Information processing system, information processing device, information processing method, and program

Abstract

The present invention provides a technique capable of generating a learned model from acquired measurement data and performing reasoning directly. According to one embodiment of the present invention, an information processing system is provided. The information processing system includes a learning unit and an inference unit. The learning unit generates a learned model by inputting the first reconstructed image and the second reconstructed image into the learning model such that an output when the first reconstructed image is input becomes the second reconstructed image. The first reconstructed image and the second reconstructed image are images reconstructed based on any of a plurality of projection images included in the measurement data, respectively. The reasoning unit obtains an output image, which is an output image, by inputting the third reconstructed image into the learned model. The third reconstructed image is an image reconstructed based on any of a plurality of projection images included in the measurement data.

Description

Technical Field

[0001] This invention relates to information processing systems, information processing devices, information processing methods, and programs. Background Technology

[0002] Patent Document 1 discloses an X-ray diagnostic device including a processing unit that processes a learned model generated by machine learning. The learned model takes second data corresponding to a second number of views as input data, which is obtained based on first data corresponding to a first number of views, and takes third data corresponding to a third number of views that is more than the second number of views as output data. Then, it inputs fourth data corresponding to a fourth number of views that is less than the first number of views into the learned model to improve the quality of the fourth data. Existing technical documents Patent documents

[0003] [Patent Document 1] Japanese Patent Publication No. 2023-58034 Summary of the Invention The problem the invention aims to solve

[0004] According to the invention described in Patent Document 1, the generation of the learned model uses data obtained from the first data, while the improvement in the quality of the reconstructed image data uses data obtained from a fourth data different from the first data. That is, the invention described in Patent Document 1 requires the preparation of measurement data that is different from the measurement data used in the reasoning to generate the learned model.

[0005] In view of the above, the present invention aims to provide a technique that can generate a learned model based on acquired measurement data and directly perform inference. Solution for solving the problem

[0006] According to one aspect of the present invention, an information processing system is provided. The information processing system includes a learning department and a reasoning department. The learning unit generates a learned model by inputting the first reconstructed image and the second reconstructed image into the learning model, such that the output when the first reconstructed image is used as input becomes the second reconstructed image. The first reconstructed image and the second reconstructed image are images reconstructed based on any one of the multiple projected images contained in the measurement data. The inference unit obtains an output image by inputting the third reconstructed image into the learned model. The third reconstructed image is an image reconstructed from any of the multiple projected images contained in the measurement data.

[0007] According to this disclosure, it is possible to generate a learned model from the acquired measurement data and perform inference directly. Attached Figure Description

[0008] Figure 1 This is a diagram illustrating an example of the system architecture and hardware architecture of information processing system 1. Figure 2 This is a diagram showing an example of the functional units possessed by the processor 21. Figure 3 This is a diagram representing an example of an activity performed by information processing system 1. Figure 4 This is a diagram used to illustrate the process of obtaining decomposed waveform data from waveform data. Figure 5 This is a diagram used to illustrate the method of determining the expiratory and inspiratory regions. Figure 6 It is a diagram used to illustrate the method of determining the exhalation and inhalation of the lungs and the diastolic and systolic phases of the heart. Figure 7 This is an example used to illustrate the generation of a learned model. Figure 8 This is an example of a graph used to illustrate the use of a learned model. Figure 9 This is an example diagram representing screen 4. Figure 10 This is an example image representing screen 5. Figure 11 It is a diagram representing an image of a cross-section containing a ferret's heart, reconstructed from all projected images contained in the measurement data. Figure 12 It means about and Figure 11 A diagram of a third reconstructed image, based on a projection image of the ferret's lungs characterized by inhalation, contained in the measurement data, using the same cross section. Figure 13 It means about and Figure 11 The same cross section, by Figure 12 The third reconstructed image is input into the learned model and the output image is plotted. Figure 14 It is a graph representing a cross-section of a ferret's lungs and diaphragm reconstructed from all the projected images contained in the measurement data. Figure 15 It means about and Figure 14A diagram of a third reconstructed image, reconstructed from a projection image of the ferret's lungs characterized by inhalation, contained in the measurement data, using the same cross section. Figure 16 It means about and Figure 14 The same cross section, by Figure 15 The third reconstructed image is input into the learned model and the output image is plotted. Detailed Implementation

[0009] [Example] The embodiments of the present invention will now be described in conjunction with the accompanying drawings. The various features shown in the following embodiments can be combined with each other.

[0010] Furthermore, in one embodiment, the program for implementing the software may be provided as a non-transitory computer-readable medium, or it may be provided by downloading from an external server, or it may be provided by launching the program on an external computer to implement its functionality on a client terminal (i.e., so-called cloud computing).

[0011] Furthermore, in various information processing methods of one embodiment, input and output based on the input can be implemented. Here, as long as the output can be obtained as the input result, the form of the information referenced in this information processing (hereinafter referred to as the reference information) is not restricted. Reference information can be, for example, rule-based information such as databases, lookup tables, predefined functions (including judgments constructed through statistical methods, such as regression equations), learned models that have pre-learned the correlation between inputs and outputs, or large language models that can output the desired results by inputting prompt words.

[0012] Furthermore, in one embodiment, "part" may include, for example, a combination of hardware resources implemented by generalized circuitry and information processing software specifically implemented through these hardware resources. In another embodiment, various types of information are processed, such as physical values ​​representing voltage and current, binary bits consisting of 0 or 1 as high or low signal values, or quantum superposition (i.e., so-called qubits), and communication and computation can be performed on a generalized circuit.

[0013] In addition, in a broader sense, a circuit refers to a circuit implemented by at least a proper combination of circuits, circuits, processors, and memory. In addition, the processor can be a general-purpose processor or a dedicated circuit. That is, including Application Specific Integrated Circuit (ASIC), Programmable Logic Device (e.g., Simple Programmable Logic Device (SPLD), Complex Programmable Logic Device (CPLD), and Field Programmable Gate Array (FPGA)).

[0014] 1. System structure and hardware structure of information processing system 1 First, refer to Figure 1 The system structure and hardware structure of the information processing system 1 in this embodiment are described. Figure 1 This is a diagram illustrating an example of the system architecture and hardware architecture of information processing system 1.

[0015] (Information Processing System 1) Figure 1 The information processing system 1 shown is capable of processing multiple projection images captured by the CT device 3. The information processing system 1 includes an information processing device 2 and a CT (Computed Tomography) device 3. The information processing device 2 and the CT device 3 communicate with each other via a communication cable or network. Therefore, the information processing device 2 and the CT device 3 can send or receive various types of information to each other. Here, the system illustrated in Information Processing System 1 is composed of one or more devices or components. Therefore, the information processing device 2 alone or the CT device 3 alone is also included in the system illustrated in the information processing system 1. The information processing device 2 and the CT device 3 are operated, for example, by the measurer, i.e., the user.

[0016] (Information processing device 2) Information processing device 2 is a PC (Personal Computer). The information processing device 2 can also be a tablet computer, smartphone, or other device that replaces the PC. The information processing device 2 can process multiple projection images captured by the CT device 3. Specifically, for example, the information processing device 2 is configured to perform arbitrary information processing on the measurement data acquired from the CT device 3, control the radiation generated from the radiation generator 34, acquire the projection image detected by the detector 35, control the movement of the sample holding part 36, and control the rotation drive part 37, etc. Note that the information processing device 2 only needs to be able to perform any information processing related to the CT device 3 as a result, and the information processing device 2 and the CT device 3 can also be connected by other information processing devices. like Figure 1 As shown, the information processing device 2 includes a processor 21, a storage unit 22, a communication unit 23, an input unit 24, and an output unit 25. These components are electrically connected through a communication bus inside the information processing device 2. Information processing device 2 performs the processing related to this embodiment.

[0017] The processor 21 performs processing and control of the overall operation related to the information processing device 2. Processor 21 is, for example, a central processing unit (CPU). The information processing performed by the program stored in the storage unit 22 is specifically implemented by the processor 21 of the hardware example, and can be executed as various functional units included in the processor 21. The processor 21 includes various functional units, for example, to implement the functions described later. Figure 3 The processing is shown. Note that processor 21 is not limited to a single processor; multiple processors 21 may be used depending on the function. In addition, it can also be a combination of these.

[0018] Storage unit 22 stores various information as defined in the above description. This can be implemented, for example, as a storage device such as a solid-state drive (SSD) that stores various programs related to the information processing device 2 executed by the processor 21, or as memory such as random access memory (RAM) that stores temporary necessary information (parameters, arrays, etc.) related to the operation of the program. The storage unit 22 stores various programs, variables, and data used by the processor 21 during program execution related to the information processing device 2, which are executed by the processor 21. Storage unit 22 can also be an example of a storage medium.

[0019] While the Communication Department 23 preferentially includes wired communication methods such as USB (Universal Serial Bus), IEEE (Institute of Electrical and Electronics Engineers) 1394, Thunderbolt (registered trademark), and wired LAN (Local Area Network) network communication, it may also include wireless LAN network communication, LTE (Long Term Evolution) / 3G (Third Generation Mobile Communication) / 4G (Fourth Generation Mobile Communication) / 5G (Fifth Generation Mobile Communication) and Bluetooth (registered trademark) communication as needed. That is, it is preferred to implement it as a collection of these multiple communication methods. That is, the information processing device 2 can communicate with the outside world in various ways through the communication unit 23.

[0020] The input unit 24 can be contained in the housing of the information processing device 2 or externally. For example, the input unit 24 can be integrated with the output unit 25 as a touch screen. If it is a touchscreen, users can perform input operations such as clicking and swiping. Of course, you can also use switches, mice, keyboards, etc. instead of touchscreens. That is, the input unit 24 receives input generated by the user's operation. The input is transmitted to the processor 21 as a command signal via the communication bus, and the processor 21 can perform predetermined control or calculation as needed.

[0021] The output unit 25 can function as a display device for the information processing device 2. The output unit 25 may be contained in the housing of the information processing device 2, or it may be external. The output unit 25 displays a user-operable graphical user interface (GUI) screen. This can be implemented, for example, by using display devices such as CRT displays, liquid crystal displays, organic EL displays, and plasma displays according to the type of information processing device 2.

[0022] (CT device 3) The CT device 3 is a device capable of irradiating a sample with radiation and obtaining a projected image of the sample from the amount of radiation transmitted. The CT device 3 may include, but is not limited to, a sample rotation type CT device that rotates the sample holding part 36, a gantry type CT device that rotates the radiation generator 34 and the detector 35 relative to the sample holding part 36, etc. The CT device 3 includes a processor 31, a storage unit 32, a communication unit 33, a radiation generator 34, a detector 35, a sample holder 36, and a rotation drive unit 37. These components are electrically connected through a communication bus inside the CT device 3. The CT device 3 performs the processing related to this embodiment. For information on the processor 31, storage unit 32 and communication unit 33 of the CT device 3, please refer to the processor 21, storage unit 22 and communication unit 23 of the information processing device 2.

[0023] The radiation generator 34 irradiates the area where the sample is located in the sample holder 36 with radiation. In addition, radiation can include any of the following: alpha rays, beta rays, gamma rays, X-rays, neutron rays, etc.

[0024] The detector 35 is configured to detect radiation passing through the sample disposed in the sample holder 36. The detected radiation is analyzed as measurement data by the information processing device 2. The measurement data was obtained by the CT device 3. The measurement data includes information representing the angle at which the image was taken and information about the projected image corresponding to that angle. The detector 35 can be a two-dimensional detector such as a CCD (Charge Coupled Device) or an imaging plate.

[0025] The sample holding part 36 is configured to hold the sample stage. The sample holding unit 36 ​​can move the sample stage in any direction according to the movement command generated by the processor 21 or the processor 31. The sample stage is configured to hold the sample.

[0026] The rotation drive unit 37 is configured to rotate the sample holding unit 36 ​​and / or the radiation generator 34 and the detector 35. The rotation drive unit 37 can also be configured to include a mechanism for adjusting the magnification of the projected image during shooting.

[0027] 2. Functional structure of processor 21 in information processing device 2 Figure 2This is a diagram showing an example of the functional units possessed by the processor 21. like Figure 2 As shown, an example of the information processing system 1, namely the processor 21 of the information processing device 2, includes an input processing unit 210, a data storage unit 211, an output processing unit 212, a condition setting unit 213, a feature acquisition unit 214, a synchronization processing unit 215, a phase recognition unit 216, a reconstruction unit 217, a learning unit 218, and an inference unit 219. As described above, the processing of software information stored in storage unit 22 is specifically implemented through hardware example, namely processor 21, and can be executed as each functional unit (step) included in processor 21. The processor 21 can execute at least the condition setting step, the feature acquisition step, the synchronization processing step, the phase recognition step, the reconstruction step, the learning step, and the inference step.

[0028] The input processing unit 210 receives or acquires various data input by the user through the input unit 24. In addition, the input processing unit 210 receives, accepts, or acquires various data from the CT device 3 via the communication unit 23.

[0029] Data storage unit 211 stores the acquired data to storage unit 22.

[0030] The output processing unit 212 sends various data to the CT device 3 via the communication unit 23. Furthermore, the output processing unit 212 is configured to control the display information displayed on the output unit 25. Note that the displayed information can be the visual information itself that is visible to the user, such as screens, images, icons, and text, or it can be the rendering information used to display screens, images, icons, and text on various terminals.

[0031] The condition setting unit 213 receives various condition settings.

[0032] The feature acquisition unit 214 calculates feature quantities from the projected image.

[0033] The synchronization processing unit 215 identifies multiple projected images that share common motion characteristics from multiple projected images.

[0034] The phase recognition unit 216 recognizes the projected image used to generate the reconstructed image.

[0035] The reconstruction section 217 reconstructs an image of the sample from multiple projection images.

[0036] The learning unit 218 performs the processing for generating the learned model.

[0037] The inference unit 219 performs processing to output the image using the learned model.

[0038] Details regarding the input processing unit 210, data storage unit 211, output processing unit 212, condition setting unit 213, feature acquisition unit 214, synchronization processing unit 215, phase recognition unit 216, reconstruction unit 217, learning unit 218, and inference unit 219 will be explained later.

[0039] 3. Operation flow of information processing system 1 Next, an example of preferred information processing performed by the information processing system 1 in this embodiment will be described. Figure 3 This is a diagram illustrating an example of an activity performed by Information Processing System 1. Note that this activity may include any exception handling not shown in the diagram. Exception handling includes interrupting the information processing or omitting any processing steps.

[0040] In this embodiment, the multiple projected images include at least a portion of the lungs or heart of the organism as the subject. Living things can include humans and animals.

[0041] (Activity A1) First, the input processing unit 210 receives the user's input settings for measurement conditions and the measurement start command from the CT device 3 (hereinafter referred to as the measurement start command) via the input unit 24. Measurement conditions include, for example, the number of shots taken, scanning speed, exposure time, and magnification of the captured images.

[0042] (Activity A2) Next, the output processing unit 212 sends the measurement conditions and measurement start command to the CT device 3 through the communication unit 23.

[0043] (Activity A3) Then, the processor 31 of the CT device 3 receives the measurement conditions and measurement start command from the information processing device 2 via the communication unit 33.

[0044] (Activity A4) Next, the CT device 3 acquires measurement data based on the received measurement conditions.

[0045] (Activity A5) Then, the processor 31 of the CT device 3 sends measurement data to the information processing device 2 via the communication unit 33.

[0046] (Activity A6) Subsequently, the input processing unit 210 receives measurement data from the CT device 3 via the communication unit 23.

[0047] (Activity A7) Next, the data storage unit 211 stores the acquired measurement data to the storage unit 22.

[0048] If it is not necessary to acquire new measurement data (e.g., to perform reconstruction using acquired measurement data), the information processing system 1 may omit the information processing of activities A1 to A7.

[0049] (Activity A8) Next, the output processing unit 212 displays the projected image on the output unit 25 as a user-visible screen 4. Details regarding screen 4 will be provided later. Figure 9 illustrate.

[0050] (Activity A9) Subsequently, the condition setting unit 213 receives the ROI range setting input through the input unit 24 on screen 4.

[0051] "ROI (Region of Interest)" refers to a local region in a projected image used to obtain feature values, also known as the region of interest. In this embodiment, the ROI is a region that includes at least a portion of the lungs and heart of an organism.

[0052] (Activity A10) Next, the feature acquisition unit 214 acquires waveform data based on the received ROI range and the projected image. For example, the feature acquisition unit 214 acquires feature values ​​for the input ROI range for each of the multiple projected images. The feature acquisition unit 214 acquires waveform data by plotting the ROI feature of each projected image on the vertical axis and plotting the number of frames shown in the projected image on the horizontal axis. The output processing unit 212 further displays the waveform data on the screen 4.

[0053] "Waveform data" refers to data obtained from multiple projection images and represented by characteristic quantities. In this embodiment, the waveform data is as follows: Figure 4 The waveform is shown below.

[0054] "Feature quantity" refers to the value obtained by integrating (accumulating) the intensity in the image. In this embodiment, the value obtained by integrating (accumulating) the intensity in the ROI is used as the feature quantity. The feature quantities are shown in Formula 1 below, where x represents the feature quantity of each of the N existing projected images. As an indicator of intensity, a brightness value can be used, but it is not limited to this.

[0055] [Formula 1] (x1,x W ,x3…,x N )

[0056] (Activity A11) Next, the condition setting unit 213 receives the user's input of hysteresis value and singular value selection information on the waveform data display screen 4 visible to the user via the input unit 24. In this way, users can observe projected images, waveform data, etc., while receiving input of condition settings for the projected images applied by singular spectrum analysis. In addition, users can obtain more suitable decomposed waveform data from the projected image based on any set hysteresis value.

[0057] The "hysteresis value" is the value of the Hankel matrix used in singular spectral analysis of specified characteristics. For example, the number of rows in a Hankel matrix is ​​represented by L, and the number of columns in a Hankel matrix is ​​represented by K. L equals the hysteresis value, and K is the number of projected images N minus L and 1, which can also be expressed as NL-1.

[0058] "Singular value selection information" refers to the information on which singular value to select from multiple singular values ​​obtained from a matrix. In this embodiment, the singular value selection information is the information that specifies which singular value to select from multiple singular values ​​obtained from the Hankel matrix to obtain the decomposed waveform data.

[0059] (Activity A12) Then, the feature acquisition unit 214 obtains the decomposed waveform data by applying singular spectrum analysis to the feature quantities of the waveform data based on the hysteresis value. More specifically, for example, the feature acquisition unit 214 obtains the Hankel matrix shown in Formula 2 by applying the Hankel transformation based on the feature quantity and hysteresis value of Formula 1. The feature acquisition unit 214 obtains multiple matrices (X1, X2, ..., Xr) as shown in Formula 3 by applying singular value decomposition to the acquired Hankel matrix. Each matrix is ​​defined according to the singular value i as shown in Formula 4. The feature acquisition unit 214 applies an inverse Hankel transform to the matrix corresponding to the largest singular value (i=1) among the singular values ​​obtained from singular spectrum analysis to obtain trend waveform data. Furthermore, the feature acquisition unit 214 applies an inverse Hankel transformation to the matrix represented by at least one singular value selected from the singular values ​​other than the largest singular value (corresponding to the singular value selected by the singular value selection information, such as i=2, etc.) obtained from the singular values ​​obtained by singular spectrum analysis, and obtains vibration waveform data. In this way, more suitable decomposed waveform data can be obtained from the projected image based on the set hysteresis and singular values.

[0060] [Formula 2] [Formula 3] X = X1 + X W +…+X r [Formula 4]

[0061] Here, refer to Figure 4 The decomposed waveform data is explained. Figure 4 This is a diagram used to illustrate the process of obtaining decomposed waveform data from waveform data. like Figure 4 As shown, waveform data, as decomposed waveform data, is decomposed into trend components (trend waveform data), vibration components (vibration waveform data), noise components, etc. Vibration waveform data and noise components can exist in more than one form. Waveform data is formed by summing the intensities of all the decomposed waveform data.

[0062] "Decomposed waveform data" refers to waveform data obtained by decomposing waveform data. Decomposing waveform data includes at least trend waveform data, vibration waveform data, and noise components.

[0063] "Trend waveform data" refers to waveform data that is decomposed as a component representing the trend of waveform data. Trend waveform data, for example, reflects relatively large behaviors such as the carcass or limbs of an organism. Trend waveform data is also known as the first component.

[0064] "Vibration waveform data" refers to waveform data that is decomposed into vibration components representing waveform data. Vibration waveform data, for example, reflects the behavior of organs such as the lungs or heart of an organism that undergo periodic movements. Vibration waveform data is obtained from the second largest singular value, also known as the second component.

[0065] In the activities A13 to A16 described below, the synchronization processing unit 215 identifies the various features of the subject contained in each of the multiple projected images based on the waveform shown in the decomposed waveform data (the waveform shown in at least one of the trend waveform data and the vibration waveform data). To illustrate this information processing, refer to Figures 5 to 7 This describes a method for determining the expiratory zone, inspiratory zone, expiration, inspiration, diastole, and systole using trend waveform data and vibration waveform data.

[0066] Figure 5 This is a diagram used to illustrate the method of determining the expiratory and inspiratory regions. Figure 5 It contains vibration waveform data. The vibration waveform data represents the vibration components with the intensity shown in the trend waveform data as the baseline (preset value is 0).

[0067] The "exhalation zone" refers to the area that can be considered as corresponding to the exhalation of an organism. In this embodiment, the exhalation region is the region where the vibration intensity shown in the vibration waveform data is less than 0. exist Figure 5 In the example, the exhalation region is the area where the vibration waveform data intensity is below 0 (the area around 40 to 70, 90 to 110, 150 to 180, etc.).

[0068] The "inhalation zone" refers to the area that can be considered as corresponding to the inhalation of an organism. In this embodiment, the intake region is the region with a vibration intensity higher than 0. exist Figure 5 In the example, the inhalation region is the region where the intensity of the vibration waveform data is higher than 0 (regions with frame numbers around 0 to 40, 70 to 90, 110 to 150, etc.). Note that a position with an intensity of 0 can be assigned to either the exhalation or inhalation zone.

[0069] Next, we will explain the process by Figure 5 The method determines the expiratory and inspiratory regions, as well as the diastolic and systolic phases of the heart. Figure 6 It is a diagram used to illustrate the method of determining the exhalation and inhalation of the lungs and the diastolic and systolic phases of the heart. exist Figure 6 In the middle, trend waveform data and data obtained by overlaying are displayed on top of the waveform data. Figure 5 The method determines the vibration waveform data reflected by the exhalation and inhalation regions. exist Figure 6In the diagram, solid lines represent waveform data, dashed lines represent trend waveform data, and dotted lines represent a combination of trend waveform data and vibration waveform data. In addition, Figure 6 The trend waveform data showed disturbances in breathing and heartbeat.

[0070] Exhalation refers to the location with the minimum feature value in waveform data corresponding to multiple exhalation regions, for example... Figure 6 The position indicated by the middle circle. Inhalation refers to the location in the waveform data corresponding to multiple inhalation regions where the feature quantity is the largest at each position, for example... Figure 6 The position indicated by the middle triangle. The diastolic phase refers to the location of the minimum value of a characteristic quantity at various positions in the waveform data corresponding to the expiratory region, for example... Figure 6 The position indicated by the middle rhombus. The systolic phase refers to the location of the maximum value of a characteristic quantity at various positions in the waveform data corresponding to the expiratory region, for example... Figure 6 The position indicated by the square.

[0071] (Activity A13) The synchronization processing unit 215 determines the region with an intensity lower than a predetermined value as the exhalation region from the vibration shown in the decomposed waveform data. The synchronous processing unit 215 identifies the projection images with the smallest feature quantity at each position in the waveform data corresponding to the multiple exhalation regions determined by the decomposed waveform data as projection images corresponding to the exhalation of the organism.

[0072] (Activity A14) Next, the synchronization processing unit 215 determines the region with an intensity higher than a predetermined value as the intake region from the vibration shown in the decomposed waveform data. The synchronous processing unit 215 identifies the projection images with the largest feature quantity at each position in the waveform data corresponding to the multiple inhalation regions determined by the decomposed waveform data as projection images corresponding to the inhalation of the organism.

[0073] (Activity A15) Then, the synchronization processing unit 215 identifies the projection image of the position corresponding to the exhalation region of the waveform data where the feature value is at its maximum as the projection image of the corresponding systolic phase of the heart.

[0074] (Activity A16) Next, the synchronization processing unit 215 identifies the projection image of the position where the feature quantity is at a minimum value corresponding to the expiratory region of the waveform data as the projection image of the corresponding diastolic phase of the heart. Note that the location corresponding to the exhalation zone refers to the location where the intensity of the vibration waveform data is less than the intensity of the trend waveform data, based on the intensity of the trend waveform data. Furthermore, at each position of the waveform data corresponding to the exhalation region, the projection image with the smallest feature quantity includes the projection image with the minimum feature quantity. Therefore, the projection image corresponding to exhalation can also be identified as the projection image corresponding to diastole.

[0075] Without the need for synchronized exhalation and inhalation, information processing system 1 can perform only the determination process in the processing of activities A13 to A14, omitting the other processes. In cases where synchronization of at least one of the systolic and diastolic phases is not required, information processing system 1 may omit the processing of at least one of activities A15 to A16.

[0076] Through activities A13 to A16, the projected images for lung and heart reconstruction can be more appropriately determined from the trend waveform data and vibration waveform data obtained from singular spectrum analysis. Furthermore, even in the event of noise or vibration, a clearer image can be reconstructed from the projected image.

[0077] Based on the user's instruction to confirm the synchronization result, the information processing system 1 can enter the setting input of ROI range (Activity A9), the setting input of hysteresis value and singular value selection information (Activity A11), and the information processing related to acquiring and synchronizing decomposed waveform data (Activity A12).

[0078] (Activity A17) After the information processing of activities A13 to A16 is completed, the output processing unit 212 can display screen 5 on the output unit 25. Details regarding screen 5 will be provided later. Figure 10 illustrate.

[0079] (Activity A18) Next, the phase recognition unit 216 assigns a reconstructed projection image for each reconstructed image for each feature of the subject identified by the synchronization processing unit 215. A reconstructed image is an image reconstructed from any of the multiple projected images contained in the measurement data. The first reconstructed image, the second reconstructed image, and the third reconstructed image are examples of reconstructed images, respectively. In this way, since projection images with common subject features can be used for their respective reconstructions, clearer first, second, and third reconstructed images can be output.

[0080] More specifically, for example, the phase recognition unit 216 selects multiple projection images containing the biological lung exhalation characteristics identified by the synchronization processing unit 215 as projection images for the reconstruction of the first reconstruction image in the checkbox of the phase setting area 51. The first reconstructed image serves as input data for generating the learned model and is used as training data. In this embodiment, the first reconstructed image can be reconstructed from multiple projection images containing the exhalation features of the lungs of an organism.

[0081] In addition, the phase recognition unit 216 selects multiple projection images containing the diastolic features of the biological heart, which are identified by the synchronization processing unit 215, as projection images for the reconstruction of the second reconstruction image in the checkbox of the phase setting area 51. The second reconstructed image serves as output data when generating the learned model, and is used as training data. In this embodiment, the second reconstructed image can be reconstructed from multiple projection images containing the diastolic features of the organism's heart. As described above, since the projection image corresponding to exhalation can also be identified as the projection image corresponding to diastole, the number of projection images used for the reconstruction of the first reconstruction image can be equal to the number of projection images used for the reconstruction of the second reconstruction image after reducing the number of projection images. In this way, since the multiple projected images used for the first reconstructed image are contained within the multiple projected images used for the second reconstructed image, the second reconstructed image can be reconstructed as an image less affected by noise or stripes than the first reconstructed image. Therefore, by using the first reconstructed image as input and the second reconstructed image as output, a learned model can be generated, thereby enabling the generation of a learned model that can improve image quality.

[0082] Furthermore, the multiple projected images used for the reconstruction of the first reconstructed image need not have an inclusion relationship equal to the number of multiple projected images used for the reconstruction of the second reconstructed image after the number of images is reduced, but can have a quantity relationship below the number of multiple projected images used for the reconstruction of the second reconstructed image. Even if the projected images used for the reconstruction of the reconstructed image do not have an inclusion relationship, the learned model can still learn the structure of the subject through alignment, as described later. Furthermore, when the projected image data used for the reconstruction of the first reconstructed image and the second reconstructed image are different, and the number of projected images of the first reconstructed image and the second reconstructed image is equal, the learned model can learn the structure of the subject excluding noise. Furthermore, when the number and amount of projected image data used for the reconstruction of the first and second reconstructed images are equal, the neural network, which serves as both the learning model and the learned model, can also use algorithms such as autoencoders, which are based on identity mappings and can realize both input and output, to perform information processing.

[0083] The projected images used for reconstruction of at least one of the first and second reconstructed images can be aligned. The targeting can be based on specific parts of an organism, such as the lungs, heart, diaphragm, blood vessels, and ribs. In this way, by aligning the projected images used to generate the learned model, a learned model that can improve image quality with higher accuracy can be generated.

[0084] In addition, the phase recognition unit 216 selects multiple projection images containing the lung inhalation characteristics of the organism identified by the synchronization processing unit 215 as projection images for reconstruction of the third reconstruction image in the checkbox of the phase setting area 51. The third reconstructed image is used as input data into the generated learned model. In this embodiment, the third reconstructed image can be reconstructed from multiple projection images that include inspiratory features of the lungs. In addition, the third reconstructed image can also be reconstructed from multiple projection images containing expiratory features of the lungs. Note that the projected image used for the reconstruction of the third reconstructed image can also be aligned.

[0085] In addition, the phase recognition unit 216 can receive selection operations for multiple checkboxes corresponding to the first reconstructed image, the second reconstructed image, and the third reconstructed image displayed in the phase setting area 51. The phase recognition unit 216 can assign projection images for the reconstruction of the first reconstructed image, the second reconstructed image, and the third reconstructed image based on the selection results of the checkboxes displayed in the phase setting area 51. In this way, the projected images used for the first, second, and third reconstructed images can be adjusted based on the user's actions on the checkboxes.

[0086] (Activity A19) Next, the condition setting unit 213 can receive the conditions for generating the learned model through the learning condition setting area 52. The conditions for generating a learned model can include at least one of the following hyperparameters: number of iterations (epochs), learning rate, batch size, stride, number of layers, number of units, loss function, optimizer, number of filters, and kernel size. In this way, learning conditions can be received for each situation, thereby generating the most suitable learned model.

[0087] The number of iterations refers to the number of times the image set selected from the first and second reconstructed images is used during learning.

[0088] The learning rate refers to the degree to which a parameter in a machine learning process is changed at one time. In this embodiment, by reducing the number of iterations while increasing the learning rate, the learning time can be shortened, and higher quality output images can be obtained.

[0089] Batch size refers to the amount of data that is segmented and processed at one time during learning of the first reconstructed image. Alternatively, the number of splits can be used instead of the batch size.

[0090] Stride refers to the stride of the filter as it moves across the first reconstructed image. In this embodiment, the larger the step size, the faster the learned model is generated, thereby reducing the information processing load.

[0091] The number of layers refers to the number of each of the input layer, intermediate layer, and output layer in a neural network. The number of layers when using DnCNN (Denoising Convolutional Neural Network) is: 1 input layer, 15 intermediate layers, and 1 output layer.

[0092] The number of units refers to the number of neurons contained in the input layer, intermediate layer, and output layer of a neural network.

[0093] The loss function is a function that quantifies the error between the predicted result and the actual value, and is used to evaluate the accuracy of the learned model. Functions can include mean squared error, mean absolute error, mean squared logarithmic error, cross-entropy error, etc.

[0094] An optimizer is a parameter tuning method used to minimize the value of a loss function.

[0095] The number of filters refers to the total number of filters.

[0096] The kernel size refers to the size of the filter.

[0097] (Activity A20) Then, the input processing unit 210 receives an instruction to generate a learned model by receiving operations such as pressing the learning execution button 53.

[0098] (Activity A21) Next, the reconstruction unit 217 outputs the first reconstructed image, the second reconstructed image, and the third reconstructed image based on the selection results of the checkboxes displayed in the phase setting area 51.

[0099] (Activity A22) Subsequently, Learning Department 218 generates the learned model based on the conditions for generating the learned model received in Activity A19. Figure 7 This is an example used to illustrate the graph generated by the learned model. like Figure 7 As shown, the learned model inputs the first reconstructed image and the second reconstructed image into the learning model, so that the output when the first reconstructed image is used as input becomes the second reconstructed image. For example, learned models and already learned models can be generated through neural networks. The neural network in this embodiment uses, for example, a CNN (Convolutional Neural Network) such as DnCNN (Denoising Convolutional Neural Network). A learning model refers to a model before it has completed learning, given that its output is a second reconstructed image when the first reconstructed image is taken as input.

[0100] The generation of the learned model and the information processing related to the output image from the learned model can be performed in the information processing device 2, or in other information processing devices such as a server.

[0101] (Activity A23) Next, the input processing unit 210 receives an instruction to obtain an output image using the learned model by receiving operations such as pressing the inference execution button 54.

[0102] (Activity A24) Next, the inference unit 219 inputs the third reconstructed image into the learned model and obtains the output image as the output image. The output processing unit 212 displays the output image on the output unit 25. Figure 8 This is an example of a graph used to illustrate the use of a learned model. like Figure 8 As shown, the learned model outputs an output image based on the input of the third reconstructed image. The output image is a third reconstructed image in which at least one of the following effects—noise, stripes, blur, and spatial resolution—is reduced. Noise can include the concept of artifacts. Therefore, by inputting a third reconstructed image, which has the same number of projected images as the first reconstructed image, into the generated learned model, the image quality of the third reconstructed image can be improved.

[0103] 4. Screenshot Example Next, we will refer to Figure 9 and Figure 10 The following explanation will be given for scenes 4 and 5, which were previously omitted in detail.

[0104] (Screen 4) Figure 9 This is an example diagram representing screen 4. Screen 4 displays measurement data-related information in a user-visible manner. Screen 4 includes a projected image display area 40, a waveform display area 41, a conditional reception area 42, a read button 43, a synchronous execution button 44, and a reconstruction execution button 45.

[0105] The projected image display area 40 is the area where the acquired projected image is displayed. To set the ROI, the projected image in the projected image display area 40 can specify the X-axis and Y-axis. The projected image display area 40 includes the ROI setting area 400. The ROI setting area 400 displays the ROI region in the projected image. The ROI range can be set through any operation, such as dragging and dropping the ROI setting area 400, changing the cursor range of the ROI setting area 400, or inputting values ​​into the condition receiving area 42 (described later). Condition setting unit 213 performs as Figure 3 The ROI range setting for Activity A9.

[0106] Waveform display area 41 is the area for displaying waveform data and / or decomposing waveform data. More specifically, for example, waveform display area 41 can be an area that displays waveform data before trend waveform data is acquired, and displays waveform data and trend waveform data overlaid after trend waveform data is acquired. In addition, the waveform display area 41 can also be an area that displays at least two of the following side by side: waveform data, trend waveform data, and vibration waveform data, for comparison.

[0107] Conditional reception area 42 is an area used to receive inputs related to waveform data singular spectrum analysis settings. The conditional acceptance region 42 includes the ROI setting region 420, the hysteresis setting region 421, and the singular value setting region 422.

[0108] The ROI setting area 420 is configured to allow setting the range from which waveform data is acquired from the projected image as the ROI, for example, such as Figure 9 As shown, the ROI is set by specifying the X-axis and Y-axis coordinates in the projected image display area 40.

[0109] The hysteresis setting region 421 is configured to allow setting hysteresis values ​​for specifying the singular spectral analysis matrix, for example, such as... Figure 9 As shown, you can input a lag value such as "100".

[0110] The singular value setting area 422 is configured to allow setting which vibration component's singular value to use to acquire vibration waveform data, for example, such as... Figure 9 The area shown is the area displaying the checkboxes for determining the singular values ​​of the vibrational components used when the singular values ​​associated with the vibrational components are arranged from largest to smallest (e.g., 1, 2, 3, etc.).

[0111] The read button 43 is used to read measurement data and display the projected image in the projected image display area 40. For example, the output processing unit 212 executes the operation based on the pressing of the read button 43. Figure 3 Activity A8.

[0112] The synchronous execution button 44 is used to perform synchronous processing. Based on the pressing operation of the synchronous execution button 44, for example, the feature acquisition unit 214 starts... Figure 3 Activity A12, followed by the synchronization processing unit 215. Figure 3 Activities A12 to A16. Figure 3 After the information processing of activities A12 to A16 is completed, the output processing unit 212 can... Figure 10 Screen 5 is shown.

[0113] (Screen 5) Figure 10 This is an example image representing screen 5. Screen 5 shows the generation and use of the learned model. Screen 5 includes a feature waveform area 50, a phase setting area 51, a learning condition setting area 52, a learning execution button 53, and an inference execution button 54.

[0114] Feature waveform area 50 displays the various features of the subject identified by the synchronous processing unit 215 on the waveform data display. Figure 10 The example shows the regions where data is marked during exhalation, inspiration, and diastole.

[0115] Phase setting area 51 contains multiple checkboxes. Multiple checkboxes can be set to select whether to use them for the reconstruction of the first, second, and third reconstructed images for each projected image. exist Figure 10 In the example, regarding phase setting region 51, the projection image with frame number 8602 is selected for the reconstruction of the first reconstructed image, the projection images with frames number 8602, 8607, and 8612 are selected for the reconstruction of the second reconstructed image, and the projection image with frame number 8604 is selected for the reconstruction of the third reconstructed image. The phase recognition unit 216 performs operations based on the phase setting area 51. Figure 3 Activity A17.

[0116] Learning condition setting area 52 is the area where learning conditions can be set. The learning condition setting area 52 may include at least one of the following: model selection area 520, model detailed settings button 521, iteration number area 522, learning rate area 523, and batch size area 524. exist Figure 10 The learning condition setting area 52 includes the iteration number area 522, the learning rate area 523, and the batch size area 524, which are used as learning conditions for the neural network. However, the displayed area can also be changed according to the model specified in the model selection area 520. The condition setting unit 213 executes the operation based on the operation of the learning condition setting area 52. Figure 3 Activity A18.

[0117] The model selection region 520 is configured to select models for learning such as CNN.

[0118] The Model Detailed Settings button 521 is used to further set detailed conditions for each model. For example, based on the pressing operation of the model detailed setting button 521, the output processing unit 212 can further display a screen that allows setting the stride, number of layers, and number of units.

[0119] The iteration count range 522 is configured to allow setting the number of iterations.

[0120] The learning rate range 523 allows you to set the learning rate.

[0121] The batch size range of 524 allows you to set the batch size.

[0122] The learning execution button 53 is used to start generating the learned model according to the conditions set in the learning condition setting area 52.

[0123] The inference execution button 54 is a button used to generate an output image using the learned model generated based on the pressing operation of the learning execution button 53.

[0124] 5. Examples Next, we will refer to Figures 11 to 16 This document describes an embodiment illustrating the changes in reconstructed images before and after synchronization. The subject in this embodiment is a ferret.

[0125] First, refer to Figures 11 to 13 This describes the situation with and without the application of this disclosure for a cross-section containing the heart of a ferret. Figure 11 It represents an image of a cross-section containing the ferret's heart, reconstructed from all projected images contained in the measurement data. Figure 12 It means about and Figure 11 The same cross section, a third reconstructed image reconstructed from a projection image of the ferret's lungs characterized by inhalation contained in the measurement data. Figure 13 It means about and Figure 11 The same cross section, by Figure 12 The third reconstructed image is input into the output image of the learned model. By comparison Figure 11 and Figure 13 It can be seen that, in Figure 11 In the reconstruction, because all projected images were used, the boundary between the heart and space was difficult to distinguish clearly, and the outline of the heart became blurred, resulting in an unclear reconstructed image. On the other hand, Figure 13 In China, for the sake of Figure 11 In similar locations, the boundary between the heart and the space becomes relatively easier to identify. In addition, through comparison Figure 12 and Figure 13 It can be seen that, in Figure 12 Overall, stripes and noise are visible, but... Figure 13 In the middle, stripes and noise are significantly reduced.

[0126] Next, refer to Figures 14 to 16 Explanation and Figures 11 to 13 Different cross-sections, including a section of the ferret's lungs and diaphragm, in the case of applying this disclosure and not applying this disclosure. Figure 14It represents an image reconstructed from all projected images contained in the measurement data, showing a cross-section of the ferret's lungs and diaphragm. Figure 15 It means about and Figure 14 The same cross section, a third reconstructed image reconstructed from a projection image of the ferret's lungs characterized by inhalation contained in the measurement data. Figure 16 It means about and Figure 14 The same cross section, by Figure 15 The third reconstructed image is input into the learned model and the output image is generated. By comparison Figure 14 and Figure 16 It can be seen that, in Figure 14 In the reconstruction, because all projected images were used, the boundary between the diaphragm and space was difficult to distinguish clearly, and the outline of the diaphragm became blurred, resulting in unclear reconstructed images. On the other hand, Figure 16 In China, for the sake of Figure 14 In similar locations, the boundary between the diaphragm and the space becomes relatively easier to identify. In addition, through comparison Figure 15 and Figure 16 It can be seen that, in Figure 15 Overall, stripes and noise are visible, but... Figure 16 In the middle, stripes and noise are significantly reduced.

[0127] According to this disclosure, a learned model can be generated from the acquired measurement data and inference can be performed directly.

[0128] [other] A program is a procedure that enables one or more computers to execute various functional units (steps). Alternatively, it could be an information processing method executed by the information processing system 1 (or the processor 21 of the information processing device 2).

[0129] In the embodiments, the case of user inputting a lag value is illustrated, but in the variations, the optimal value can be automatically calculated and used. That is, the feature acquisition unit 214 calculates the period of the frequency containing the maximum peak value by applying frequency analysis to the waveform data. The feature acquisition unit 214 acquires the value of the minimum amount of data included in the calculated period as the setting of the hysteresis value. In this way, more suitable decomposed waveform data can be obtained from the projected image based on the calculated hysteresis value. Furthermore, by calculating and setting a reasonable lag value, users can be spared the trouble of exploring the lag value.

[0130] In one embodiment, the subject of the projected image is the lungs or heart of an organism, but in a variation, other moving objects may also be included as the subject. For example, in a variation, the subject of the projected image may include biological parts such as the head and abdomen of an organism, artificial lungs and hearts, or objects used in industrial applications such as working mechanical parts (e.g., operational testing of mechanical parts such as gears and bearings) and fluids (e.g., observing fluid dynamics within a container).

[0131] In the embodiments, the predetermined analysis method is singular spectral analysis, and an example of applying Hankel matrix to singular value decomposition as a singular spectral analysis is illustrated. In a variation, the predetermined analysis method can be any method that can obtain decomposed waveform data from waveform data, or it can be singular spectrum analysis using the Toeplitz matrix or other methods.

[0132] In addition, multiple projected images with the same motion characteristics of the subject can be identified without using feature quantities. For example, the synchronous processing unit 215 can use other devices besides the CT device 3, and identify multiple projected images of the subject with the same motion characteristics based on the data obtained from the device. The device may include an electrocardiograph, an X-ray device, a PET (Positron Emission Tomography) device, an MRI (Magnetic Resonance Imaging) device, an ultrasound diagnostic device, a bone densitometer, a heart rate monitor, a blood pressure monitor, an endoscope, a displacement meter, etc.

[0133] The first reconstructed image, the second reconstructed image, and the third reconstructed image are not limited to the cases described in the embodiments, and can be reconstructed from multiple projected images containing motion features of any subject.

[0134] The embodiments illustrate the use of neural networks as learning models and learned models, but are not limited thereto. The learning model and the learned model can be models built through learning methods such as supervised learning, unsupervised learning, or self-supervised learning.

[0135] As specific algorithms for machine learning that learn models and already learned models, in addition to neural networks, deep learning methods such as nearest neighbor, naive Bayes, decision trees, and support vector machines can also be applied as needed.

[0136] The learning model and the learned model can include generative AI, such as AI (Artificial Intelligence) that includes Transformers such as GPT (Generative Pretrained Transformer, including GPT-3.5, GPT-4, etc.), BERT model (Bidirectional Encoder Representations from Transformers), BART model (Bidirectional and Auto-regressive Transformer) or language models such as Recurrent Neural Network (RNN).

[0137] Learning models and learned models, as artificial intelligence, can also refer to general natural language processing learning models such as large language models (LLMs) that have learned from massive amounts of data. Such learning models can include language models that can handle various tasks without fine-tuning, such as one-shot learning or few-shot learning.

[0138] Both learned and already learned models can use Generative Adversarial Networks (GANs). In addition, learned models can also include algorithms for any generative model such as DCGAN (Deep Convolutional GAN), Conditional GAN, Pix2Pix, PGGAN (Progressive Growing GAN), and StyleGAN.

[0139] This disclosure can also be applied to conditions where biological rhythms change during the measurement process, such as when photographing a biological organism on a heated bed, when anesthesia on the organism begins to take effect, or when anesthesia on the organism begins to wear off.

[0140] In addition, it can also be provided through the various methods described below.

[0141] (1) An information processing system, comprising: a learning unit and an inference unit; the learning unit generates a learned model by inputting a first reconstructed image and a second reconstructed image into a learning model, such that the output when the first reconstructed image is used as input becomes the second reconstructed image; the first reconstructed image and the second reconstructed image are respectively images reconstructed based on any one of a plurality of projected images contained in measurement data; the inference unit obtains an output image as an output image by inputting a third reconstructed image into the learned model; the third reconstructed image is an image reconstructed based on any one of the plurality of projected images contained in the measurement data.

[0142] In this way, a learned model can be generated from the acquired measurement data and inference can be performed directly.

[0143] (2) The information processing system as described in (1) above, wherein the number of multiple projected images used for the reconstruction of the first reconstructed image is less than the number of multiple projected images used for the reconstruction of the second reconstructed image.

[0144] In this way, by training the learned model with the number of projected images of the first reconstructed image as input being less than the number of projected images of the second reconstructed image as output, the learned model can learn the structure of the subject.

[0145] (3) The information processing system as described in (1) or (2) above, further comprising a condition setting unit, wherein the condition setting unit receives the setting of at least one of the following conditions when generating the learned model: number of iterations, learning rate, batch size, stride, number of layers, number of units, loss function, optimizer, convolution kernel size and number of filters, wherein the learned model is generated by a neural network.

[0146] In this way, learning conditions can be received for each situation, thereby generating the most suitable learned model and outputting a third reconstructed image.

[0147] (4) The information processing system as described in any one of (1) to (3) above, further comprising a feature acquisition unit and a synchronization processing unit, wherein the feature acquisition unit acquires waveform data as feature quantities obtained by each of the plurality of projected images included in the measurement data, and decomposes the waveform data to obtain decomposed waveform data, and the synchronization processing unit identifies the respective features of the subjects included in each of the plurality of projected images based on the waveforms shown by the decomposed waveform data.

[0148] In this way, even in the presence of factors such as noise or vibration, the characteristics of the subject can be identified more accurately.

[0149] (5) The information processing system as described in (4) above, further comprising a phase recognition unit, wherein the phase recognition unit assigns a projected image for reconstruction of the first reconstructed image, the second reconstructed image and the third reconstructed image for each feature of the subject.

[0150] In this way, because each reconstruction is performed using projected images with common subject features, clearer first, second, and third reconstructed images can be output.

[0151] (6) The information processing system as described in any one of (1) to (5) above, further comprising a phase recognition unit, wherein the phase recognition unit allocates projection images for the reconstruction of the first reconstructed image, the second reconstructed image and the third reconstructed image based on the selection results of a plurality of check boxes, wherein the plurality of check boxes are configured to allow selection of whether to use the plurality of projection images for the reconstruction of the first reconstructed image, the second reconstructed image and the third reconstructed image respectively.

[0152] In this way, the projected images used for the first, second, and third reconstructed images can be adjusted according to the user's instructions.

[0153] (7) The information processing system as described in any one of (1) to (6) above, wherein the number of multiple projected images used for the reconstruction of the first reconstructed image is equal to the number of multiple projected images used for the reconstruction of the second reconstructed image after reduction.

[0154] In this way, since the multiple projected images used for the first reconstructed image are contained within the multiple projected images used for the second reconstructed image, a learned model that can improve image quality with higher accuracy can be generated.

[0155] (8) The information processing system as described in any one of (1) to (7) above, wherein the projected images for reconstruction of at least one of the first reconstructed image and the second reconstructed image are aligned.

[0156] In this way, by aligning the projected images used to generate the learned model, a learned model that can improve image quality with higher accuracy can be generated.

[0157] (9) The information processing system as described in any one of (1) to (8) above, wherein the first reconstructed image is reconstructed from a plurality of projection images characterized by exhalation of the lungs of an organism, the second reconstructed image is reconstructed from a plurality of projection images characterized by diastole of the heart of the organism, and the third reconstructed image is reconstructed from a plurality of projection images characterized by exhalation or inhalation of the lungs.

[0158] In this way, by using the first and second reconstructed images as training data, the image quality of a third reconstructed image containing the exhalation or inhalation features of an organism's lungs can be improved.

[0159] (10) An information processing apparatus, comprising: a learning unit and an inference unit; the learning unit generates a learned model by inputting a first reconstructed image and a second reconstructed image into a learning model, such that the output when the first reconstructed image is used as input becomes the second reconstructed image; the first reconstructed image and the second reconstructed image are respectively images reconstructed based on any one of a plurality of projected images contained in measurement data; the inference unit obtains an output image by inputting a third reconstructed image into the learned model; the third reconstructed image is an image reconstructed based on any one of the plurality of projected images contained in the measurement data.

[0160] (11) An information processing method, comprising: causing a learning unit to generate a learned model by inputting a first reconstructed image and a second reconstructed image into a learning model, such that the output when the first reconstructed image is used as input becomes the second reconstructed image; the first reconstructed image and the second reconstructed image are respectively images reconstructed based on any one of a plurality of projected images contained in measurement data; causing an inference unit to obtain an output image by inputting a third reconstructed image into the learned model; the third reconstructed image is an image reconstructed based on any one of the plurality of projected images contained in the measurement data.

[0161] (12) A program that enables a computer to function as a functional part of the information processing system described in any one of (1) to (9) above. Of course, it's not limited to this.

[0162] Finally, although various embodiments related to the present invention have been described, these are provided by way of example only and are not intended to limit the scope of the invention. This new embodiment can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and variations thereof are included within the scope and spirit of the invention, and are included within the scope of the invention described in the claims and its equivalents. Explanation of reference numerals in the attached figures

[0163] 1: Information processing system 2: Information processing device 21: Processor 210: Input Processing Unit 211: Data Storage Department 212: Output Processing Unit 213: Condition Setting Department 214: Feature Acquisition Department 215: Synchronization Processing Department 216: Phase recognition unit, 217: Reconstruction Department 218: Study Department 219: Reasoning Department, 22: Storage Department 23: Ministry of Communications, 24: Input Department 25: Output section 3: CT scanner, 31: Processor 32: Storage Department 33: Ministry of Communications, 34: Radiation generator 35: Detector 36: Sample holding section 37: Rotary drive unit 4: Screen, 40: Projected image display area 400: ROI setting area 41: Waveform display area 42: Conditional Acceptance Range 420: ROI setting area 421: Delayed setting of region, 422: Singularity Setting Region 43: Read button 44: Synchronous Execution Button 45: Rebuild the execute button. 5: Screen, 50: Characteristic waveform region 51: Phase setting area 52: Learning conditions setting area, 520: Model selection area 521: Model detailed settings button, 522: Iteration count range 523: Learning rate range 524: Batch size range 53: Learn to execute buttons, 54: Reasoning Execution Button.

Claims

1. An information processing system, comprising: Study Club and Reasoning Club; The learning unit generates a learned model by inputting the first reconstructed image and the second reconstructed image into the learning model, such that the output when the first reconstructed image is used as input becomes the second reconstructed image. The first reconstructed image and the second reconstructed image are images reconstructed based on any one of the multiple projected images contained in the measurement data; The inference unit obtains an output image as the output image by inputting the third reconstructed image into the learned model; The third reconstructed image is an image reconstructed based on any of the plurality of projected images contained in the measurement data.

2. The information processing system according to claim 1, wherein, The number of multiple projected images used for the reconstruction of the first reconstructed image is less than the number of multiple projected images used for the reconstruction of the second reconstructed image.

3. The information processing system according to claim 1 or 2, wherein, It also includes a condition setting section. The condition setting unit receives settings for at least one of the following conditions when generating the learned model: number of iterations, learning rate, batch size, stride, number of layers, number of units, loss function, optimizer, kernel size, and number of filters. The learned model is generated through a neural network.

4. The information processing system according to any one of claims 1 to 3, wherein, It also includes a feature acquisition unit and a synchronization processing unit. The feature acquisition unit acquires waveform data as feature quantities obtained from each of the multiple projected images included in the measurement data, and decomposes the waveform data to obtain decomposed waveform data. The synchronization processing unit identifies the respective features of the subjects contained in each of the multiple projected images based on the waveforms shown in the decomposed waveform data.

5. The information processing system according to claim 4, wherein, It also includes a phase recognition unit, The phase recognition unit assigns a projected image for reconstruction of the first reconstructed image, the second reconstructed image, and the third reconstructed image for each feature of the subject.

6. The information processing system according to any one of claims 1 to 5, wherein, It also includes a phase recognition unit, The phase recognition unit assigns projection images for the reconstruction of the first reconstructed image, the second reconstructed image, and the third reconstructed image based on the selection results of multiple checkboxes. The multiple checkboxes are configured to allow selection of whether to use the multiple projected images for the reconstruction of the first reconstructed image, the second reconstructed image, and the third reconstructed image, respectively.

7. The information processing system according to any one of claims 1 to 6, wherein, The number of multiple projected images used for the reconstruction of the first reconstructed image is equal to the number of multiple projected images used for the reconstruction of the second reconstructed image after reduction.

8. The information processing system according to any one of claims 1 to 7, wherein, Align the projected images used for the reconstruction of at least one of the first reconstructed image and the second reconstructed image.

9. The information processing system according to any one of claims 1 to 8, wherein, The first reconstructed image is reconstructed from multiple projection images featuring the exhalation of an organism's lungs. The second reconstructed image is reconstructed from multiple projected images characterized by the diastolic phase of the organism's heart. The third reconstructed image is reconstructed from multiple projection images characterized by exhalation or inhalation of the lungs.

10. An information processing apparatus, comprising: Study Club and Reasoning Club; The learning unit generates a learned model by inputting the first reconstructed image and the second reconstructed image into the learning model, such that the output when the first reconstructed image is used as input becomes the second reconstructed image. The first reconstructed image and the second reconstructed image are images reconstructed based on any one of the multiple projected images contained in the measurement data; The inference unit obtains an output image by inputting a third reconstructed image into the learned model; The third reconstructed image is an image reconstructed based on any of the plurality of projected images contained in the measurement data.

11. An information processing method, comprising: The learning unit generates a learned model by inputting the first reconstructed image and the second reconstructed image into the learning model, such that the output when the first reconstructed image is used as input becomes the second reconstructed image; The first reconstructed image and the second reconstructed image are images reconstructed based on any one of the multiple projected images contained in the measurement data; The inference unit obtains an output image by inputting a third reconstructed image into the learned model; The third reconstructed image is an image reconstructed based on any of the plurality of projected images contained in the measurement data.

12. A program, The computer is made to function as a functional part of the information processing system according to any one of claims 1 to 9.

Images