the acquired dynamic images to generate dynamic analysis images, and searches for and outputs similar case images that are similar to the acquired dynamic images or case candidates related to the acquired dynamic images. Thus, the diagnostic console 3 is a case search device that supports doctors in making diagnoses. As shown in FIG. 1, the diagnostic console 3 comprises a control unit 31, a storage unit 32, an operation unit 33, a display unit , and a communication unit , and each unit is connected by a bus .

[0049] The control unit 31 is composed of a CPU, RAM, etc. In response to an operation of the operation unit 33, the CPU of the control unit 31 reads out a system program and various processing programs stored in the storage unit 32, expands them in the RAM, and executes various processes in accordance with the expanded programs. The CPU of the control unit 31 also reads out a program 32a stored in the storage unit 32, expands it in the RAM, and executes a case learning process and a case search process (described later) in accordance with the expanded program 32a. Furthermore, the control unit 31 performs a search using the dynamic image and outputs similar case images that are similar to the dynamic image or case candidates related to the dynamic image. Here, the control unit 31 functions as a control unit.

[0050] The storage unit 32 is configured with a non-volatile semiconductor memory, a hard disk, etc. The storage unit 32 stores a program 32a for executing case learning processing and case search processing in the control unit 31, various programs, parameters required for executing processing by the programs, data such as processing results, etc. These various programs are stored in the form of readable program code, and the control unit 31 sequentially executes operations in accordance with the program code.

[0051] The storage unit 32 also stores dynamic images previously acquired by dynamic imaging and dynamic analysis images obtained by analyzing dynamic images, in association with an identification ID for identifying the dynamic images or dynamic analysis images, basic patient information, additional patient information, examination information, information on image features focused on in diagnosis, diagnostic results including disease names, medical record information (chief complaints, objective information, etc.), medical history, and label information for bookmarks, conferences, etc. Here, the diagnostic results are the results of diagnosing the dynamic images or dynamic analysis images, such as information entered by a doctor after diagnosing an image, information on a definitive diagnosis obtained by a pathological examination, and information resulting from automatic analysis of medical images using CAD or the like. The storage unit 32 also stores dynamic images and dynamic analysis images obtained by imaging a plurality of patients. Furthermore, the storage unit 32 may store information on a region of interest that a doctor focused on in diagnosis in association with previously acquired dynamic images and dynamic analysis images. Furthermore, the storage unit 32 stores a group of frame images for one dynamic cycle included in the dynamic image or dynamic analysis image, in association with feature amounts calculated from the group of frame images based on information about image features focused on in diagnosis, and information about groups determined by machine learning based on the feature amounts. Here, a group is a group of frame images included in the dynamic image or dynamic analysis image divided into several groups based on a predetermined criterion (for example, feature amounts within the frame images).

[0052] Here, information about image features focused on in diagnosis will be described. When making a diagnosis based on dynamic images or dynamic analysis images generated from the dynamic images, a doctor will focus on image features, such as a long exhalation time compared to an inhalation time, a long breathing time, little concentration change, or poor diaphragm movement, to make the diagnosis. Therefore, when the control unit 31 of the diagnostic console 3 displays the dynamic images or the dynamic analysis images on the display unit 34, it also displays a user interface for inputting or specifying information about the image features focused on by the doctor. Information about the image features input or specified through this user interface by the operation unit 33 is stored in the memory unit 32 in association with the dynamic images. In this embodiment, when the diagnostic target is ventilation, the ratio (or difference) between expiration time and inspiration time, breathing time, concentration change amount, diaphragm movement amount, or average change amount of concentration or diaphragm movement amount during expiration and inspiration can be input or specified as the image feature of interest.When the diagnostic target is pulmonary blood flow, the time of one cycle, concentration change amount, average change amount of maximum to minimum value (or minimum to maximum value) of concentration change within one cycle, etc. can be input or specified as the image feature of interest. It is assumed that the past dynamic images stored are dynamic images made up of a group of frame images for one dynamic period used for diagnosis.

[0053] The operation unit 33 is configured with a keyboard having cursor keys, numeric input keys, various function keys, etc., and a pointing device such as a mouse, and outputs instruction signals input by operating the keys on the keyboard or the mouse to the control unit 31. The operation unit 33 may also have a touch panel on the display screen of the display unit 34, and in this case, outputs instruction signals input via the touch panel to the control unit 31.

[0054] The display unit 34 is configured with a monitor such as an LCD or CRT, and performs various displays according to instructions of a display signal input from the control unit 31.

[0055] The communication unit 35 includes a LAN adapter, a modem, a TA, etc., and controls data transmission and reception between each device connected to the communication network NT.

[0056] Here, a dynamic analysis image generated by analyzing the dynamic image will be described. Dynamic analysis images are images generated by performing analytical processing on dynamic images, such as blood flow analysis images that perform dynamic analysis of blood flow function, ventilation analysis images that perform dynamic analysis of ventilation function, or adhesion analysis images that perform dynamic analysis of adhesions. The analysis process for the dynamic analysis image may include, for example, a time-domain frequency filter process. For example, if the diagnostic target is ventilation, the density changes in the frame images are subjected to a time-domain low-pass filter process (e.g., a cutoff frequency of 0.85 Hz) to generate a dynamic analysis image that extracts density changes due to ventilation. For example, if the diagnostic target is pulmonary blood flow, the frame images are subjected to a time-domain high-pass filter process (e.g., a cutoff frequency of 0.85 Hz) to generate a dynamic analysis image that extracts density changes due to pulmonary blood flow. Note that the density changes in the frame images may also be filtered using a band-pass filter (e.g., a low-pass cutoff frequency of 0.8 Hz and a high-pass cutoff frequency of 2.4 Hz) to extract density changes due to pulmonary blood flow.

[0057] The analysis process may involve matching pixels at the same position in each frame image of the frame image group with each other and performing frequency filtering in the time direction on a pixel-by-pixel basis, or it may involve dividing each frame image of the frame image group into small regions consisting of multiple pixels, calculating a representative value (e.g., average, median, etc.) of the density value of each divided small region, matching the divided small regions between frame images (e.g., matching small regions at the same pixel position), and performing frequency filtering in the time direction on a small region-by-small region basis.

[0058] In addition, a representative value (e.g., variance value) in the time direction may be calculated for each pixel (or each small region) of the frame image group that has undergone analysis processing, and a single image with the calculated value as the pixel value may be generated as a dynamic analysis image.

[0059] [Operation of the case search system 100] Next, the operation of the case retrieval system 100 will be described.

[0060] (Operation of imaging device 1 and imaging console 2) First, the imaging operation performed by the imaging device 1 and the imaging console 2 will be described. 2 shows an imaging control process executed in the control unit 21 of the imaging console 2. The imaging control process is executed by the control unit 21 in cooperation with a program stored in the storage unit 22.

[0061] First, the control unit 21 accepts input of the subject's basic patient information (patient's name, height, weight, age, sex, etc.) and examination information (imaged area (here, chest), type of diagnostic target (ventilation, pulmonary blood flow, etc.)) by the person performing the imaging via the operation unit 23 of the imaging console 2 (step S1).

[0062] Next, the control unit 21 reads out the radiation irradiation conditions from the storage unit 22 and sets them in the radiation irradiation control device 12, and reads out the image reading conditions from the storage unit 22 and sets them in the reading control device 14 (step S2).

[0063] Next, the control unit 21 determines whether or not the radiographer has issued an instruction to irradiate radiation via the operation unit 23 (step S3). Here, the radiographer positions the subject M by placing it between the radiation source 11 and the radiation detection unit 13. The radiographer also instructs the subject to relax and encourages quiet breathing. Alternatively, the radiographer may guide the subject to take deep breaths by, for example, instructing the subject to "inhale and exhale." Furthermore, when the diagnostic target is pulmonary blood flow, for example, it is easier to extract image features by performing an image capture while the subject is holding their breath, so the radiographer may instruct the subject to hold their breath. When radiographing preparations are complete, the radiographer operates the operation unit 23 to input an instruction to irradiate radiation.

[0064] When the person performing the imaging operation inputs a radiation irradiation instruction via the operation unit 23 (step S3; YES), the control unit 21 outputs an imaging start instruction to the radiation irradiation control device 12 and the reading control device 14 to start dynamic imaging (step S4). That is, the radiation source 11 irradiates radiation at pulse intervals set in the radiation irradiation control device 12, and the radiation detection unit 13 acquires frame images.

[0065] When the predetermined number of frames have been captured, the control unit 21 outputs an instruction to end the capture to the radiation irradiation control device 12 and the reading control device 14, and stops the capture operation. The number of frames to be captured is the number that can capture at least one respiratory cycle, or the number that can capture m respiratory cycles (m>1, m is an integer).

[0066] Next, the control unit 21 stores the frame images acquired by shooting in the storage unit 22 in association with numbers (frame numbers) indicating the shooting order (step S5). Next, the control unit 21 displays the frame images acquired by imaging on the display unit 24 (step S6). The imaging operator checks the positioning and the like using the displayed dynamic image, and determines whether an image suitable for diagnosis has been acquired by imaging (imaging OK) or whether reimaging is necessary (imaging NG).

[0067] Next, the control unit 21 receives an input of whether or not the photographing is OK by the person performing the photographing via the operation unit 23, and determines whether or not a determination result indicating that the photographing is OK has been input (step S7). If it is determined that a judgment result indicating that imaging is OK has been input (step S7; YES), the control unit 21 attaches information such as an identification ID for identifying the dynamic image, basic patient information, examination information, radiation irradiation conditions, image reading conditions, and a number indicating the imaging order (frame number) to each of the series of frame images acquired by dynamic imaging in step S4 (for example, writes this in the header area of ​​the image data in DICOM format), and transmits the attached information to the diagnostic console 3 via the communication unit 25 (step S8). Then, the control unit 31 terminates this process. The control unit 31 of the diagnostic console 3 receives the series of frame images acquired by dynamic imaging via the communication unit 35 and stores them in the memory unit 32. On the other hand, if it is determined that a determination result indicating that photography is NG has been input (step S7; NO), the control unit 21 deletes the series of frame images stored in the storage unit 22 (step S9) and ends this process. In this case, photography will need to be retaken.

[0068] (Diagnostic console 3 operation) The operation of the diagnostic console 3 will now be described. First, the case learning process shown in FIG. 3 will be described. The control unit 31, for example, works in cooperation with a program 32a stored in the memory unit 32 to execute a case learning process, thereby learning group classifications for dynamic images and dynamic analysis images stored in the memory unit 32, with the disease name being the correct answer. The learning of group classification in which disease names are used as correct answers for dynamic images and dynamic analysis images is not limited to the example shown in FIG. 3, and may be performed by other methods.

[0069] The flow of the case learning process will be described below with reference to FIG. First, the control unit 31 acquires a dynamic image or a dynamic analysis image to be used for learning from the storage unit 32 (step S11). Next, the control unit 31 divides the acquired dynamic image or dynamic analysis image into a group of frame images for each dynamic cycle (step S12). The division in step S12 uses, for example, density changes throughout the image. For example, a representative density value (e.g., average, median, etc.) is calculated for each frame image of the dynamic image, and the calculated representative density values ​​are plotted in chronological order (in the order of the frame images) as shown in Fig. 4 to determine the waveform of density changes. The dynamic image is divided into a group of frame images corresponding to one dynamic cycle of the subject M by dividing the image at frame images with extreme values ​​(maximum or minimum values). Alternatively, a target region (e.g., lung region) may be extracted from the dynamic image, and the dynamic image may be divided into a group of frame images corresponding to one dynamic cycle using density changes within the extracted region. For example, if the diagnostic target is ventilation, it is preferable to perform segmentation after subjecting the concentration changes to a time-domain low-pass filter (e.g., a cutoff frequency of 0.85 Hz). This removes high-frequency signal changes such as those due to pulmonary blood flow, allowing for accurate extraction of concentration changes due to ventilation. Furthermore, for example, when the diagnostic target is pulmonary blood flow, it is preferable to perform high-pass filtering (e.g., cutoff frequency 0.85 Hz) on the concentration changes in the time direction before segmentation. This allows low-frequency signal changes due to ventilation, etc. to be removed, and concentration changes due to pulmonary blood flow to be extracted with high accuracy. Alternatively, concentration changes due to pulmonary blood flow can be extracted using a band-pass filter (e.g., low-frequency cutoff frequency 0.8 Hz, high-frequency cutoff frequency 2.4 Hz).

[0070] Furthermore, when the diagnostic target is ventilation, the image may be divided into a plurality of frame image groups using changes in the amount of movement of the diaphragm. For example, the diaphragm is recognized in each frame image of a dynamic image, the y coordinate of a certain x coordinate position on the recognized diaphragm is calculated, and the distance between the calculated y coordinate and a reference y coordinate (for example, the distance from the y coordinate of the resting expiratory position (or the distance between the calculated y coordinate and the apex of the lung)) is plotted in time series to obtain a waveform of the change in the amount of movement of the diaphragm over time, and the image is divided at frame images with extreme values ​​(maximum or minimum values). This divides the dynamic image into frame image groups (frame image group 1 to frame image group n (n>1, n is an integer)) for each dynamic cycle of the subject. Here, the left-right direction of each frame image is defined as the x direction, and the up-down direction is defined as the y direction. The diaphragm can be recognized, for example, by recognizing a lung field region from a frame image and recognizing the contour of the lower part of the recognized lung field region as the diaphragm. Any method can be used to extract the lung field region. For example, a threshold is determined by discriminant analysis from a histogram of the signal values ​​of each pixel in the frame image from which the lung field region is to be recognized, and regions with signals higher than this threshold are primarily extracted as lung field region candidates. Next, edge detection is performed near the boundary of the primarily extracted lung field region candidate, and the boundary of the lung field region can be extracted by extracting the point along the boundary where the edge is maximum in a small region near the boundary.

[0071] Next, the control unit 31 acquires from the storage unit 32 image feature information stored in association with the dynamic image or dynamic analysis image acquired in step S11. The control unit 31 then calculates image feature quantities R1 to Rn for each of the frame image groups 1 to n divided in step S12 based on the acquired image feature information (step S13). In calculating the image feature quantities R1 to Rn, various parameters such as histogram, gray value, pixel average, center of gravity, entropy, edge, and contrast of the frame image groups 1 to n may be used. Furthermore, the amount of change (time-course graph) of multiple biological sites obtained from the dynamic analysis images may also be used.

[0072] As described above, when the diagnostic target is ventilation, the image features include the ratio (or difference) between the expiration time and the inspiration time, the breathing time, the amount of concentration change, the amount of diaphragm movement, and the average change in concentration or amount of diaphragm movement during expiration and inspiration.When the diagnostic target is pulmonary blood flow, the image features include the time of one cycle, the amount of concentration change, and the average change in concentration from maximum to minimum (or minimum to maximum) within one cycle.

[0073] The above-mentioned image feature amounts R1 to Rn can be calculated based on the density change or the amount of movement of the diaphragm in the frame image group. The ratio of the expiration time to the inspiration time can be obtained by calculating the time it takes for the concentration or the amount of movement of the diaphragm in the frame image group to go from a maximum to a minimum to obtain the expiration time, and by calculating the time it takes for the concentration or the amount of movement of the diaphragm in the frame image group to go from a minimum to a maximum to obtain the inspiration time, and then calculating the ratio between the two. The breathing time can be obtained by adding the expiration time and the inspiration time. The amount of density change can be found by calculating the amplitude value of the density change in the frame image group. The amount of movement of the diaphragm can be found by calculating the amplitude value of the amount of movement of the diaphragm in the group of frame images. The time for one cycle of pulmonary blood flow can be found by calculating the time it takes for the density of the frame images to go from a maximum (minimum) to the next maximum (minimum).

[0074] When the diagnostic target is ventilation, it is preferable to calculate the feature quantities R1 to Rn after applying a time-domain low-pass filter (e.g., a cutoff frequency of 0.85 Hz) to the density changes in each frame image group. This removes high-frequency signal changes such as those due to pulmonary blood flow, allowing for accurate extraction of density changes due to ventilation. Furthermore, when the diagnostic target is pulmonary blood flow, it is preferable to calculate the feature quantities R1 to Rn after applying a high-pass filter process (e.g., a cutoff frequency of 0.85 Hz) to the density changes in each frame image group in the time direction. This allows low-frequency signal changes due to ventilation, etc. to be removed, and density changes due to pulmonary blood flow to be extracted with high accuracy. Alternatively, density changes due to pulmonary blood flow may be extracted using a band-pass filter (e.g., a low-frequency cutoff frequency of 0.8 Hz and a high-frequency cutoff frequency of 2.4 Hz). Furthermore, by extracting a lung field area from each frame image and then calculating density changes using pixels within that area, it becomes possible to calculate the feature amounts R1 to Rn relating to ventilation and pulmonary blood flow with higher accuracy.

[0075] Next, the control unit 31 determines whether or not a region of interest in the dynamic image or dynamic analysis image has been set in advance (step S14). If no region of interest is set (step S14; NO), the control unit 31 performs machine learning based on the feature amounts R1 to Rn calculated in step S13, thereby learning group classification (step S15). A known machine learning model may be used for the machine learning in step S15. Next, the control unit 31 classifies the dynamic image or dynamic analysis image acquired in step S11 into groups based on the group classification learned in step S15 (step S16). Next, the control unit 31 associates the group classification information, feature amounts R1 to Rn, and other information with the dynamic image or dynamic analysis image classified in step S16, stores them in the storage unit 32 (step S17), and ends the process. The other information includes an identification ID stored in association with the dynamic image or dynamic analysis image, basic patient information, additional patient information, examination information, information on image features focused on in diagnosis, disease name, medical record information (chief complaint, objective information, etc.), medical history, label information for bookmarks, conferences, etc.

[0076] Furthermore, if a region of interest has been set (step S14; YES), the control unit 31 calculates the feature amounts R1r to Rnr of the image features in the region of interest r of the frame image groups 1 to n (step S18). Next, the control unit 31 performs machine learning based on the feature amounts R1r to Rnr calculated in step S18, thereby learning group classification (step S19). The machine learning in step S19 may use a known machine learning model, similar to step S15. Next, the control unit 31 classifies the dynamic image or dynamic analysis image acquired in step S11 into groups based on the group classification learned in step S19 (step S20). Next, the control unit 31 associates the group classification information, feature amounts R1r to Rnr, and other information with the dynamic image or dynamic analysis image classified in step S20, and stores them in the storage unit 32 (step S21), and the process ends.

[0077] The control unit 31 also executes the case learning process using dynamic images and dynamic analysis images obtained by imaging multiple patients. Also, when learning group classification for dynamic images and dynamic analysis images by a method other than the case learning process, the control unit 31 uses dynamic images and dynamic analysis images obtained by imaging multiple patients. Furthermore, in the case learning process described above, the machine learning model is trained on dynamic images or dynamic analysis images, but this is not limiting. Still images may also be trained in addition to dynamic images or dynamic analysis images. Furthermore, the groups in the group classification learned in steps S15 and S19 of the case study process may have different groupings for each feature, such as grouping by disease name or grouping by further classifying diseases into types I, II, III, IV, etc. Furthermore, in learning group classification using disease names as the correct answer, ideally, it is desirable to classify dynamic images or dynamic analysis images associated with information on a single disease name into one group, but dynamic images or dynamic analysis images associated with information on a single disease name may be divided and belong to multiple groups.

[0078] Next, the case search process shown in FIG. 5 will be described. When a user selects a search target image on a search screen 341 shown in Fig. 6 (described later) via the operation unit 33 of the diagnostic console 3 and issues an instruction to execute a search, a case search process is executed in cooperation with the control unit 31 and the program 32a stored in the storage unit 32. Here, the search target image is a dynamic image used for the search, and more specifically, is an image that is undiagnosed and that the user intends to diagnose, among the dynamic images or dynamic analysis images stored in the storage unit 32. Furthermore, when the user selects an image to be searched, the user may specify a region of interest on the search screen 341, which is a region that the user focuses on in diagnosis.

[0079] The flow of the case search process will be described below with reference to FIG. First, the control unit 31 acquires a dynamic image or a dynamic analysis image, which is a search target image selected by the user, from the storage unit 32 (step S31). Step S31 is an acquisition step. Next, the control unit 31 divides the search target image acquired in step S31 into a group of frame images for each dynamic period (step S32). The division in step S32 may be performed in the same manner as in step S12 of the case learning process. Next, the control unit 31 calculates image feature quantities R1 to Rn for each of the frame image groups 1 to n divided in step S32 (step S33). The calculation of the feature quantities R1 to Rn in step S33 may use various parameters such as histograms, gray values, pixel averages, centroids, entropy, edges, and contrasts for the frame image groups 1 to n, as in step S13 of the case study processing. Furthermore, the amount of change (time-course graph) of multiple biological sites obtained from dynamic analysis images may also be used.

[0080] Next, the control unit 31 determines whether or not a region of interest has been set in the image to be searched (step S34). If a region of interest has not been set (step S34; NO), the control unit 31 determines the group to which the image to be searched belongs based on the features R1 to Rn calculated in step S33, for example, using a machine learning model that has been trained in the case study process (step S35). Next, the control unit 31 compares the feature Ry of the frame image group y constituting the dynamic images or dynamic analysis images that belong to the group determined in step S35 and that have been learned in the case learning process, for example, with the feature Rn calculated in step S33. Then, the control unit 31 sets the dynamic images or dynamic analysis images constituted by the frame image group y corresponding to the feature Ry that is closest in distance to the feature Rn in the feature space as display candidates in order of closest distance, that is, in order of similarity (step S36). Here, the dynamic images or dynamic analysis images that belong to the group determined in step S35 are set as similar case images that are similar to the search target image. Next, the control unit 31 retrieves the diagnosis results stored in association with the similar case images from the storage unit 32 as case candidates, and references the disease name from the diagnosis results. The control unit 31 then displays on the display unit 34 the search target image, similar case images, and the diagnosis results of the similar case images as case candidates related to the search target image in order of similarity for each referenced disease name (step S37), and ends the process. That is, the control unit 31 outputs similar case images that are similar to the search target image (dynamic image) and associated with the case candidates, and the case candidates related to the search target image (dynamic image). Here, a similar case image is an image that is output as a case similar to the search target image (dynamic image). The similarity may be determined based on the similarity of the image itself, or based on case information other than the image. Here, steps S36 and S37 are search steps.

[0081] Furthermore, if a region of interest has been set (step S34; YES), the control unit 31 calculates feature amounts R1r to Rnr of the image features in the region of interest r of the frame image groups 1 to n (step S38). Steps S33 and S38 are feature amount calculation steps. Next, the control unit 31 determines the group to which the search target image belongs based on the feature amounts R1r to Rnr calculated in step S38, for example, by using the machine learning model trained in the case learning process (step S39). Next, the control unit 31 proceeds to step S36. In step S36, the control unit 31 compares the feature amount Ryr of the region of interest r of the frame image group y constituting the dynamic image or dynamic analysis image that belongs to the group determined in step S39 and has been learned in the case learning process, for example, with the feature amount Rnr calculated in step S38. The control unit 31 then selects, as display candidates, the dynamic image or dynamic analysis image constituted by the frame image group y corresponding to the feature amount Ryr closest to the feature amount Rnr in the feature amount space, in order of proximity, that is, in order of similarity. Here, the dynamic image or dynamic analysis image that belongs to the group determined in step S39 is selected as a similar case image that is similar to the search target image.

[0082] FIG. 6 shows an example of a search screen 341 that the control unit 31 displays on the display unit 34 of the diagnostic console 3. In the example shown in FIG. 6, a search target image selected by the user is displayed in a field A. Furthermore, if a region of interest in the image to be searched is set in advance, it is marked on the image by region B. Alternatively, if the search screen 341 is configured so that the user can specify a region of interest, the specified region may be displayed as region B. Furthermore, button C is a search button, and the user can press button C to instruct the execution of a search. In addition, in column D, similar case images 342 selected as display candidates in step S36 of the case search process and their diagnosis results 343 are displayed in order of similarity for each disease name. By selecting a disease name via the operation unit 33, the user can refer to similar case images associated with information on the disease name they want to use as reference, and can compare the similar case images with the search target image. In addition, in a field E, a similar case image selected by the user from field D via the operation unit 33 is displayed. In addition, in a column F, the diagnosis result of the similar case image displayed in the column E is displayed. In the example shown in FIG. 6, the column D may not display the diagnosis results of the similar case images, but may display only the similar case images. In addition, columns D and F may display not only the diagnosis results of the similar case images, but also the identification ID associated with the similar case images, basic patient information, additional patient information, test information, information on image features focused on in the diagnosis, disease name, medical record information (chief complaint, objective information, etc.), medical history, label information for bookmarks and conferences, etc. In addition, column G may display an identification ID associated with the image being searched for, basic patient information, additional patient information, examination information, information on image features focused on in diagnosis, disease name, medical record information (chief complaint, objective information, etc.), medical history, label information for bookmarks and conferences, etc.

[0083] <Modification> This modified example will be described below. The configuration and the operations of the imaging device 1 and imaging console 2 in this modified example are similar to those described in the above embodiment, so the explanation will be used to describe the operation of the diagnostic console 3.

[0084] The flow of the case search process of this modified example will be described below with reference to FIG. First, the control unit 31 performs steps S41 to S45, which are similar to steps S31 to S35 of the case search process of the above embodiment. Next, the control unit 31 extracts dynamic images or dynamic analysis images from the dynamic images or dynamic analysis images belonging to the group determined in step S45, for example, from those that have been trained in the case learning process, based on information associated with the search target image or information associated with the trained dynamic images or dynamic analysis images (step S46). Specifically, the control unit 31 extracts dynamic images or dynamic analysis images associated with the same patient basic information as the patient basic information associated with the search target image. The control unit 31 also extracts dynamic images or dynamic analysis images associated with information matching the user's search criteria. Examples of information matching the user's search criteria include an identification ID, patient basic information, patient supplementary information, test information, information on image features focused on in diagnosis, diagnosis results including disease name, medical record information (chief complaint, objective information, etc.), medical history, and label information for bookmarks, conferences, etc. The extracted dynamic images or dynamic analysis images are regarded as similar case images. Next, the control unit 31 compares the feature amount Rz of the frame image group z constituting the dynamic image or dynamic analysis image extracted in step S46 with the feature amount Rn calculated in step S43. Then, the control unit 31 selects, as display candidates, the dynamic image or dynamic analysis image constituted by the frame image group z corresponding to the feature amount Rz closest in distance to the feature amount Rn in the feature amount space, in order of proximity, that is, in order of similarity (step S47). Next, the control unit 31 retrieves the diagnosis results stored in association with the similar case images from the storage unit 32 as case candidates, and references the disease name from the diagnosis results. The control unit 31 then displays on the display unit 34 the search target image, similar case images, and the diagnosis results of the similar case images as case candidates related to the search target image in order of similarity for each referenced disease name (step S48), and ends the process. That is, the control unit 31 outputs similar case images that are similar to the search target image (dynamic image) and associated with the case candidates, and the case candidates related to the search target image (dynamic image). Here, a similar case image is an image that is output as a case similar to the search target image (dynamic image). The similarity may be determined based on the similarity of the image itself, or based on case information other than the image. Here, steps S46 to S48 are search steps.

[0085] Furthermore, if a region of interest has been set (step S44; YES), the control unit 31 performs steps S49 and S50 similar to steps S38 and S39 in the case search process of the above embodiment.

[0086] By performing step S46 of the case search process of the above modified example, similar case images that are candidates for display can be narrowed down, thereby shortening the processing time in steps S47 and S48.

[0087] The search target image and similar case images may be composed of a group of frame images (multiple frame images) constituting a dynamic image, or may be composed of a single frame image. In other words, the search target image and similar case images may be composed of at least one frame image constituting a dynamic image. When the search target image and similar case images are composed of a single frame image, a case search can be performed using one of the multiple frame images constituting a dynamic image, which contains more information than a still image, thereby enabling more accurate case searches. When the search target image and similar case images are composed of multiple frame images, a case search can be performed using multiple of the multiple frame images constituting a dynamic image, which contains more information than a still image, thereby enabling even more accurate case searches. Furthermore, when the search target image and the similar case images are configured from a group of frame images (a plurality of frame images) that form a dynamic image, the group of frame images are consecutive frame images. The search target image and the similar case image may be images of a region of interest within one of the frame images constituting the dynamic image. In this case, by performing a case search on the region of interest set in the frame image, more accurate case search can be performed. Furthermore, the search target image and similar case images may be only some of the frame images constituting the dynamic image. In this case, the case search can be performed using some of the frame images constituting the dynamic image, which contains more information than a still image, thereby enabling more accurate case search. Furthermore, when the search target image and the similar case images are configured from a single frame image that constitutes a dynamic image, the single frame image may be selected by the user. This allows the case search to be performed using the frame image that the user determines to be optimal, thereby enabling more accurate case search.

[0088] In addition, in steps S37 and S48 of the case search process, the search target image, similar case images, and diagnosis results of the similar case images are displayed in order of similarity for each disease name, but this is not limited to this. Only the disease names of the search target image and similar case images may be displayed, or only the search target image and similar case images may be displayed. That is, the control unit 31 displays (outputs) the disease names of similar case images that are similar to the search target image and associated with case candidates or similar case images that are case candidates related to the search target image. At least one of the similar case images and the case candidates related to the search target image may be output. That is, either one or both may be output. Furthermore, the control unit 31 may output multiple case candidates related to the similar case image or the search target image, or may output only one. Outputting multiple case candidates related to the similar case image or the search target image allows the doctor to refer to more information to help with diagnosis. Furthermore, the control unit 31 may display the search target image, similar case images, and diagnostic results of the similar case images in order of similarity, without separating them by disease name.

[0089] Furthermore, although the machine-learned model used in the case search process is trained by the case learning process, this is not limitative and the case search process may be performed using an external machine-learned model. Furthermore, a model obtained by additionally training a machine-learned model using dynamic images captured by the imaging control process or dynamic analysis images generated based on the dynamic images may be used in the case search process. Furthermore, the dynamic images or dynamic analysis images used for additional training may be weighted more than other dynamic images or dynamic analysis images. The weighted dynamic images or dynamic analysis images are trained by the machine learning model so as to be more frequently searched during the case search process. Examples of weighting include a case where a diagnosis based on the output result of the case search process differs from a definitive diagnosis (medical record or pathological diagnosis), and additional training is performed using the definitive diagnosis; a case where additional training is performed using the diagnosis of a particular doctor known to be an expert; or a case where multiple doctors frequently request detailed display of dynamic images or dynamic analysis images, and the diagnosis is likely to be mistaken for that case, and additional training is performed using that case.

[0090] Furthermore, the region of interest that is set in advance in the dynamic image or dynamic analysis image in the above-described embodiment and modified examples may be one or more within the frame image.

[0091] Furthermore, in step S36 of the case search process, dynamic images or dynamic analysis images belonging to the group determined in step S35 are determined as similar case images similar to the search target image. Furthermore, in step S46 of the modified case search process, dynamic images or dynamic analysis images are extracted from the learned dynamic images or dynamic analysis images belonging to the group determined in step S45 based on predetermined information, and the extracted dynamic images or dynamic analysis images are determined as similar case images. However, this is not limited to this. Dynamic images or dynamic analysis images belonging to adjacent groups with close feature distances as well as those belonging to the same group as the search target image may also be determined as similar case images. This makes it possible to search for many dynamic images or dynamic analysis images with similar image features but associated with information on different disease names.

[0092] In the case search process, it is preferable that the dynamic image or dynamic analysis image selected by the user as the image to be searched has a clear lesion area within the image. By using a dynamic image or dynamic analysis image with a clear lesion area in the case search process, a highly accurate search can be performed.

[0093] Furthermore, in the above embodiment and modified examples, the dynamic images are images captured by radiography, ultrasound, magnetism, etc., but it is preferable that the dynamic images used in the case learning process and case search process be images captured using radiography. Because radiography is most commonly used for screening primary examinations in clinics and other facilities that perform dynamic radiography, the volume of previously captured radiographic dynamic images is the largest. Case search can be performed with higher accuracy by using a model trained on previously captured radiographic dynamic images.

[0094] Furthermore, the dynamic analysis images used in the above embodiment and modified examples contain the results of the above-described analysis processing, and therefore contain more information than dynamic images. Therefore, by using the dynamic analysis images to perform the case search processing, a more accurate search can be performed.

[0095] Furthermore, dynamic images and dynamic analysis images obtained by photographing multiple patients are used to train the machine learning model, which has already been trained for group classification and is used in steps S35 and S45 of the case search process. Therefore, in steps S37 and S48 of the case search process, it is possible to display similar case images of patients other than the patient photographed in the search target image, or diagnostic results of other patients as case candidates related to the search target image. In other words, images and diagnostic results of other patients can be output as search results, allowing doctors to refer to more information.

[0096] Furthermore, in the case search process described above, the similar case images are dynamic images or dynamic analysis images, but still images (still images) may also be displayed as similar case images. In addition, the diagnostic console 3 may have a function that allows the user to select whether the similar case images to be output in the case search process are dynamic images or dynamic analysis images, still images, or dynamic images or dynamic analysis images and still images.

[0097] As described above, the control program causes the case search device to perform a case search, and the control program causes the control unit 31 (computer) of the diagnostic console 3 (case search device) to perform a search step (steps S36, S37, S46 to S48) in which the control program causes the control unit 31 (computer) of the diagnostic console 3 (case search device) to search using dynamic images and output similar case images that are similar to the dynamic images or case candidates related to the dynamic images. Therefore, by using dynamic images, which contain more information than still images, case retrieval can be performed using more information, resulting in more accurate case retrieval.

[0098] In addition, the control unit 31 of the diagnostic console 3 has an acquisition step (steps S31, S41) for acquiring a dynamic image, and a feature calculation step (steps S33, S38, S43, S49) for calculating the features of a first image from the dynamic image acquired by the acquisition step, and a search step outputs similar case images that are similar to the dynamic image or case candidates related to the dynamic image based on the features of the first image (features Rn, Rnr) calculated by the feature calculation step and the features of a second image (features Ry, Ryr) calculated in advance by learning multiple frame images of an arbitrary dynamic image. Therefore, it is possible to perform a more accurate case search based on dynamic images captured in the past.

[0099] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic image used for the search is at least one frame image out of the multiple frame images that make up the dynamic image. Therefore, by using frame images that constitute dynamic images that contain more information than still images in case retrieval, it is possible to perform case retrieval with higher accuracy.

[0100] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic image used for the search is an image of a region of interest in at least one frame image out of the multiple frame images that make up the dynamic image. Therefore, it is possible to perform a more accurate case search for the region of interest set in the frame image.

[0101] In the case search executed by the control unit 31 of the diagnostic console 3, the region of interest is a region designated by the user. Therefore, a case search can be performed for a region of interest that the user determines to be optimal, thereby enabling a more accurate case search.

[0102] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic images used for the search are a plurality of frame images that make up the dynamic image. Therefore, by using a plurality of frame images as search target images, a case search can be performed using more information, and therefore a more accurate case search can be performed.

[0103] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic image used for the search is one frame image that constitutes the dynamic image. Therefore, a case search can be performed using one of a plurality of frame images that make up a dynamic image, which contains more information than a still image, thereby enabling a more accurate case search.

[0104] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic image used for the search is an image of a region of interest within one of the frame images that constitute the dynamic image. Therefore, it is possible to perform a more accurate case search for the region of interest set in the frame image.

[0105] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic images used for the search are only some of the frame images among the multiple frame images that make up the dynamic images. Therefore, a case search can be performed using some of the frame images among the multiple frame images that make up a dynamic image, which contains more information than a still image, thereby enabling more accurate case search.

[0106] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic image used for the search is an image selected by the user from among a plurality of frame images that make up the dynamic image. Therefore, the case search can be performed using the image that the user determines to be the most suitable for use in the search, thereby enabling more accurate case search.

[0107] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic images used for the search are consecutive frame images among the multiple frame images that make up the dynamic image. Therefore, by using consecutive frame images as search target images, a case search can be performed using more information, resulting in a more accurate case search.

[0108] In the case search executed by the control unit 31 of the diagnostic console 3, dynamic images are images obtained by continuously radiographing the dynamics of a target area along the time axis. Therefore, by using dynamic images, which contain more information than still images, case retrieval can be performed using more information, resulting in more accurate case retrieval.

[0109] In the case search executed by the control unit 31 of the diagnostic console 3, dynamic images are images obtained by continuously radiographing the dynamics of a periodic target region along the time axis. Therefore, by using dynamic images, which contain more information than still images, case retrieval can be performed using more information, resulting in more accurate case retrieval.

[0110] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the dynamic image used for the search is a dynamic analysis image obtained by analyzing the dynamic image. Therefore, by using a dynamic analysis image with a larger amount of information, a case search can be performed using more information, and therefore a more accurate case search can be performed.

[0111] In addition, in the case search performed by the control unit 31 of the diagnostic console 3, the dynamic analysis image is one of a blood flow analysis image in which the blood flow function is dynamically analyzed, a ventilation analysis image in which the ventilation function is dynamically analyzed, and an adhesion analysis image in which adhesions are dynamically analyzed. Therefore, by using a blood flow analysis image, ventilation analysis image, or adhesion analysis image, which contains more information, case searches can be performed using more information, resulting in more accurate case searches.

[0112] In the case search executed by the control unit 31 of the diagnostic console 3, the similar case images are dynamic images. Therefore, dynamic images, which contain more information than still images, are output as similar case images, so that more information can be obtained as case search results.

[0113] In the case search executed by the control unit 31 of the diagnostic console 3, the similar case image is at least one frame image among the plurality of frame images that make up the dynamic image. Therefore, by outputting frame images that constitute dynamic images, which contain more information than still images, as similar case images, it is possible to obtain more information as case search results.

[0114] In the case search executed by the control unit 31 of the diagnostic console 3, the similar case images are images obtained by continuously radiographing the dynamics of the target area along the time axis. Therefore, dynamic images, which contain more information than still images, are output as similar case images, so that more information can be obtained as case search results.

[0115] In the case search executed by the control unit 31 of the diagnostic console 3, the similar case images are images obtained by continuously radiographing the dynamics of a target region having periodicity along the time axis. Therefore, dynamic images, which contain more information than still images, are output as similar case images, so that more information can be obtained as case search results.

[0116] In the case search executed by the control unit 31 of the diagnostic console 3, the similar case images are dynamic analysis images obtained by analyzing dynamic images. Therefore, by outputting a dynamic analysis image with a larger amount of information as a similar case image, more information can be obtained as a case search result.

[0117] In addition, in the case search performed by the control unit 31 of the diagnostic console 3, the dynamic analysis image is one of a blood flow analysis image in which the blood flow function is dynamically analyzed, a ventilation analysis image in which the ventilation function is dynamically analyzed, and an adhesion analysis image in which adhesions are dynamically analyzed. Therefore, by outputting a blood flow analysis image, ventilation analysis image, or adhesion analysis image with a larger amount of information as a similar case image, it is possible to obtain more information as a case search result.

[0118] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the search step outputs a plurality of similar case images that are similar to the dynamic image or case candidates that are related to the dynamic image. Thus, doctors have more information to refer to and use to aid in diagnosis.

[0119] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the search step causes similar case images similar to the dynamic image or case candidates related to the dynamic image to be output in order of similarity. Therefore, it is possible to refer to similar case images that are more similar to the search target image.

[0120] Furthermore, in the case search executed by the control unit 31 of the diagnostic console 3, the search step causes similar case images similar to the dynamic image or case candidates related to the dynamic image to be output for each disease name. Therefore, more accurate case search can be performed using the disease name as a criterion.

[0121] In the case search executed by the control unit 31 of the diagnostic console 3, the search step outputs similar case images similar to the dynamic image or case candidates related to the dynamic image by disease name and in order of similarity. Therefore, more accurate case searches can be performed using disease names as criteria, and similar case images that are more similar to the search target image can be referenced.

[0122] The description in this embodiment is an example of a suitable case search system according to the present invention, and the present invention is not limited to this.

[0123] For example, in the above embodiment, the present invention has been described as being applied to dynamic images of the chest, but the present invention is not limited to this and may be applied to dynamic images of other parts of the body.

[0124] Furthermore, the storage unit 32 of the diagnostic console 3 of the above embodiment and modified example may store images and diagnostic reports of all diseases diagnosed by image diagnosticians or clinicians, which can be used to educate trainees, students, etc.

[0125] The storage unit 32 of the diagnostic console 3 in the above embodiment and modified example may store an electronic medical book, which is a computerized practical reference book for image diagnosticians. The electronic medical book may be displayed on the display unit 34 so that the user can view it when filling out a diagnostic report. It is also desirable to configure the electronic medical book so that the user can search for keywords.

[0126] In the above description, examples have been disclosed in which a hard disk or a semiconductor nonvolatile memory is used as a computer-readable medium for the program according to the present invention, but the present invention is not limited to these examples. Portable recording media such as CD-ROMs can also be used as other computer-readable media. Furthermore, carrier waves can also be used as a medium for providing data for the program according to the present invention via a communication line.

[0127] In addition, the detailed configuration and detailed operation of each device constituting the case search system 100 may be modified as appropriate without departing from the spirit of the present invention. [Explanation of symbols]

[0128] 100 Case Search System 1. Imaging device 11 Radiation source 12 Radiation exposure control device 13 Radiation detection unit 14 Reading control device 2. Filming console 21 Control section 22 Memory section 23 Control section 24 Display 25 Communications Department 26 Bus 3 Diagnostic console (case search device) 31 Control Unit (Control Unit) 32 Storage section 33 Operation section 34 Display section 35 Communications Department 36 Bus

Claims

1. A control program for causing a case search device to perform a case search, The control program is configured to: an acquisition step of acquiring a dynamic image; a feature amount calculation step of calculating a feature amount of a first image from the dynamic image acquired by the acquisition step; a search step of performing a search using the dynamic image and outputting similar case images that are similar to the dynamic image or case candidates that are related to the dynamic image; The search step determines a group to which the dynamic image belongs by a trained model based on the feature amount of the first image calculated in the feature amount calculation step, and outputs the similar case image from the dynamic image belonging to the determined group based on the feature amount of the first image and the feature amount of a second image calculated from a frame image of the dynamic image belonging to the determined group, or outputs the case candidate from a diagnosis result associated with the dynamic image belonging to the determined group; The trained model is a control program that has trained group classification using disease information as the correct answer based on a second feature calculated in advance from multiple frame images of any dynamic image.

2. A control program as described in claim 1, wherein the information regarding the disease is the name of the disease.

3. 3. The control program according to claim 1, wherein the dynamic image used for the search is at least one frame image among a plurality of frame images that constitute the dynamic image.

4. 4. The control program according to claim 3, wherein the dynamic image used for the search is an image of a region of interest in at least one frame image among a plurality of frame images that constitute the dynamic image.

5. 5. The control program according to claim 4, wherein the region of interest is a region designated by a user.

6. The control program according to claim 3 , wherein the dynamic image used for the search is a plurality of frame images that constitute the dynamic image.

7. 6. The control program according to claim 3, wherein the dynamic image used for the search is one frame image that constitutes the dynamic image.

8. 6. The control program according to claim 3, wherein the dynamic image used for the search is an image of a region of interest within one of the frame images that constitute the dynamic image.

9. 6. The control program according to claim 3, wherein the dynamic image used for the search is only a portion of a plurality of frame images constituting the dynamic image.

10. 10. The control program according to claim 3, wherein the dynamic image used in the search is an image selected by a user from a plurality of frame images that constitute the dynamic image.

11. 6. The control program according to claim 3, wherein the dynamic image used for the search is a sequence of frame images among a plurality of frame images that form the dynamic image.

12. The control program according to claim 1 , wherein the dynamic image is an image obtained by continuously radiographing the dynamic state of a target region along a time axis.

13. The control program according to claim 1 , wherein the dynamic image is an image obtained by continuously radiographing the dynamics of a periodic target region along a time axis.

14. The control program according to claim 1 , wherein the dynamic image used for the search is a dynamic analysis image obtained by analyzing a dynamic image.

15. The control program according to claim 14, wherein the dynamic analysis image is one of a blood flow analysis image obtained by dynamic analysis of blood flow function, a ventilation analysis image obtained by dynamic analysis of ventilation function, and an adhesion analysis image obtained by dynamic analysis of adhesion.

16. The control program according to claim 1 , wherein the similar case image is a dynamic image.

17. The control program according to claim 16 , wherein the similar case image is at least one frame image among a plurality of frame images constituting a dynamic image.

18. 18. The control program according to claim 16, wherein the similar case images are images obtained by continuously radiographing the dynamics of a target region along a time axis.

19. The control program according to claim 16 , wherein the similar case images are images obtained by continuously radiographing the dynamics of a target region having periodicity along a time axis.

20. The control program according to claim 16 , wherein the similar case image is a dynamic analysis image obtained by analyzing a dynamic image.

21. A control program as described in Claim 20, wherein the dynamic analysis image as the similar case image is any one of a blood flow analysis image in which blood flow function is dynamically analyzed, a ventilation analysis image in which ventilation function is dynamically analyzed, and an adhesion analysis image in which adhesions are dynamically analyzed.

22. The searching step The control program according to claim 1 , which outputs a plurality of similar case images similar to the dynamic image or a plurality of case candidates related to the dynamic image.

23. The searching step The control program according to claim 1 , further comprising: outputting similar case images similar to the dynamic image or case candidates related to the dynamic image in order of similarity.

24. The control program according to claim 22 , wherein the search step outputs similar case images similar to the dynamic image or case candidates related to the dynamic image for each disease name.

25. The searching step 25. The control program according to claim 24, wherein similar case images similar to the dynamic image or case candidates related to the dynamic image are output for each disease name in order of similarity.

26. Acquire a dynamic image, Calculating a feature amount of a first image from the acquired dynamic image; a control unit that performs a search using the dynamic image and outputs a similar case image that is similar to the dynamic image or a case candidate that is related to the dynamic image; the control unit determines a group to which the dynamic image belongs using a trained model based on the feature amount of the first image, and outputs the similar case image from the dynamic image belonging to the determined group based on the feature amount of the first image and the feature amount of a second image calculated from a frame image of the dynamic image belonging to the determined group, or outputs the case candidate from a diagnosis result associated with the dynamic image belonging to the determined group; The trained model is a case search device that has been trained to perform group classification using disease information as the correct answer based on second features calculated in advance from multiple frame images of any dynamic image.

Images