Medical image processing apparatus and computer-readable non-temporary recording medium

Abstract

The present invention addresses the problem of providing a medical image processing device and a computer-readable non-transitory recording medium that, when a learning model is used to infer a medical image, make the inference result using the learning model close to the user's request. The learning model (30) learns so as to output, on the basis of input data representing an ultrasound image that is a medical image, output data representing an inference result relating to the ultrasound image. An inference parameter changing unit (36) changes an inference parameter, which is a parameter to be adjusted in the learning process of the learning model (30) and which affects the output data, in accordance with an instruction from a user. The inference processing unit (34) outputs an inference result relating to a target ultrasound image on the basis of output data when the target ultrasound image to be inferred is input to the learning model (30) having the changed inference parameters.

Description

Technical Field

[0001] This specification discloses improvements to a medical image processing device and a medical image processing program. Background Art

[0002] Medical images are mainly used for the diagnosis or treatment of diseases, and are images that represent the internal structure or function of a subject. Examples of medical images include X-ray images, computed tomography (CT) images, magnetic resonance images (MRI), and ultrasonic images. On the other hand, there are known (machine) learning models that adjust internal parameters by performing learning processing using learning data. A fully learned learning model can predict (infer) data corresponding to each of the various input data with high accuracy, and output its inference results as output data.

[0003] Conventionally, it has been proposed to incorporate a learning model into a medical image processing apparatus for processing medical images.

[0004] For example, Patent Document 1 discloses an ultrasonic image forming device that constructs a learned model for learning the correlation between ultrasonic image data and an evaluation value of the ultrasonic image data, and changes parameters related to the acquisition of ultrasonic image data when the evaluation value output by the learned model to which the ultrasonic image data to be processed has been input is less than a threshold value.

[0005] Patent Document 1: Japanese Patent Application Publication No. 2020-92739 Summary of the invention

[0006] Consider learning a learning model so as to output output data representing an inference result related to a medical image. An example of a learning model that outputs an inference result related to a medical image is a learning model that infers the contour of the heart (inner cavity) when the heart is shown in a medical image.

[0007] At this time, the learning model uses a large number of medical images as learning data to learn, thereby setting its internal parameters (referred to as "inference parameters" in this specification). However, the inference results of the learned learning model are sometimes not completely consistent with the results required by a specific user. For example, using the above example, when a user wants to roughly know the outline of the heart, the learning model outputs the outline of the heart in considerable detail, and as a result, sometimes its smoothness is lost.

[0008] An object of the medical image processing apparatus disclosed in this specification is to provide a medical image processing apparatus that can make an inference result using a learning model close to a user's request when inferring a medical image using a learning model.

[0009] The medical image processing device disclosed in the present specification is characterized in that it comprises: a learning model that learns in a manner that outputs output data representing an inference result related to the medical image based on input data representing the medical image; an inference parameter changing unit that changes an inference parameter based on a user's instruction, wherein the inference parameter is a parameter that is an adjustment object in the learning process of the learning model and is a parameter that affects the output data; and an inference processing unit that outputs an inference result related to the object medical image based on the output data when the object medical image that is a medical image to be inferred and processed is input to the learning model with the changed inference parameter.

[0010] Preferably, the device further comprises a display control unit that displays the target medical image and an inference result related to the target medical image on a display, and changes display content of the inference result on the display according to a change in the inference parameter.

[0011] Preferably, when a user corrects the inference result related to the target medical image displayed on the display, the inference parameter changing unit may change the inference parameter according to the corrected inference result.

[0012] Preferably, it is as follows: there is also a parameter change history database, in which the changed inference parameters changed according to the user's instructions are stored, and when the object medical image is inferred according to the instructions from the user, the inference parameter changing unit changes the inference parameters according to the changed inference parameters stored in the parameter change history database without relying on the instructions from the user.

[0013] Preferably, it is as follows: in the parameter change history database, a user identifier that identifies a user who has indicated a change in the inference parameter is associated with and stored in the changed inference parameter that has been changed according to the instruction of the user; when the inference parameter related to the user is stored in the parameter change history database, when inference processing is performed on the object medical image according to the instruction from the user, the inference parameter changing unit changes the inference parameter according to the changed inference parameter that is associated with the user identifier of the user in the parameter change history database, without relying on the instruction from the user.

[0014] Preferably, the medical image is an ultrasonic image formed by irradiating ultrasonic waves toward the heart of the subject, and the inference processing unit outputs information for measuring the heart using the ultrasonic image as the target medical image as an inference result of the target medical image.

[0015] Furthermore, the medical image processing program disclosed in the present specification is characterized in that a computer is caused to function as the following parts: a learning model that performs learning in a manner that outputs output data representing an inference result related to the medical image based on input data representing the medical image; an inference parameter changing unit that changes an inference parameter in accordance with a user's instruction, wherein the inference parameter is a parameter that is an adjustment object in the learning processing of the learning model and is a parameter that affects the output data; and an inference processing unit that outputs an inference result related to the object medical image based on the output data when the object medical image that is a medical image to be inferred and processed is input to the learning model having the changed inference parameter.

[0016] Effects of the Invention

[0017] According to the medical image processing apparatus disclosed in this specification, when inference is performed on a medical image using a learning model, it is possible to make the inference result using the learning model close to the user's request. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram of the structure of the ultrasonic diagnostic apparatus according to this embodiment.

[0019] Figure 2 It is a diagram showing a display example of the inference result related to the target ultrasonic image.

[0020] Figure 3 This is a diagram showing an example of a screen for changing inference parameters.

[0021] Figure 4 It is a diagram showing a display example of the inference result using the changed inference parameters.

[0022] Figure 5 This is a conceptual diagram showing an example of the content of the parameter change history DB.

[0023] Figure 6 This is a flowchart showing the flow of processing by the ultrasonic diagnostic apparatus according to the present embodiment.

[0024] Explanation of symbols

[0025] 10-ultrasonic diagnostic device, 12-ultrasonic probe, 14-transceiver unit, 16-signal processing unit, 18-image forming unit, 20-display control unit, 22-display, 24-input interface, 26-control unit, 28-memory, 30-learning model, 32-parameter change history DB, 34-inference processing unit, 36-inference parameter change unit. DETAILED DESCRIPTION

[0026] Figure 11 is a schematic diagram of the configuration of an ultrasonic diagnostic apparatus 10 as a medical image processing apparatus according to the present embodiment. The ultrasonic diagnostic apparatus 10 is a medical apparatus installed in a medical institution such as a hospital.

[0027] The ultrasonic diagnostic apparatus 10 is an apparatus that scans an ultrasonic beam on a subject and generates an ultrasonic image as a medical image based on the received signal obtained thereby. In particular, in the present embodiment, the ultrasonic diagnostic apparatus 10 forms an ultrasonic tomographic image (B-mode image) by converting the amplitude intensity of the reflected wave from the scanning surface into brightness based on the received signal. In addition, the ultrasonic diagnostic apparatus 10 can also form other ultrasonic images such as a Doppler image that shows the movement speed of tissue in the subject formed by the difference (Doppler shift) in frequency between the transmitted wave and the received wave.

[0028] In addition, the transceiver 14, the signal processing unit 16, the image forming unit 18, the display control unit 20, the inference processing unit 34, and the inference parameter changing unit 36 ​​of the ultrasonic diagnostic device 10 are composed of a processor. The processor is composed of at least one of a general-purpose processing device (for example, a CPU (Central Processing Unit), etc.), and a dedicated processing device (for example, a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a programmable logic device, etc.). As a processor, it is not composed of a single processing device, but can be composed of the cooperation of a plurality of processing devices existing in physically separated positions. In addition, the above-mentioned units can also be realized by the cooperation of hardware such as a processor and software.

[0029] The ultrasonic probe 12 is a device that transmits and receives ultrasonic waves to a subject. The ultrasonic probe 12 has a transducer element array composed of a plurality of transducer elements that transmit and receive ultrasonic waves to a subject.

[0030] The transceiver 14 transmits a transmission signal to the ultrasonic probe 12 (specifically, each transducer element of the transducer element array) under the control of the control unit 26 (described later). As a result, ultrasonic waves are transmitted from each transducer element toward the subject.

[0031] Furthermore, the transceiver 14 receives a reception signal from each vibration element that receives the reflected wave from the subject. The transceiver 14 has an adder and a plurality of delays corresponding to each vibration element, and performs a phase-aligned addition process of aligning and adding the phases of the reception signals from each vibration element by the adder and the plurality of delays. Thus, a reception beam signal is formed in which information indicating the signal strength of the reflected wave from the subject is arranged in the depth direction of the subject.

[0032] The signal processing unit 16 performs various signal processing including filtering processing using a bandpass filter and detection processing on the reception beam signal from the transmission and reception unit 14 .

[0033] The image forming unit 18 forms an ultrasonic tomographic image (B-mode image) based on the received beam signal processed by the signal processing unit 16. First, the image forming unit 18 converts the received beam signal into data on the coordinate space of the ultrasonic image. Then, the image forming unit 18 forms an ultrasonic tomographic image based on the coordinate conversion signal.

[0034] The display control unit 20 controls the display 22 to display the ultrasonic tomographic image formed by the image forming unit 18 and other various information. The display 22 is, for example, a liquid crystal display or an organic EL (Electro Luminescence) display.

[0035] The input interface 24 is composed of, for example, buttons, a trackball, a touch panel, etc. The input interface 24 is used to input user commands to the ultrasonic diagnostic apparatus 10. In this embodiment, the display 22 is a touch panel, and the display 22 also functions as the input interface 24.

[0036] The control unit 26 is configured to include at least one of a general-purpose processor (e.g., a CPU (Central Processing Unit), etc.) and a dedicated processor (e.g., a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a programmable logic device, etc.). The control unit 26 may not be composed of a single processing device, but may be composed of a plurality of processing devices that are physically separated. The control unit 26 controls each unit of the ultrasonic diagnostic apparatus 10 according to a medical image processing program stored in a memory 28 described later.

[0037] The memory 28 is configured to include a HDD (Hard Disk Drive), an SSD (Solid State Drive), an eMMC (embedded Multi Media Card), or a ROM (Read Only Memory). The memory 28 stores a medical image processing program to operate each part of the ultrasonic diagnostic apparatus 10. In addition, the medical image processing program can also be stored in a non-temporary storage medium readable by a computer, such as a USB (Universal Serial Bus) memory or a CD-ROM. The ultrasonic diagnostic apparatus 10 can read and execute the medical image processing program from such a storage medium.

[0038] And, if Figure 1 As shown, a learning model 30 is stored in the memory 28. In the learning model 30, data representing a medical image is input as data, and reasoning related to the medical image involved in the input data is performed, and the reasoning result is output. In this specification, an example is described in which the learning model 30 inputs an ultrasonic tomographic image showing a cross section of the heart and outputs information for measuring the heart using the ultrasonic tomographic image as a reasoning result. In particular, in this specification, an example is described in which the learning model 30 infers the contour line of the heart and the characteristic position of the heart (for example, the apex position or the valve ring position, etc.) as information for measuring the heart and outputs the reasoning result. The input data representing the medical image input to the learning model 30 is not limited to the medical image itself. For example, in the case where the medical image is an ultrasonic image, it may be a receiving beam signal before the ultrasonic image is formed. In addition, the medical image is not limited to the ultrasonic image, and may be, for example, an X-ray image, a CT image, or an MRI. In addition, the reasoning content of the learning model 30 may be any content as long as it is related to the input medical image.

[0039] The learning model 30 is, for example, a model using an ASM (Active Shape Model). ASM is a model that expresses the shape of an object (in this embodiment, the contour of the heart) using a collection of points. The ASM incorporated into the learning model 30 in this embodiment calculates the average shape of multiple contours of the heart as learning data, and performs principal component analysis on the contours of the heart to obtain a combination of eigenvalues ​​and eigenvectors of the multiple contours. Then, a weight is applied to each eigenvector. By changing the weight of each eigenvector, the shape of the contour of the heart changes. For example, by increasing the weight of a high-level principal component (for example, the first principal component) and reducing the weight of a low-level principal component, the learning model 30 infers the approximate shape of the contour of the heart (in other words, the contour of the heart becomes smooth). On the other hand, by reducing the weight of a high-level principal component and increasing the weight of a low-level principal component, the learning model 30 emphasizes the details of the inferred contour of the heart. In this way, the weight of each eigenvector is a parameter that affects the output data of the learning model 30.

[0040] The weight of each feature vector in the ASM is a parameter that is an adjustment object in the learning process of the learning model 30. The learning model 30 uses a combination of an ultrasonic image showing a heart and the contour of the heart in the ultrasonic image as learning data, and adjusts the inference parameters including the weight of each feature vector in the above-mentioned ASM so that the contour or characteristic position of the heart shown in the ultrasonic image can be inferred. In addition, the inference parameters in this specification are parameters that are an adjustment object in the learning process of the learning model 30, and represent parameters that affect the inference result of the learning model 30 (that is, the output data of the learning model 30). The learning model 30 that has been fully learned can output the contour and characteristic position of the heart in the ultrasonic image with high accuracy based on the ultrasonic image showing the heart.

[0041] Moreover, if Figure 1 As shown, a parameter change history DB (DataBase) 32 is stored in the memory 28. The details of the parameter change history DB 32 and the method of using it will be described later together with the details of the inference parameter changing unit 36.

[0042] The inference processing unit 34 performs inference processing on a target medical image (target ultrasonic image in this embodiment) as a medical image to be inferred, and outputs the inference result. Specifically, the inference processing unit 34 inputs the target ultrasonic image to the learned learning model 30, and outputs the inference result on the target ultrasonic image based on the output data of the learning model 30 for the input data. In this embodiment, as described above, the inference processing unit 34 infers the contour line and characteristic position of the heart for the target ultrasonic image showing the heart, and outputs the inference result.

[0043] The display control unit 20 causes the display 22 to display the inference result of the inference processing unit 34 regarding the target ultrasonic image. Figure 2 FIG. 2 is a diagram showing a display example of the inference result related to the target ultrasonic image USI. Figure 2 The target ultrasonic image USI shown shows a four-chamber cross section of the heart, and shows a contour line OL and a characteristic position P of the heart (particularly, the left ventricle) as a result of inference related to the target ultrasonic image USI.

[0044] In addition, the control unit 26 may have a function as a measuring unit, and may perform measurement related to the heart shown in the target ultrasonic image USI based on at least one of the contour line OL and the characteristic position P inferred by the inference processing unit 34. For example, the control unit 26 may measure the end-diastolic long-axis length, the end-systolic long-axis length, the end-diastolic volume, the end-systolic volume, or the ejection fraction of the heart. The display control unit 20 may display the measurement result R of the control unit 26 together with the inference result of the inference processing unit 34 on the display 22.

[0045] The inference parameter changing unit 36 ​​changes the inference parameters of the learning model 30 according to the user's instructions. As described above, the inference parameters of the learning model 30 that performs inference related to the target ultrasonic image are adjusted by the learning process, but in this embodiment, the inference parameter changing unit 36 ​​can change at least a part of the inference parameters of the learning model 30 according to the user's instructions. In other words, the user can change at least a part of the inference parameters of the learning model 30. As described above, the inference parameter is a parameter that affects the inference result of the learning model 30, so if the inference parameter changing unit 36 ​​changes the inference parameter, the inference result of the learning model 30 will change.

[0046] Figure 32 is a diagram showing an example of a screen for changing inference parameters. In the present embodiment, the display control unit 20 displays a button B for the user to change the inference parameters together with the inference result of the inference processing unit 34 on the display 22 which is a touch panel. The user operates the button B to input an instruction to change the inference parameters of the learning model 30 to the inference parameter changing unit 36. The inference parameter changing unit 36 ​​changes the inference parameters of the learning model 30 according to the input change instruction.

[0047] The learning model 30 generally has a plurality of inference parameters, and it is sometimes difficult for the user to understand how which inference parameters will affect the inference result of the learning model 30. Therefore, it is preferable that inference parameter information that associates the inference parameters of the learning model 30 with the items of the inference result of the learning model 30 affected by the inference parameters is stored in the memory 28 in advance, and the display control unit 20 associates the item included in the inference parameter information with the button B for changing the inference parameter corresponding to the item and displays it on the display 22.

[0048] For example, in Figure 3 In the example of FIG. 1 , the display control unit 20 displays a button B for the item “smoothness (of the heart contour OL)”, a button B for the item “apex position”, and a button B for the item “outside / inside”. The buttons B for each item are composed of a button B for changing the inference parameter for that item in the first direction (for example, a decreasing direction) (in the Figure 3 In the example, a button B showing a left arrow is shown and a button B for changing the inference parameter for the item in a second direction (for example, an increasing direction) opposite to the first direction is shown. Figure 3 In the example of , a group structure of button B) showing a right arrow is shown.

[0049] Furthermore, the display control unit 20 preferably displays an index value (in Figure 3 In the example of , the index value of each item is "5"). The index value can take a value between 0 and 10, for example, and the value of each item determined by the learning process is set to the middle value "5".

[0050] For example, the inference parameter for the item "smoothness" is a weight set of each feature vector in the above-mentioned ASM. For example, the closer the user instructs the index value of the item "smoothness" to a value of "0", the more the inference parameter changing unit 36 ​​increases the weight of the high-level principal component and decreases the weight of the low-level principal component for the weight of each feature vector in the ASM, and the closer the user instructs the index value of the item "smoothness" to a value of "10", the more the inference parameter changing unit 36 ​​decreases the weight of the high-level principal component and increases the weight of the low-level principal component for the weight of each feature vector in the ASM.

[0051] Furthermore, the item "outside / inside" is an adjustment item for the position of the heart contour line OL. For example, the closer the index value of the item "outside / inside" is to "0", the closer the contour line OL is to the heart (inside the heart). Figure 3 In the example, the left ventricle moves inward in the direction of decreasing volume. The closer the index value of the item "outside / inside" is to "10", the more the contour line OL moves to the outside in the direction of increasing the volume of the heart.

[0052] Inference parameters are parameters that are adjusted through learning processing so that inference results related to medical images (ultrasound images in this embodiment) can be output with high accuracy. Therefore, if the inference parameters are changed significantly, the learning model 30 may not be able to output inference results related to medical images with high accuracy. Therefore, the inference parameter changing unit 36 ​​preferably limits the amount of changeable inference parameters based on user instructions to a specified amount so that the inference accuracy of the learning model 30 is maintained at a certain level of accuracy.

[0053] The inference processing unit 34 again inputs the target ultrasonic image to the learning model 30 having the changed inference parameters by the inference parameter changing unit 36. Then, the inference processing unit 34 outputs an inference result related to the target ultrasonic image based on the output data of the learning model 30 for the input data.

[0054] The display control unit 20 changes the display content of the inference result of the inference processing unit 34 on the display 22 according to the change of the inference parameter. Figure 4 2 is a diagram showing a display example of an inference result using the changed inference parameter. Thus, the user can make the inference result related to the medical image using the learning model 30 close to his / her own requirements (preferences).

[0055] Whenever the user operates button B and the inference processing unit 34 performs inference processing based on the changed inference parameters, the display control unit 20 preferably displays the inference result based on the changed inference parameters on the display 22. Thus, the user can easily obtain a preferred inference result while fine-tuning the inference parameters.

[0056] The user can also directly correct the inference result of the medical image (in this example, the contour line OL or the characteristic position P) without operating the button B. In this case, the inference parameter changing unit 36 ​​preferably changes the inference parameter of the learning model 30 according to the corrected inference result. For example, when the user manually corrects the shape of the contour line OL, the inference parameter changing unit 36 ​​quantifies the smoothness of the contour line OL and changes the inference parameter corresponding to the item "smoothness" (i.e., the weight of each feature vector in the ASM) according to the numerical value.

[0057] When the inference parameter is changed according to the user's instruction, the inference parameter changing unit 36 ​​preferably stores the changed inference parameter in the parameter change history DB 32. In particular, when the inference parameter is changed according to the user's instruction, the inference parameter changing unit 36 ​​preferably associates the user identifier that uniquely identifies the user who instructed the change of the inference parameter with the changed inference parameter and stores the associated inference parameter in the parameter change history DB 32.

[0058] Figure 5 32 is a conceptual diagram showing an example of the content of the parameter change history DB 32. In the present embodiment, the parameter change history DB 32 includes a user ID as a user identifier, an inference parameter ( Figure 3 or Figure 4 The item in ( ) and the index value indicated by the user for the item are associated with each other and stored. In addition, the user ID of the user can be obtained by the control unit 26 authenticating the user using a password or other method before the user starts using the ultrasonic diagnostic device 10.

[0059] According to the instruction from the user, the target ultrasonic image USI (reference Figure 3 or Figure 4 ) performs inference processing, the inference parameter changing unit 36 ​​can change the inference parameters of the learning model 30 according to the changed inference parameters stored in the parameter change history DB 32 without relying on the instruction from the user. For example, when the content of the parameter change history DB 32 is Figure 5 In the case of the content shown, the inference parameter changing unit 36 ​​does not rely on instructions from the user, in other words, automatically determines, for example, based on the index values ​​"3" and "5" for the item "smoothness" set by the user in the past, the inference parameter corresponding to the item "smoothness" of the learning model 30 in the current inference. For example, the inference parameter changing unit 36 ​​first calculates a representative value (for example, an average value) of a plurality of values ​​for the item "smoothness" stored in the parameter change history DB 32. Here, "4" is calculated as the average value of the plurality of values ​​"3" and "5" for the item "smoothness". Then, the inference parameter changing unit 36 ​​changes the inference parameter of the learning model 30 to a value corresponding to the calculated index value "4". The inference parameter changing unit 36 ​​performs the same processing on other items "apex position" and "outside / inside".

[0060] In particular, it is preferable that, when the user ID is associated with the changed inference parameter and stored in the parameter change history DB 32, when the target medical image is inferred according to the instruction from the user, the inference parameter change unit 36 ​​changes the inference parameter of the learning model 30 according to the changed inference parameter associated with the user ID of the user in the parameter change history DB 32, regardless of the instruction from the user. For example, when the content of the parameter change history DB 32 is Figure 5 In the case of the contents shown in FIG. 1 , when the target ultrasonic image USI is inferred according to the instruction from the user represented by the user ID “U0001”, the inference parameter changing unit 36 ​​changes the inference parameters of the learning model 30 to values ​​corresponding to the index value “3” for the item “smoothness”, the index value “8” for the item “apex position”, and the index value “8” for the item “outside / inside”, without depending on the instruction from the user. On the other hand, when the target ultrasonic image USI is inferred according to the instruction from the user represented by the user ID “U0002”, the inference parameter changing unit 36 ​​changes the inference parameters of the learning model 30 to values ​​corresponding to the index value “5” for the item “smoothness”, the index value “6” for the item “apex position”, and the index value “6” for the item “outside / inside”, without depending on the instruction from the user.

[0061] The schematic structure of the ultrasonic diagnostic apparatus 10 according to the present embodiment is as described above. Figure 6 The flowchart shown in FIG. 1 is used to explain the processing flow of the ultrasonic diagnostic apparatus 10. Figure 6 At the start time of the flowchart shown, the learning model 30 has been completed.

[0062] In step S10 , the control unit 26 authenticates the user and specifies the user ID of the user.

[0063] In step S12, the image forming unit 18 forms an ultrasonic tomographic image based on the reception beam signal subjected to the signal processing in the signal processing unit 16. It is assumed that a four-chamber cross section of the heart is shown in the ultrasonic tomographic image.

[0064] In step S14, the user sets the inference type. Here, it is assumed that the inference of the contour line and characteristic position of the heart is set.

[0065] In step S16 , the inference processing unit 34 determines a frame to be processed (target ultrasonic image).

[0066] In step S18, the inference parameter changing unit 36 ​​determines whether the data related to the user ID specified in step S10 is stored in the parameter change history DB 32. If the data related to the user ID is stored in the parameter change history DB 32, the process proceeds to step S20, and if not, the process skips step S20 and proceeds to step S22.

[0067] In step S20 , the inference parameter changing unit 36 ​​changes the inference parameter of the learning model 30 based on the changed inference parameter associated with the user ID of the user in the parameter change history DB 32 .

[0068] In step S22 , the inference processing unit 34 inputs the target ultrasonic image to the learning model 30 , and outputs an inference result related to the target ultrasonic image based on the output data of the learning model 30 for the input data.

[0069] In step S24 , the display control unit 20 causes the display 22 to display the inference result of the inference processing unit 34 regarding the target ultrasonic image.

[0070] In step S26, the inference parameter changing unit 36 ​​determines whether an instruction to change the inference parameters of the learning model 30 has been received from the user. If no instruction to change the inference parameters has been received, the process ends, and if an instruction to change the inference parameters has been received, the process proceeds to step S28.

[0071] In step S28, the inference parameter changing unit 36 ​​changes the inference parameters of the learning model 30 according to the user instruction received in step S26. At the same time, the inference parameter changing unit 36 ​​associates the user ID of the user with the changed inference parameters and stores them in the parameter change history DB 32.

[0072] In step S22 after step S28 , the inference processing unit 34 inputs the target ultrasonic image to the learning model 30 having the inference parameters changed in step S28 , and outputs an inference result related to the target ultrasonic image based on output data of the learning model 30 for the input data.

[0073] In step S24 again, the display control unit 20 changes the display content of the inference result of the inference processing unit 34 on the display 22 according to the change of the inference parameter.

[0074] Every time the user issues an instruction to change the inference parameter, the processing of step S22 to step S28 is repeated.

[0075] As mentioned above, although the embodiment according to the present invention has been described, the present invention is not limited to the above-mentioned embodiment, and various modifications can be made without departing from the gist of the present invention.

[0076] For example, in the above-mentioned embodiment, the medical image processing device is an ultrasonic diagnostic device 10, but the medical image processing device is not limited to the ultrasonic diagnostic device 10. For example, the medical image processing device may also be a personal computer (PC (Personal Computer)) or a server. In this case, the PC or server as a medical image processing device has a display control unit 20, a display 22, an input interface 24, a control unit 26, a memory 28, an inference processing unit 34, and an inference parameter change unit 36. The learning model 30 and the parameter change history DB 32 are stored in the memory 28. The PC or server as a medical image processing device acquires a target medical image from an X-ray device, a CT device, an MRI device, or an ultrasonic diagnostic device, and performs inference processing on the acquired target medical image.

Claims

1. A medical image processing device, characterized in that: have: A learning model is learned in such a way as to output output data representing a result of inference related to the medical image based on input data representing the medical image; an inference parameter changing unit that changes an inference parameter according to a user's instruction, the inference parameter being a parameter to be adjusted in a learning process of the learning model and being a parameter that affects the output data; and The inference processing unit outputs an inference result related to the target medical image based on output data when a target medical image as a medical image to be processed by inference is input to the learning model having the changed inference parameters.

2. The medical image processing device according to claim 1, characterized in that: The medical image processing apparatus further includes a display control unit that displays a target medical image and an inference result related to the target medical image on a display, and changes display content of the inference result on the display according to a change in the inference parameter.

3. The medical image processing device according to claim 2, characterized in that: When a user corrects the inference result related to the target medical image displayed on the display, the inference parameter changing unit changes the inference parameter according to the corrected inference result.

4. The medical image processing apparatus according to claim 1, characterized in that: The medical image processing apparatus further includes a parameter change history database in which the changed inference parameters changed according to a user's instruction are stored. When the target medical image is inferred according to an instruction from a user, the inference parameter changing unit changes the inference parameter according to the changed inference parameter stored in the parameter change history database without depending on the instruction from the user.

5. The medical image processing apparatus according to claim 4, characterized in that: In the parameter change history database, a user identifier for identifying a user who instructed a change in the inference parameter is associated with the changed inference parameter that has been changed according to the instruction of the user and stored. In a case where the inference parameters related to the user are stored in the parameter change history database, when the object medical image is inferred based on instructions from the user, the inference parameter changing unit changes the inference parameters based on the changed inference parameters associated with the user identifier of the user in the parameter change history database, without relying on the instructions from the user.

6. The medical image processing apparatus according to claim 1, characterized in that: The medical image is an ultrasonic image formed by irradiating ultrasonic waves toward the heart of the subject. The inference processing unit outputs information for measurement of the heart using the ultrasonic image as the target medical image as an inference result of the target medical image.

7. A computer-readable non-transitory recording medium having a medical image processing program recorded thereon, characterized in that: The medical image processing program enables the computer to function as the following parts: A learning model is learned in such a way as to output output data representing a result of inference related to the medical image based on input data representing the medical image; an inference parameter changing unit that changes an inference parameter according to a user's instruction, the inference parameter being a parameter to be adjusted in a learning process of the learning model and being a parameter that affects the output data; and The inference processing unit outputs an inference result related to the target medical image based on output data when a target medical image as a medical image to be processed by inference is input to the learning model having the changed inference parameters.

Images